Index: stable/10/sys/cam/ata/ata_da.c =================================================================== --- stable/10/sys/cam/ata/ata_da.c (revision 271237) +++ stable/10/sys/cam/ata/ata_da.c (revision 271238) @@ -1,2095 +1,2096 @@ /*- * Copyright (c) 2009 Alexander Motin * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer, * without modification, immediately at the beginning of the file. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include "opt_ada.h" #include #ifdef _KERNEL #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #endif /* _KERNEL */ #ifndef _KERNEL #include #include #endif /* _KERNEL */ #include #include #include #include #include #include #include /* geometry translation */ #ifdef _KERNEL #define ATA_MAX_28BIT_LBA 268435455UL typedef enum { ADA_STATE_RAHEAD, ADA_STATE_WCACHE, ADA_STATE_NORMAL } ada_state; typedef enum { ADA_FLAG_CAN_48BIT = 0x0002, ADA_FLAG_CAN_FLUSHCACHE = 0x0004, ADA_FLAG_CAN_NCQ = 0x0008, ADA_FLAG_CAN_DMA = 0x0010, ADA_FLAG_NEED_OTAG = 0x0020, ADA_FLAG_WAS_OTAG = 0x0040, ADA_FLAG_CAN_TRIM = 0x0080, ADA_FLAG_OPEN = 0x0100, ADA_FLAG_SCTX_INIT = 0x0200, ADA_FLAG_CAN_CFA = 0x0400, ADA_FLAG_CAN_POWERMGT = 0x0800, ADA_FLAG_CAN_DMA48 = 0x1000, ADA_FLAG_DIRTY = 0x2000 } ada_flags; typedef enum { ADA_Q_NONE = 0x00, ADA_Q_4K = 0x01, } ada_quirks; #define ADA_Q_BIT_STRING \ "\020" \ "\0014K" typedef enum { ADA_CCB_RAHEAD = 0x01, ADA_CCB_WCACHE = 0x02, ADA_CCB_BUFFER_IO = 0x03, ADA_CCB_DUMP = 0x05, ADA_CCB_TRIM = 0x06, ADA_CCB_TYPE_MASK = 0x0F, } ada_ccb_state; /* Offsets into our private area for storing information */ #define ccb_state ppriv_field0 #define ccb_bp ppriv_ptr1 struct disk_params { u_int8_t heads; u_int8_t secs_per_track; u_int32_t cylinders; u_int32_t secsize; /* Number of bytes/logical sector */ u_int64_t sectors; /* Total number sectors */ }; #define TRIM_MAX_BLOCKS 8 #define TRIM_MAX_RANGES (TRIM_MAX_BLOCKS * ATA_DSM_BLK_RANGES) struct trim_request { uint8_t data[TRIM_MAX_RANGES * ATA_DSM_RANGE_SIZE]; TAILQ_HEAD(, bio) bps; }; struct ada_softc { struct bio_queue_head bio_queue; struct bio_queue_head trim_queue; int outstanding_cmds; /* Number of active commands */ int refcount; /* Active xpt_action() calls */ ada_state state; ada_flags flags; ada_quirks quirks; int sort_io_queue; int trim_max_ranges; int trim_running; int read_ahead; int write_cache; #ifdef ADA_TEST_FAILURE int force_read_error; int force_write_error; int periodic_read_error; int periodic_read_count; #endif struct disk_params params; struct disk *disk; struct task sysctl_task; struct sysctl_ctx_list sysctl_ctx; struct sysctl_oid *sysctl_tree; struct callout sendordered_c; struct trim_request trim_req; }; struct ada_quirk_entry { struct scsi_inquiry_pattern inq_pat; ada_quirks quirks; }; static struct ada_quirk_entry ada_quirk_table[] = { { /* Hitachi Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "Hitachi H??????????E3*", "*" }, /*quirks*/ADA_Q_4K }, { /* Samsung Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "SAMSUNG HD155UI*", "*" }, /*quirks*/ADA_Q_4K }, { /* Samsung Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "SAMSUNG HD204UI*", "*" }, /*quirks*/ADA_Q_4K }, { /* Seagate Barracuda Green Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "ST????DL*", "*" }, /*quirks*/ADA_Q_4K }, { /* Seagate Barracuda Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "ST???DM*", "*" }, /*quirks*/ADA_Q_4K }, { /* Seagate Barracuda Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "ST????DM*", "*" }, /*quirks*/ADA_Q_4K }, { /* Seagate Momentus Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "ST9500423AS*", "*" }, /*quirks*/ADA_Q_4K }, { /* Seagate Momentus Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "ST9500424AS*", "*" }, /*quirks*/ADA_Q_4K }, { /* Seagate Momentus Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "ST9640423AS*", "*" }, /*quirks*/ADA_Q_4K }, { /* Seagate Momentus Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "ST9640424AS*", "*" }, /*quirks*/ADA_Q_4K }, { /* Seagate Momentus Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "ST9750420AS*", "*" }, /*quirks*/ADA_Q_4K }, { /* Seagate Momentus Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "ST9750422AS*", "*" }, /*quirks*/ADA_Q_4K }, { /* Seagate Momentus Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "ST9750423AS*", "*" }, /*quirks*/ADA_Q_4K }, { /* Seagate Momentus Thin Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "ST???LT*", "*" }, /*quirks*/ADA_Q_4K }, { /* WDC Caviar Green Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "WDC WD????RS*", "*" }, /*quirks*/ADA_Q_4K }, { /* WDC Caviar Green Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "WDC WD????RX*", "*" }, /*quirks*/ADA_Q_4K }, { /* WDC Caviar Green Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "WDC WD??????RS*", "*" }, /*quirks*/ADA_Q_4K }, { /* WDC Caviar Green Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "WDC WD??????RX*", "*" }, /*quirks*/ADA_Q_4K }, { /* WDC Scorpio Black Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "WDC WD???PKT*", "*" }, /*quirks*/ADA_Q_4K }, { /* WDC Scorpio Black Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "WDC WD?????PKT*", "*" }, /*quirks*/ADA_Q_4K }, { /* WDC Scorpio Blue Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "WDC WD???PVT*", "*" }, /*quirks*/ADA_Q_4K }, { /* WDC Scorpio Blue Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "WDC WD?????PVT*", "*" }, /*quirks*/ADA_Q_4K }, /* SSDs */ { /* * Corsair Force 2 SSDs * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "Corsair CSSD-F*", "*" }, /*quirks*/ADA_Q_4K }, { /* * Corsair Force 3 SSDs * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "Corsair Force 3*", "*" }, /*quirks*/ADA_Q_4K }, { /* * Corsair Neutron GTX SSDs * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "Corsair Neutron GTX*", "*" }, /*quirks*/ADA_Q_4K }, { /* * Corsair Force GT & GS SSDs * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "Corsair Force G*", "*" }, /*quirks*/ADA_Q_4K }, { /* * Crucial M4 SSDs * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "M4-CT???M4SSD2*", "*" }, /*quirks*/ADA_Q_4K }, { /* * Crucial RealSSD C300 SSDs * 4k optimised */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "C300-CTFDDAC???MAG*", "*" }, /*quirks*/ADA_Q_4K }, { /* * Intel 320 Series SSDs * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "INTEL SSDSA2CW*", "*" }, /*quirks*/ADA_Q_4K }, { /* * Intel 330 Series SSDs * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "INTEL SSDSC2CT*", "*" }, /*quirks*/ADA_Q_4K }, { /* * Intel 510 Series SSDs * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "INTEL SSDSC2MH*", "*" }, /*quirks*/ADA_Q_4K }, { /* * Intel 520 Series SSDs * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "INTEL SSDSC2BW*", "*" }, /*quirks*/ADA_Q_4K }, { /* * Intel X25-M Series SSDs * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "INTEL SSDSA2M*", "*" }, /*quirks*/ADA_Q_4K }, { /* * Kingston E100 Series SSDs * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "KINGSTON SE100S3*", "*" }, /*quirks*/ADA_Q_4K }, { /* * Kingston HyperX 3k SSDs * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "KINGSTON SH103S3*", "*" }, /*quirks*/ADA_Q_4K }, { /* * Marvell SSDs (entry taken from OpenSolaris) * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "MARVELL SD88SA02*", "*" }, /*quirks*/ADA_Q_4K }, { /* * OCZ Agility 2 SSDs * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "OCZ-AGILITY2*", "*" }, /*quirks*/ADA_Q_4K }, { /* * OCZ Agility 3 SSDs * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "OCZ-AGILITY3*", "*" }, /*quirks*/ADA_Q_4K }, { /* * OCZ Deneva R Series SSDs * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "DENRSTE251M45*", "*" }, /*quirks*/ADA_Q_4K }, { /* * OCZ Vertex 2 SSDs (inc pro series) * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "OCZ?VERTEX2*", "*" }, /*quirks*/ADA_Q_4K }, { /* * OCZ Vertex 3 SSDs * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "OCZ-VERTEX3*", "*" }, /*quirks*/ADA_Q_4K }, { /* * OCZ Vertex 4 SSDs * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "OCZ-VERTEX4*", "*" }, /*quirks*/ADA_Q_4K }, { /* * Samsung 830 Series SSDs * 4k optimised */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "SAMSUNG SSD 830 Series*", "*" }, /*quirks*/ADA_Q_4K }, { /* * Samsung 840 SSDs * 4k optimised */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "Samsung SSD 840*", "*" }, /*quirks*/ADA_Q_4K }, { /* * SuperTalent TeraDrive CT SSDs * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "FTM??CT25H*", "*" }, /*quirks*/ADA_Q_4K }, { /* * XceedIOPS SATA SSDs * 4k optimised */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "SG9XCS2D*", "*" }, /*quirks*/ADA_Q_4K }, { /* Default */ { T_ANY, SIP_MEDIA_REMOVABLE|SIP_MEDIA_FIXED, /*vendor*/"*", /*product*/"*", /*revision*/"*" }, /*quirks*/0 }, }; static disk_strategy_t adastrategy; static dumper_t adadump; static periph_init_t adainit; static void adaasync(void *callback_arg, u_int32_t code, struct cam_path *path, void *arg); static void adasysctlinit(void *context, int pending); static periph_ctor_t adaregister; static periph_dtor_t adacleanup; static periph_start_t adastart; static periph_oninv_t adaoninvalidate; static void adadone(struct cam_periph *periph, union ccb *done_ccb); static int adaerror(union ccb *ccb, u_int32_t cam_flags, u_int32_t sense_flags); static void adagetparams(struct cam_periph *periph, struct ccb_getdev *cgd); static timeout_t adasendorderedtag; static void adashutdown(void *arg, int howto); static void adasuspend(void *arg); static void adaresume(void *arg); #ifndef ADA_DEFAULT_LEGACY_ALIASES #define ADA_DEFAULT_LEGACY_ALIASES 1 #endif #ifndef ADA_DEFAULT_TIMEOUT #define ADA_DEFAULT_TIMEOUT 30 /* Timeout in seconds */ #endif #ifndef ADA_DEFAULT_RETRY #define ADA_DEFAULT_RETRY 4 #endif #ifndef ADA_DEFAULT_SEND_ORDERED #define ADA_DEFAULT_SEND_ORDERED 1 #endif #ifndef ADA_DEFAULT_SPINDOWN_SHUTDOWN #define ADA_DEFAULT_SPINDOWN_SHUTDOWN 1 #endif #ifndef ADA_DEFAULT_SPINDOWN_SUSPEND #define ADA_DEFAULT_SPINDOWN_SUSPEND 1 #endif #ifndef ADA_DEFAULT_READ_AHEAD #define ADA_DEFAULT_READ_AHEAD 1 #endif #ifndef ADA_DEFAULT_WRITE_CACHE #define ADA_DEFAULT_WRITE_CACHE 1 #endif #define ADA_RA (softc->read_ahead >= 0 ? \ softc->read_ahead : ada_read_ahead) #define ADA_WC (softc->write_cache >= 0 ? \ softc->write_cache : ada_write_cache) #define ADA_SIO (softc->sort_io_queue >= 0 ? \ softc->sort_io_queue : cam_sort_io_queues) /* * Most platforms map firmware geometry to actual, but some don't. If * not overridden, default to nothing. */ #ifndef ata_disk_firmware_geom_adjust #define ata_disk_firmware_geom_adjust(disk) #endif static int ada_legacy_aliases = ADA_DEFAULT_LEGACY_ALIASES; static int ada_retry_count = ADA_DEFAULT_RETRY; static int ada_default_timeout = ADA_DEFAULT_TIMEOUT; static int ada_send_ordered = ADA_DEFAULT_SEND_ORDERED; static int ada_spindown_shutdown = ADA_DEFAULT_SPINDOWN_SHUTDOWN; static int ada_spindown_suspend = ADA_DEFAULT_SPINDOWN_SUSPEND; static int ada_read_ahead = ADA_DEFAULT_READ_AHEAD; static int ada_write_cache = ADA_DEFAULT_WRITE_CACHE; static SYSCTL_NODE(_kern_cam, OID_AUTO, ada, CTLFLAG_RD, 0, "CAM Direct Access Disk driver"); SYSCTL_INT(_kern_cam_ada, OID_AUTO, legacy_aliases, CTLFLAG_RW, &ada_legacy_aliases, 0, "Create legacy-like device aliases"); TUNABLE_INT("kern.cam.ada.legacy_aliases", &ada_legacy_aliases); SYSCTL_INT(_kern_cam_ada, OID_AUTO, retry_count, CTLFLAG_RW, &ada_retry_count, 0, "Normal I/O retry count"); TUNABLE_INT("kern.cam.ada.retry_count", &ada_retry_count); SYSCTL_INT(_kern_cam_ada, OID_AUTO, default_timeout, CTLFLAG_RW, &ada_default_timeout, 0, "Normal I/O timeout (in seconds)"); TUNABLE_INT("kern.cam.ada.default_timeout", &ada_default_timeout); SYSCTL_INT(_kern_cam_ada, OID_AUTO, send_ordered, CTLFLAG_RW, &ada_send_ordered, 0, "Send Ordered Tags"); TUNABLE_INT("kern.cam.ada.send_ordered", &ada_send_ordered); SYSCTL_INT(_kern_cam_ada, OID_AUTO, spindown_shutdown, CTLFLAG_RW, &ada_spindown_shutdown, 0, "Spin down upon shutdown"); TUNABLE_INT("kern.cam.ada.spindown_shutdown", &ada_spindown_shutdown); SYSCTL_INT(_kern_cam_ada, OID_AUTO, spindown_suspend, CTLFLAG_RW, &ada_spindown_suspend, 0, "Spin down upon suspend"); TUNABLE_INT("kern.cam.ada.spindown_suspend", &ada_spindown_suspend); SYSCTL_INT(_kern_cam_ada, OID_AUTO, read_ahead, CTLFLAG_RW, &ada_read_ahead, 0, "Enable disk read-ahead"); TUNABLE_INT("kern.cam.ada.read_ahead", &ada_read_ahead); SYSCTL_INT(_kern_cam_ada, OID_AUTO, write_cache, CTLFLAG_RW, &ada_write_cache, 0, "Enable disk write cache"); TUNABLE_INT("kern.cam.ada.write_cache", &ada_write_cache); /* * ADA_ORDEREDTAG_INTERVAL determines how often, relative * to the default timeout, we check to see whether an ordered * tagged transaction is appropriate to prevent simple tag * starvation. Since we'd like to ensure that there is at least * 1/2 of the timeout length left for a starved transaction to * complete after we've sent an ordered tag, we must poll at least * four times in every timeout period. This takes care of the worst * case where a starved transaction starts during an interval that * meets the requirement "don't send an ordered tag" test so it takes * us two intervals to determine that a tag must be sent. */ #ifndef ADA_ORDEREDTAG_INTERVAL #define ADA_ORDEREDTAG_INTERVAL 4 #endif static struct periph_driver adadriver = { adainit, "ada", TAILQ_HEAD_INITIALIZER(adadriver.units), /* generation */ 0 }; PERIPHDRIVER_DECLARE(ada, adadriver); static MALLOC_DEFINE(M_ATADA, "ata_da", "ata_da buffers"); static int adaopen(struct disk *dp) { struct cam_periph *periph; struct ada_softc *softc; int error; periph = (struct cam_periph *)dp->d_drv1; if (cam_periph_acquire(periph) != CAM_REQ_CMP) { return(ENXIO); } cam_periph_lock(periph); if ((error = cam_periph_hold(periph, PRIBIO|PCATCH)) != 0) { cam_periph_unlock(periph); cam_periph_release(periph); return (error); } CAM_DEBUG(periph->path, CAM_DEBUG_TRACE | CAM_DEBUG_PERIPH, ("adaopen\n")); softc = (struct ada_softc *)periph->softc; softc->flags |= ADA_FLAG_OPEN; cam_periph_unhold(periph); cam_periph_unlock(periph); return (0); } static int adaclose(struct disk *dp) { struct cam_periph *periph; struct ada_softc *softc; union ccb *ccb; int error; periph = (struct cam_periph *)dp->d_drv1; softc = (struct ada_softc *)periph->softc; cam_periph_lock(periph); CAM_DEBUG(periph->path, CAM_DEBUG_TRACE | CAM_DEBUG_PERIPH, ("adaclose\n")); /* We only sync the cache if the drive is capable of it. */ if ((softc->flags & ADA_FLAG_DIRTY) != 0 && (softc->flags & ADA_FLAG_CAN_FLUSHCACHE) != 0 && (periph->flags & CAM_PERIPH_INVALID) == 0 && cam_periph_hold(periph, PRIBIO) == 0) { ccb = cam_periph_getccb(periph, CAM_PRIORITY_NORMAL); cam_fill_ataio(&ccb->ataio, 1, adadone, CAM_DIR_NONE, 0, NULL, 0, ada_default_timeout*1000); if (softc->flags & ADA_FLAG_CAN_48BIT) ata_48bit_cmd(&ccb->ataio, ATA_FLUSHCACHE48, 0, 0, 0); else ata_28bit_cmd(&ccb->ataio, ATA_FLUSHCACHE, 0, 0, 0); error = cam_periph_runccb(ccb, adaerror, /*cam_flags*/0, /*sense_flags*/0, softc->disk->d_devstat); if (error != 0) xpt_print(periph->path, "Synchronize cache failed\n"); else softc->flags &= ~ADA_FLAG_DIRTY; xpt_release_ccb(ccb); cam_periph_unhold(periph); } softc->flags &= ~ADA_FLAG_OPEN; while (softc->refcount != 0) cam_periph_sleep(periph, &softc->refcount, PRIBIO, "adaclose", 1); cam_periph_unlock(periph); cam_periph_release(periph); return (0); } static void adaschedule(struct cam_periph *periph) { struct ada_softc *softc = (struct ada_softc *)periph->softc; if (softc->state != ADA_STATE_NORMAL) return; /* Check if we have more work to do. */ if (bioq_first(&softc->bio_queue) || (!softc->trim_running && bioq_first(&softc->trim_queue))) { xpt_schedule(periph, CAM_PRIORITY_NORMAL); } } /* * Actually translate the requested transfer into one the physical driver * can understand. The transfer is described by a buf and will include * only one physical transfer. */ static void adastrategy(struct bio *bp) { struct cam_periph *periph; struct ada_softc *softc; periph = (struct cam_periph *)bp->bio_disk->d_drv1; softc = (struct ada_softc *)periph->softc; cam_periph_lock(periph); CAM_DEBUG(periph->path, CAM_DEBUG_TRACE, ("adastrategy(%p)\n", bp)); /* * If the device has been made invalid, error out */ if ((periph->flags & CAM_PERIPH_INVALID) != 0) { cam_periph_unlock(periph); biofinish(bp, NULL, ENXIO); return; } /* * Place it in the queue of disk activities for this disk */ if (bp->bio_cmd == BIO_DELETE) { KASSERT((softc->flags & ADA_FLAG_CAN_TRIM) || ((softc->flags & ADA_FLAG_CAN_CFA) && !(softc->flags & ADA_FLAG_CAN_48BIT)), ("BIO_DELETE but no supported TRIM method.")); bioq_disksort(&softc->trim_queue, bp); } else { if (ADA_SIO) bioq_disksort(&softc->bio_queue, bp); else bioq_insert_tail(&softc->bio_queue, bp); } /* * Schedule ourselves for performing the work. */ adaschedule(periph); cam_periph_unlock(periph); return; } static int adadump(void *arg, void *virtual, vm_offset_t physical, off_t offset, size_t length) { struct cam_periph *periph; struct ada_softc *softc; u_int secsize; union ccb ccb; struct disk *dp; uint64_t lba; uint16_t count; int error = 0; dp = arg; periph = dp->d_drv1; softc = (struct ada_softc *)periph->softc; cam_periph_lock(periph); secsize = softc->params.secsize; lba = offset / secsize; count = length / secsize; if ((periph->flags & CAM_PERIPH_INVALID) != 0) { cam_periph_unlock(periph); return (ENXIO); } if (length > 0) { xpt_setup_ccb(&ccb.ccb_h, periph->path, CAM_PRIORITY_NORMAL); ccb.ccb_h.ccb_state = ADA_CCB_DUMP; cam_fill_ataio(&ccb.ataio, 0, adadone, CAM_DIR_OUT, 0, (u_int8_t *) virtual, length, ada_default_timeout*1000); if ((softc->flags & ADA_FLAG_CAN_48BIT) && (lba + count >= ATA_MAX_28BIT_LBA || count >= 256)) { ata_48bit_cmd(&ccb.ataio, ATA_WRITE_DMA48, 0, lba, count); } else { ata_28bit_cmd(&ccb.ataio, ATA_WRITE_DMA, 0, lba, count); } xpt_polled_action(&ccb); error = cam_periph_error(&ccb, 0, SF_NO_RECOVERY | SF_NO_RETRY, NULL); if ((ccb.ccb_h.status & CAM_DEV_QFRZN) != 0) cam_release_devq(ccb.ccb_h.path, /*relsim_flags*/0, /*reduction*/0, /*timeout*/0, /*getcount_only*/0); if (error != 0) printf("Aborting dump due to I/O error.\n"); cam_periph_unlock(periph); return (error); } if (softc->flags & ADA_FLAG_CAN_FLUSHCACHE) { xpt_setup_ccb(&ccb.ccb_h, periph->path, CAM_PRIORITY_NORMAL); ccb.ccb_h.ccb_state = ADA_CCB_DUMP; cam_fill_ataio(&ccb.ataio, 0, adadone, CAM_DIR_NONE, 0, NULL, 0, ada_default_timeout*1000); if (softc->flags & ADA_FLAG_CAN_48BIT) ata_48bit_cmd(&ccb.ataio, ATA_FLUSHCACHE48, 0, 0, 0); else ata_28bit_cmd(&ccb.ataio, ATA_FLUSHCACHE, 0, 0, 0); xpt_polled_action(&ccb); error = cam_periph_error(&ccb, 0, SF_NO_RECOVERY | SF_NO_RETRY, NULL); if ((ccb.ccb_h.status & CAM_DEV_QFRZN) != 0) cam_release_devq(ccb.ccb_h.path, /*relsim_flags*/0, /*reduction*/0, /*timeout*/0, /*getcount_only*/0); if (error != 0) xpt_print(periph->path, "Synchronize cache failed\n"); } cam_periph_unlock(periph); return (error); } static void adainit(void) { cam_status status; /* * Install a global async callback. This callback will * receive async callbacks like "new device found". */ status = xpt_register_async(AC_FOUND_DEVICE, adaasync, NULL, NULL); if (status != CAM_REQ_CMP) { printf("ada: Failed to attach master async callback " "due to status 0x%x!\n", status); } else if (ada_send_ordered) { /* Register our event handlers */ if ((EVENTHANDLER_REGISTER(power_suspend, adasuspend, NULL, EVENTHANDLER_PRI_LAST)) == NULL) printf("adainit: power event registration failed!\n"); if ((EVENTHANDLER_REGISTER(power_resume, adaresume, NULL, EVENTHANDLER_PRI_LAST)) == NULL) printf("adainit: power event registration failed!\n"); if ((EVENTHANDLER_REGISTER(shutdown_post_sync, adashutdown, NULL, SHUTDOWN_PRI_DEFAULT)) == NULL) printf("adainit: shutdown event registration failed!\n"); } } /* * Callback from GEOM, called when it has finished cleaning up its * resources. */ static void adadiskgonecb(struct disk *dp) { struct cam_periph *periph; periph = (struct cam_periph *)dp->d_drv1; cam_periph_release(periph); } static void adaoninvalidate(struct cam_periph *periph) { struct ada_softc *softc; softc = (struct ada_softc *)periph->softc; /* * De-register any async callbacks. */ xpt_register_async(0, adaasync, periph, periph->path); /* * Return all queued I/O with ENXIO. * XXX Handle any transactions queued to the card * with XPT_ABORT_CCB. */ bioq_flush(&softc->bio_queue, NULL, ENXIO); bioq_flush(&softc->trim_queue, NULL, ENXIO); disk_gone(softc->disk); } static void adacleanup(struct cam_periph *periph) { struct ada_softc *softc; softc = (struct ada_softc *)periph->softc; cam_periph_unlock(periph); /* * If we can't free the sysctl tree, oh well... */ if ((softc->flags & ADA_FLAG_SCTX_INIT) != 0 && sysctl_ctx_free(&softc->sysctl_ctx) != 0) { xpt_print(periph->path, "can't remove sysctl context\n"); } disk_destroy(softc->disk); callout_drain(&softc->sendordered_c); free(softc, M_DEVBUF); cam_periph_lock(periph); } static void adaasync(void *callback_arg, u_int32_t code, struct cam_path *path, void *arg) { struct ccb_getdev cgd; struct cam_periph *periph; struct ada_softc *softc; periph = (struct cam_periph *)callback_arg; switch (code) { case AC_FOUND_DEVICE: { struct ccb_getdev *cgd; cam_status status; cgd = (struct ccb_getdev *)arg; if (cgd == NULL) break; if (cgd->protocol != PROTO_ATA) break; /* * Allocate a peripheral instance for * this device and start the probe * process. */ status = cam_periph_alloc(adaregister, adaoninvalidate, adacleanup, adastart, "ada", CAM_PERIPH_BIO, path, adaasync, AC_FOUND_DEVICE, cgd); if (status != CAM_REQ_CMP && status != CAM_REQ_INPROG) printf("adaasync: Unable to attach to new device " "due to status 0x%x\n", status); break; } case AC_GETDEV_CHANGED: { softc = (struct ada_softc *)periph->softc; xpt_setup_ccb(&cgd.ccb_h, periph->path, CAM_PRIORITY_NORMAL); cgd.ccb_h.func_code = XPT_GDEV_TYPE; xpt_action((union ccb *)&cgd); if ((cgd.ident_data.capabilities1 & ATA_SUPPORT_DMA) && (cgd.inq_flags & SID_DMA)) softc->flags |= ADA_FLAG_CAN_DMA; else softc->flags &= ~ADA_FLAG_CAN_DMA; if (cgd.ident_data.support.command2 & ATA_SUPPORT_ADDRESS48) { softc->flags |= ADA_FLAG_CAN_48BIT; if (cgd.inq_flags & SID_DMA48) softc->flags |= ADA_FLAG_CAN_DMA48; else softc->flags &= ~ADA_FLAG_CAN_DMA48; } else softc->flags &= ~(ADA_FLAG_CAN_48BIT | ADA_FLAG_CAN_DMA48); if ((cgd.ident_data.satacapabilities & ATA_SUPPORT_NCQ) && (cgd.inq_flags & SID_DMA) && (cgd.inq_flags & SID_CmdQue)) softc->flags |= ADA_FLAG_CAN_NCQ; else softc->flags &= ~ADA_FLAG_CAN_NCQ; if ((cgd.ident_data.support_dsm & ATA_SUPPORT_DSM_TRIM) && (cgd.inq_flags & SID_DMA)) softc->flags |= ADA_FLAG_CAN_TRIM; else softc->flags &= ~ADA_FLAG_CAN_TRIM; cam_periph_async(periph, code, path, arg); break; } case AC_ADVINFO_CHANGED: { uintptr_t buftype; buftype = (uintptr_t)arg; if (buftype == CDAI_TYPE_PHYS_PATH) { struct ada_softc *softc; softc = periph->softc; disk_attr_changed(softc->disk, "GEOM::physpath", M_NOWAIT); } break; } case AC_SENT_BDR: case AC_BUS_RESET: { softc = (struct ada_softc *)periph->softc; cam_periph_async(periph, code, path, arg); if (softc->state != ADA_STATE_NORMAL) break; xpt_setup_ccb(&cgd.ccb_h, periph->path, CAM_PRIORITY_NORMAL); cgd.ccb_h.func_code = XPT_GDEV_TYPE; xpt_action((union ccb *)&cgd); if (ADA_RA >= 0 && cgd.ident_data.support.command1 & ATA_SUPPORT_LOOKAHEAD) softc->state = ADA_STATE_RAHEAD; else if (ADA_WC >= 0 && cgd.ident_data.support.command1 & ATA_SUPPORT_WRITECACHE) softc->state = ADA_STATE_WCACHE; else break; if (cam_periph_acquire(periph) != CAM_REQ_CMP) softc->state = ADA_STATE_NORMAL; else xpt_schedule(periph, CAM_PRIORITY_DEV); } default: cam_periph_async(periph, code, path, arg); break; } } static void adasysctlinit(void *context, int pending) { struct cam_periph *periph; struct ada_softc *softc; char tmpstr[80], tmpstr2[80]; periph = (struct cam_periph *)context; /* periph was held for us when this task was enqueued */ if ((periph->flags & CAM_PERIPH_INVALID) != 0) { cam_periph_release(periph); return; } softc = (struct ada_softc *)periph->softc; snprintf(tmpstr, sizeof(tmpstr), "CAM ADA unit %d", periph->unit_number); snprintf(tmpstr2, sizeof(tmpstr2), "%d", periph->unit_number); sysctl_ctx_init(&softc->sysctl_ctx); softc->flags |= ADA_FLAG_SCTX_INIT; softc->sysctl_tree = SYSCTL_ADD_NODE(&softc->sysctl_ctx, SYSCTL_STATIC_CHILDREN(_kern_cam_ada), OID_AUTO, tmpstr2, CTLFLAG_RD, 0, tmpstr); if (softc->sysctl_tree == NULL) { printf("adasysctlinit: unable to allocate sysctl tree\n"); cam_periph_release(periph); return; } SYSCTL_ADD_INT(&softc->sysctl_ctx, SYSCTL_CHILDREN(softc->sysctl_tree), OID_AUTO, "read_ahead", CTLFLAG_RW | CTLFLAG_MPSAFE, &softc->read_ahead, 0, "Enable disk read ahead."); SYSCTL_ADD_INT(&softc->sysctl_ctx, SYSCTL_CHILDREN(softc->sysctl_tree), OID_AUTO, "write_cache", CTLFLAG_RW | CTLFLAG_MPSAFE, &softc->write_cache, 0, "Enable disk write cache."); SYSCTL_ADD_INT(&softc->sysctl_ctx, SYSCTL_CHILDREN(softc->sysctl_tree), OID_AUTO, "sort_io_queue", CTLFLAG_RW | CTLFLAG_MPSAFE, &softc->sort_io_queue, 0, "Sort IO queue to try and optimise disk access patterns"); #ifdef ADA_TEST_FAILURE /* * Add a 'door bell' sysctl which allows one to set it from userland * and cause something bad to happen. For the moment, we only allow * whacking the next read or write. */ SYSCTL_ADD_INT(&softc->sysctl_ctx, SYSCTL_CHILDREN(softc->sysctl_tree), OID_AUTO, "force_read_error", CTLFLAG_RW | CTLFLAG_MPSAFE, &softc->force_read_error, 0, "Force a read error for the next N reads."); SYSCTL_ADD_INT(&softc->sysctl_ctx, SYSCTL_CHILDREN(softc->sysctl_tree), OID_AUTO, "force_write_error", CTLFLAG_RW | CTLFLAG_MPSAFE, &softc->force_write_error, 0, "Force a write error for the next N writes."); SYSCTL_ADD_INT(&softc->sysctl_ctx, SYSCTL_CHILDREN(softc->sysctl_tree), OID_AUTO, "periodic_read_error", CTLFLAG_RW | CTLFLAG_MPSAFE, &softc->periodic_read_error, 0, "Force a read error every N reads (don't set too low)."); #endif cam_periph_release(periph); } static int adagetattr(struct bio *bp) { int ret; struct cam_periph *periph; periph = (struct cam_periph *)bp->bio_disk->d_drv1; cam_periph_lock(periph); ret = xpt_getattr(bp->bio_data, bp->bio_length, bp->bio_attribute, periph->path); cam_periph_unlock(periph); if (ret == 0) bp->bio_completed = bp->bio_length; return ret; } static cam_status adaregister(struct cam_periph *periph, void *arg) { struct ada_softc *softc; struct ccb_pathinq cpi; struct ccb_getdev *cgd; char announce_buf[80], buf1[32]; struct disk_params *dp; caddr_t match; u_int maxio; int legacy_id, quirks; cgd = (struct ccb_getdev *)arg; if (cgd == NULL) { printf("adaregister: no getdev CCB, can't register device\n"); return(CAM_REQ_CMP_ERR); } softc = (struct ada_softc *)malloc(sizeof(*softc), M_DEVBUF, M_NOWAIT|M_ZERO); if (softc == NULL) { printf("adaregister: Unable to probe new device. " "Unable to allocate softc\n"); return(CAM_REQ_CMP_ERR); } bioq_init(&softc->bio_queue); bioq_init(&softc->trim_queue); if ((cgd->ident_data.capabilities1 & ATA_SUPPORT_DMA) && (cgd->inq_flags & SID_DMA)) softc->flags |= ADA_FLAG_CAN_DMA; if (cgd->ident_data.support.command2 & ATA_SUPPORT_ADDRESS48) { softc->flags |= ADA_FLAG_CAN_48BIT; if (cgd->inq_flags & SID_DMA48) softc->flags |= ADA_FLAG_CAN_DMA48; } if (cgd->ident_data.support.command2 & ATA_SUPPORT_FLUSHCACHE) softc->flags |= ADA_FLAG_CAN_FLUSHCACHE; if (cgd->ident_data.support.command1 & ATA_SUPPORT_POWERMGT) softc->flags |= ADA_FLAG_CAN_POWERMGT; if ((cgd->ident_data.satacapabilities & ATA_SUPPORT_NCQ) && (cgd->inq_flags & SID_DMA) && (cgd->inq_flags & SID_CmdQue)) softc->flags |= ADA_FLAG_CAN_NCQ; if ((cgd->ident_data.support_dsm & ATA_SUPPORT_DSM_TRIM) && (cgd->inq_flags & SID_DMA)) { softc->flags |= ADA_FLAG_CAN_TRIM; softc->trim_max_ranges = TRIM_MAX_RANGES; if (cgd->ident_data.max_dsm_blocks != 0) { softc->trim_max_ranges = min(cgd->ident_data.max_dsm_blocks * ATA_DSM_BLK_RANGES, softc->trim_max_ranges); } } if (cgd->ident_data.support.command2 & ATA_SUPPORT_CFA) softc->flags |= ADA_FLAG_CAN_CFA; periph->softc = softc; /* * See if this device has any quirks. */ match = cam_quirkmatch((caddr_t)&cgd->ident_data, (caddr_t)ada_quirk_table, sizeof(ada_quirk_table)/sizeof(*ada_quirk_table), sizeof(*ada_quirk_table), ata_identify_match); if (match != NULL) softc->quirks = ((struct ada_quirk_entry *)match)->quirks; else softc->quirks = ADA_Q_NONE; bzero(&cpi, sizeof(cpi)); xpt_setup_ccb(&cpi.ccb_h, periph->path, CAM_PRIORITY_NONE); cpi.ccb_h.func_code = XPT_PATH_INQ; xpt_action((union ccb *)&cpi); TASK_INIT(&softc->sysctl_task, 0, adasysctlinit, periph); /* * Register this media as a disk */ (void)cam_periph_hold(periph, PRIBIO); cam_periph_unlock(periph); snprintf(announce_buf, sizeof(announce_buf), "kern.cam.ada.%d.quirks", periph->unit_number); quirks = softc->quirks; TUNABLE_INT_FETCH(announce_buf, &quirks); softc->quirks = quirks; softc->read_ahead = -1; snprintf(announce_buf, sizeof(announce_buf), "kern.cam.ada.%d.read_ahead", periph->unit_number); TUNABLE_INT_FETCH(announce_buf, &softc->read_ahead); softc->write_cache = -1; snprintf(announce_buf, sizeof(announce_buf), "kern.cam.ada.%d.write_cache", periph->unit_number); TUNABLE_INT_FETCH(announce_buf, &softc->write_cache); /* Disable queue sorting for non-rotational media by default. */ - if (cgd->ident_data.media_rotation_rate == 1) + if (cgd->ident_data.media_rotation_rate == ATA_RATE_NON_ROTATING) softc->sort_io_queue = 0; else softc->sort_io_queue = -1; adagetparams(periph, cgd); softc->disk = disk_alloc(); + softc->disk->d_rotation_rate = cgd->ident_data.media_rotation_rate; softc->disk->d_devstat = devstat_new_entry(periph->periph_name, periph->unit_number, softc->params.secsize, DEVSTAT_ALL_SUPPORTED, DEVSTAT_TYPE_DIRECT | XPORT_DEVSTAT_TYPE(cpi.transport), DEVSTAT_PRIORITY_DISK); softc->disk->d_open = adaopen; softc->disk->d_close = adaclose; softc->disk->d_strategy = adastrategy; softc->disk->d_getattr = adagetattr; softc->disk->d_dump = adadump; softc->disk->d_gone = adadiskgonecb; softc->disk->d_name = "ada"; softc->disk->d_drv1 = periph; maxio = cpi.maxio; /* Honor max I/O size of SIM */ if (maxio == 0) maxio = DFLTPHYS; /* traditional default */ else if (maxio > MAXPHYS) maxio = MAXPHYS; /* for safety */ if (softc->flags & ADA_FLAG_CAN_48BIT) maxio = min(maxio, 65536 * softc->params.secsize); else /* 28bit ATA command limit */ maxio = min(maxio, 256 * softc->params.secsize); softc->disk->d_maxsize = maxio; softc->disk->d_unit = periph->unit_number; softc->disk->d_flags = DISKFLAG_DIRECT_COMPLETION; if (softc->flags & ADA_FLAG_CAN_FLUSHCACHE) softc->disk->d_flags |= DISKFLAG_CANFLUSHCACHE; if (softc->flags & ADA_FLAG_CAN_TRIM) { softc->disk->d_flags |= DISKFLAG_CANDELETE; softc->disk->d_delmaxsize = softc->params.secsize * ATA_DSM_RANGE_MAX * softc->trim_max_ranges; } else if ((softc->flags & ADA_FLAG_CAN_CFA) && !(softc->flags & ADA_FLAG_CAN_48BIT)) { softc->disk->d_flags |= DISKFLAG_CANDELETE; softc->disk->d_delmaxsize = 256 * softc->params.secsize; } else softc->disk->d_delmaxsize = maxio; if ((cpi.hba_misc & PIM_UNMAPPED) != 0) softc->disk->d_flags |= DISKFLAG_UNMAPPED_BIO; strlcpy(softc->disk->d_descr, cgd->ident_data.model, MIN(sizeof(softc->disk->d_descr), sizeof(cgd->ident_data.model))); strlcpy(softc->disk->d_ident, cgd->ident_data.serial, MIN(sizeof(softc->disk->d_ident), sizeof(cgd->ident_data.serial))); softc->disk->d_hba_vendor = cpi.hba_vendor; softc->disk->d_hba_device = cpi.hba_device; softc->disk->d_hba_subvendor = cpi.hba_subvendor; softc->disk->d_hba_subdevice = cpi.hba_subdevice; softc->disk->d_sectorsize = softc->params.secsize; softc->disk->d_mediasize = (off_t)softc->params.sectors * softc->params.secsize; if (ata_physical_sector_size(&cgd->ident_data) != softc->params.secsize) { softc->disk->d_stripesize = ata_physical_sector_size(&cgd->ident_data); softc->disk->d_stripeoffset = (softc->disk->d_stripesize - ata_logical_sector_offset(&cgd->ident_data)) % softc->disk->d_stripesize; } else if (softc->quirks & ADA_Q_4K) { softc->disk->d_stripesize = 4096; softc->disk->d_stripeoffset = 0; } softc->disk->d_fwsectors = softc->params.secs_per_track; softc->disk->d_fwheads = softc->params.heads; ata_disk_firmware_geom_adjust(softc->disk); if (ada_legacy_aliases) { #ifdef ATA_STATIC_ID legacy_id = xpt_path_legacy_ata_id(periph->path); #else legacy_id = softc->disk->d_unit; #endif if (legacy_id >= 0) { snprintf(announce_buf, sizeof(announce_buf), "kern.devalias.%s%d", softc->disk->d_name, softc->disk->d_unit); snprintf(buf1, sizeof(buf1), "ad%d", legacy_id); setenv(announce_buf, buf1); } } else legacy_id = -1; /* * Acquire a reference to the periph before we register with GEOM. * We'll release this reference once GEOM calls us back (via * adadiskgonecb()) telling us that our provider has been freed. */ if (cam_periph_acquire(periph) != CAM_REQ_CMP) { xpt_print(periph->path, "%s: lost periph during " "registration!\n", __func__); cam_periph_lock(periph); return (CAM_REQ_CMP_ERR); } disk_create(softc->disk, DISK_VERSION); cam_periph_lock(periph); cam_periph_unhold(periph); dp = &softc->params; snprintf(announce_buf, sizeof(announce_buf), "%juMB (%ju %u byte sectors: %dH %dS/T %dC)", (uintmax_t)(((uintmax_t)dp->secsize * dp->sectors) / (1024*1024)), (uintmax_t)dp->sectors, dp->secsize, dp->heads, dp->secs_per_track, dp->cylinders); xpt_announce_periph(periph, announce_buf); xpt_announce_quirks(periph, softc->quirks, ADA_Q_BIT_STRING); if (legacy_id >= 0) printf("%s%d: Previously was known as ad%d\n", periph->periph_name, periph->unit_number, legacy_id); /* * Create our sysctl variables, now that we know * we have successfully attached. */ if (cam_periph_acquire(periph) == CAM_REQ_CMP) taskqueue_enqueue(taskqueue_thread, &softc->sysctl_task); /* * Add async callbacks for bus reset and * bus device reset calls. I don't bother * checking if this fails as, in most cases, * the system will function just fine without * them and the only alternative would be to * not attach the device on failure. */ xpt_register_async(AC_SENT_BDR | AC_BUS_RESET | AC_LOST_DEVICE | AC_GETDEV_CHANGED | AC_ADVINFO_CHANGED, adaasync, periph, periph->path); /* * Schedule a periodic event to occasionally send an * ordered tag to a device. */ callout_init_mtx(&softc->sendordered_c, cam_periph_mtx(periph), 0); callout_reset(&softc->sendordered_c, (ada_default_timeout * hz) / ADA_ORDEREDTAG_INTERVAL, adasendorderedtag, softc); if (ADA_RA >= 0 && cgd->ident_data.support.command1 & ATA_SUPPORT_LOOKAHEAD) { softc->state = ADA_STATE_RAHEAD; } else if (ADA_WC >= 0 && cgd->ident_data.support.command1 & ATA_SUPPORT_WRITECACHE) { softc->state = ADA_STATE_WCACHE; } else { softc->state = ADA_STATE_NORMAL; return(CAM_REQ_CMP); } if (cam_periph_acquire(periph) != CAM_REQ_CMP) softc->state = ADA_STATE_NORMAL; else xpt_schedule(periph, CAM_PRIORITY_DEV); return(CAM_REQ_CMP); } static void ada_dsmtrim(struct ada_softc *softc, struct bio *bp, struct ccb_ataio *ataio) { struct trim_request *req = &softc->trim_req; uint64_t lastlba = (uint64_t)-1; int c, lastcount = 0, off, ranges = 0; bzero(req, sizeof(*req)); TAILQ_INIT(&req->bps); do { uint64_t lba = bp->bio_pblkno; int count = bp->bio_bcount / softc->params.secsize; bioq_remove(&softc->trim_queue, bp); /* Try to extend the previous range. */ if (lba == lastlba) { c = min(count, ATA_DSM_RANGE_MAX - lastcount); lastcount += c; off = (ranges - 1) * ATA_DSM_RANGE_SIZE; req->data[off + 6] = lastcount & 0xff; req->data[off + 7] = (lastcount >> 8) & 0xff; count -= c; lba += c; } while (count > 0) { c = min(count, ATA_DSM_RANGE_MAX); off = ranges * ATA_DSM_RANGE_SIZE; req->data[off + 0] = lba & 0xff; req->data[off + 1] = (lba >> 8) & 0xff; req->data[off + 2] = (lba >> 16) & 0xff; req->data[off + 3] = (lba >> 24) & 0xff; req->data[off + 4] = (lba >> 32) & 0xff; req->data[off + 5] = (lba >> 40) & 0xff; req->data[off + 6] = c & 0xff; req->data[off + 7] = (c >> 8) & 0xff; lba += c; count -= c; lastcount = c; ranges++; /* * Its the caller's responsibility to ensure the * request will fit so we don't need to check for * overrun here */ } lastlba = lba; TAILQ_INSERT_TAIL(&req->bps, bp, bio_queue); bp = bioq_first(&softc->trim_queue); if (bp == NULL || bp->bio_bcount / softc->params.secsize > (softc->trim_max_ranges - ranges) * ATA_DSM_RANGE_MAX) break; } while (1); cam_fill_ataio(ataio, ada_retry_count, adadone, CAM_DIR_OUT, 0, req->data, ((ranges + ATA_DSM_BLK_RANGES - 1) / ATA_DSM_BLK_RANGES) * ATA_DSM_BLK_SIZE, ada_default_timeout * 1000); ata_48bit_cmd(ataio, ATA_DATA_SET_MANAGEMENT, ATA_DSM_TRIM, 0, (ranges + ATA_DSM_BLK_RANGES - 1) / ATA_DSM_BLK_RANGES); } static void ada_cfaerase(struct ada_softc *softc, struct bio *bp, struct ccb_ataio *ataio) { uint64_t lba = bp->bio_pblkno; uint16_t count = bp->bio_bcount / softc->params.secsize; cam_fill_ataio(ataio, ada_retry_count, adadone, CAM_DIR_NONE, 0, NULL, 0, ada_default_timeout*1000); if (count >= 256) count = 0; ata_28bit_cmd(ataio, ATA_CFA_ERASE, 0, lba, count); } static void adastart(struct cam_periph *periph, union ccb *start_ccb) { struct ada_softc *softc = (struct ada_softc *)periph->softc; struct ccb_ataio *ataio = &start_ccb->ataio; CAM_DEBUG(periph->path, CAM_DEBUG_TRACE, ("adastart\n")); switch (softc->state) { case ADA_STATE_NORMAL: { struct bio *bp; u_int8_t tag_code; /* Run TRIM if not running yet. */ if (!softc->trim_running && (bp = bioq_first(&softc->trim_queue)) != 0) { if (softc->flags & ADA_FLAG_CAN_TRIM) { ada_dsmtrim(softc, bp, ataio); } else if ((softc->flags & ADA_FLAG_CAN_CFA) && !(softc->flags & ADA_FLAG_CAN_48BIT)) { ada_cfaerase(softc, bp, ataio); } else { panic("adastart: BIO_DELETE without method, not possible."); } softc->trim_running = 1; start_ccb->ccb_h.ccb_state = ADA_CCB_TRIM; start_ccb->ccb_h.flags |= CAM_UNLOCKED; goto out; } /* Run regular command. */ bp = bioq_first(&softc->bio_queue); if (bp == NULL) { xpt_release_ccb(start_ccb); break; } bioq_remove(&softc->bio_queue, bp); if ((bp->bio_flags & BIO_ORDERED) != 0 || (softc->flags & ADA_FLAG_NEED_OTAG) != 0) { softc->flags &= ~ADA_FLAG_NEED_OTAG; softc->flags |= ADA_FLAG_WAS_OTAG; tag_code = 0; } else { tag_code = 1; } switch (bp->bio_cmd) { case BIO_WRITE: softc->flags |= ADA_FLAG_DIRTY; /* FALLTHROUGH */ case BIO_READ: { uint64_t lba = bp->bio_pblkno; uint16_t count = bp->bio_bcount / softc->params.secsize; #ifdef ADA_TEST_FAILURE int fail = 0; /* * Support the failure ioctls. If the command is a * read, and there are pending forced read errors, or * if a write and pending write errors, then fail this * operation with EIO. This is useful for testing * purposes. Also, support having every Nth read fail. * * This is a rather blunt tool. */ if (bp->bio_cmd == BIO_READ) { if (softc->force_read_error) { softc->force_read_error--; fail = 1; } if (softc->periodic_read_error > 0) { if (++softc->periodic_read_count >= softc->periodic_read_error) { softc->periodic_read_count = 0; fail = 1; } } } else { if (softc->force_write_error) { softc->force_write_error--; fail = 1; } } if (fail) { bp->bio_error = EIO; bp->bio_flags |= BIO_ERROR; biodone(bp); xpt_release_ccb(start_ccb); adaschedule(periph); return; } #endif KASSERT((bp->bio_flags & BIO_UNMAPPED) == 0 || round_page(bp->bio_bcount + bp->bio_ma_offset) / PAGE_SIZE == bp->bio_ma_n, ("Short bio %p", bp)); cam_fill_ataio(ataio, ada_retry_count, adadone, (bp->bio_cmd == BIO_READ ? CAM_DIR_IN : CAM_DIR_OUT) | ((bp->bio_flags & BIO_UNMAPPED) != 0 ? CAM_DATA_BIO : 0), tag_code, ((bp->bio_flags & BIO_UNMAPPED) != 0) ? (void *)bp : bp->bio_data, bp->bio_bcount, ada_default_timeout*1000); if ((softc->flags & ADA_FLAG_CAN_NCQ) && tag_code) { if (bp->bio_cmd == BIO_READ) { ata_ncq_cmd(ataio, ATA_READ_FPDMA_QUEUED, lba, count); } else { ata_ncq_cmd(ataio, ATA_WRITE_FPDMA_QUEUED, lba, count); } } else if ((softc->flags & ADA_FLAG_CAN_48BIT) && (lba + count >= ATA_MAX_28BIT_LBA || count > 256)) { if (softc->flags & ADA_FLAG_CAN_DMA48) { if (bp->bio_cmd == BIO_READ) { ata_48bit_cmd(ataio, ATA_READ_DMA48, 0, lba, count); } else { ata_48bit_cmd(ataio, ATA_WRITE_DMA48, 0, lba, count); } } else { if (bp->bio_cmd == BIO_READ) { ata_48bit_cmd(ataio, ATA_READ_MUL48, 0, lba, count); } else { ata_48bit_cmd(ataio, ATA_WRITE_MUL48, 0, lba, count); } } } else { if (count == 256) count = 0; if (softc->flags & ADA_FLAG_CAN_DMA) { if (bp->bio_cmd == BIO_READ) { ata_28bit_cmd(ataio, ATA_READ_DMA, 0, lba, count); } else { ata_28bit_cmd(ataio, ATA_WRITE_DMA, 0, lba, count); } } else { if (bp->bio_cmd == BIO_READ) { ata_28bit_cmd(ataio, ATA_READ_MUL, 0, lba, count); } else { ata_28bit_cmd(ataio, ATA_WRITE_MUL, 0, lba, count); } } } break; } case BIO_FLUSH: cam_fill_ataio(ataio, 1, adadone, CAM_DIR_NONE, 0, NULL, 0, ada_default_timeout*1000); if (softc->flags & ADA_FLAG_CAN_48BIT) ata_48bit_cmd(ataio, ATA_FLUSHCACHE48, 0, 0, 0); else ata_28bit_cmd(ataio, ATA_FLUSHCACHE, 0, 0, 0); break; } start_ccb->ccb_h.ccb_state = ADA_CCB_BUFFER_IO; start_ccb->ccb_h.flags |= CAM_UNLOCKED; out: start_ccb->ccb_h.ccb_bp = bp; softc->outstanding_cmds++; softc->refcount++; cam_periph_unlock(periph); xpt_action(start_ccb); cam_periph_lock(periph); softc->refcount--; /* May have more work to do, so ensure we stay scheduled */ adaschedule(periph); break; } case ADA_STATE_RAHEAD: case ADA_STATE_WCACHE: { cam_fill_ataio(ataio, 1, adadone, CAM_DIR_NONE, 0, NULL, 0, ada_default_timeout*1000); if (softc->state == ADA_STATE_RAHEAD) { ata_28bit_cmd(ataio, ATA_SETFEATURES, ADA_RA ? ATA_SF_ENAB_RCACHE : ATA_SF_DIS_RCACHE, 0, 0); start_ccb->ccb_h.ccb_state = ADA_CCB_RAHEAD; } else { ata_28bit_cmd(ataio, ATA_SETFEATURES, ADA_WC ? ATA_SF_ENAB_WCACHE : ATA_SF_DIS_WCACHE, 0, 0); start_ccb->ccb_h.ccb_state = ADA_CCB_WCACHE; } start_ccb->ccb_h.flags |= CAM_DEV_QFREEZE; xpt_action(start_ccb); break; } } } static void adadone(struct cam_periph *periph, union ccb *done_ccb) { struct ada_softc *softc; struct ccb_ataio *ataio; struct ccb_getdev *cgd; struct cam_path *path; int state; softc = (struct ada_softc *)periph->softc; ataio = &done_ccb->ataio; path = done_ccb->ccb_h.path; CAM_DEBUG(path, CAM_DEBUG_TRACE, ("adadone\n")); state = ataio->ccb_h.ccb_state & ADA_CCB_TYPE_MASK; switch (state) { case ADA_CCB_BUFFER_IO: case ADA_CCB_TRIM: { struct bio *bp; int error; cam_periph_lock(periph); if ((done_ccb->ccb_h.status & CAM_STATUS_MASK) != CAM_REQ_CMP) { error = adaerror(done_ccb, 0, 0); if (error == ERESTART) { /* A retry was scheduled, so just return. */ cam_periph_unlock(periph); return; } if ((done_ccb->ccb_h.status & CAM_DEV_QFRZN) != 0) cam_release_devq(path, /*relsim_flags*/0, /*reduction*/0, /*timeout*/0, /*getcount_only*/0); } else { if ((done_ccb->ccb_h.status & CAM_DEV_QFRZN) != 0) panic("REQ_CMP with QFRZN"); error = 0; } bp = (struct bio *)done_ccb->ccb_h.ccb_bp; bp->bio_error = error; if (error != 0) { bp->bio_resid = bp->bio_bcount; bp->bio_flags |= BIO_ERROR; } else { if (state == ADA_CCB_TRIM) bp->bio_resid = 0; else bp->bio_resid = ataio->resid; if (bp->bio_resid > 0) bp->bio_flags |= BIO_ERROR; } softc->outstanding_cmds--; if (softc->outstanding_cmds == 0) softc->flags |= ADA_FLAG_WAS_OTAG; xpt_release_ccb(done_ccb); if (state == ADA_CCB_TRIM) { TAILQ_HEAD(, bio) queue; struct bio *bp1; TAILQ_INIT(&queue); TAILQ_CONCAT(&queue, &softc->trim_req.bps, bio_queue); softc->trim_running = 0; adaschedule(periph); cam_periph_unlock(periph); while ((bp1 = TAILQ_FIRST(&queue)) != NULL) { TAILQ_REMOVE(&queue, bp1, bio_queue); bp1->bio_error = error; if (error != 0) { bp1->bio_flags |= BIO_ERROR; bp1->bio_resid = bp1->bio_bcount; } else bp1->bio_resid = 0; biodone(bp1); } } else { cam_periph_unlock(periph); biodone(bp); } return; } case ADA_CCB_RAHEAD: { if ((done_ccb->ccb_h.status & CAM_STATUS_MASK) != CAM_REQ_CMP) { if (adaerror(done_ccb, 0, 0) == ERESTART) { out: /* Drop freeze taken due to CAM_DEV_QFREEZE */ cam_release_devq(path, 0, 0, 0, FALSE); return; } else if ((done_ccb->ccb_h.status & CAM_DEV_QFRZN) != 0) { cam_release_devq(path, /*relsim_flags*/0, /*reduction*/0, /*timeout*/0, /*getcount_only*/0); } } /* * Since our peripheral may be invalidated by an error * above or an external event, we must release our CCB * before releasing the reference on the peripheral. * The peripheral will only go away once the last reference * is removed, and we need it around for the CCB release * operation. */ cgd = (struct ccb_getdev *)done_ccb; xpt_setup_ccb(&cgd->ccb_h, path, CAM_PRIORITY_NORMAL); cgd->ccb_h.func_code = XPT_GDEV_TYPE; xpt_action((union ccb *)cgd); if (ADA_WC >= 0 && cgd->ident_data.support.command1 & ATA_SUPPORT_WRITECACHE) { softc->state = ADA_STATE_WCACHE; xpt_release_ccb(done_ccb); xpt_schedule(periph, CAM_PRIORITY_DEV); goto out; } softc->state = ADA_STATE_NORMAL; xpt_release_ccb(done_ccb); /* Drop freeze taken due to CAM_DEV_QFREEZE */ cam_release_devq(path, 0, 0, 0, FALSE); adaschedule(periph); cam_periph_release_locked(periph); return; } case ADA_CCB_WCACHE: { if ((done_ccb->ccb_h.status & CAM_STATUS_MASK) != CAM_REQ_CMP) { if (adaerror(done_ccb, 0, 0) == ERESTART) { goto out; } else if ((done_ccb->ccb_h.status & CAM_DEV_QFRZN) != 0) { cam_release_devq(path, /*relsim_flags*/0, /*reduction*/0, /*timeout*/0, /*getcount_only*/0); } } softc->state = ADA_STATE_NORMAL; /* * Since our peripheral may be invalidated by an error * above or an external event, we must release our CCB * before releasing the reference on the peripheral. * The peripheral will only go away once the last reference * is removed, and we need it around for the CCB release * operation. */ xpt_release_ccb(done_ccb); /* Drop freeze taken due to CAM_DEV_QFREEZE */ cam_release_devq(path, 0, 0, 0, FALSE); adaschedule(periph); cam_periph_release_locked(periph); return; } case ADA_CCB_DUMP: /* No-op. We're polling */ return; default: break; } xpt_release_ccb(done_ccb); } static int adaerror(union ccb *ccb, u_int32_t cam_flags, u_int32_t sense_flags) { return(cam_periph_error(ccb, cam_flags, sense_flags, NULL)); } static void adagetparams(struct cam_periph *periph, struct ccb_getdev *cgd) { struct ada_softc *softc = (struct ada_softc *)periph->softc; struct disk_params *dp = &softc->params; u_int64_t lbasize48; u_int32_t lbasize; dp->secsize = ata_logical_sector_size(&cgd->ident_data); if ((cgd->ident_data.atavalid & ATA_FLAG_54_58) && cgd->ident_data.current_heads && cgd->ident_data.current_sectors) { dp->heads = cgd->ident_data.current_heads; dp->secs_per_track = cgd->ident_data.current_sectors; dp->cylinders = cgd->ident_data.cylinders; dp->sectors = (u_int32_t)cgd->ident_data.current_size_1 | ((u_int32_t)cgd->ident_data.current_size_2 << 16); } else { dp->heads = cgd->ident_data.heads; dp->secs_per_track = cgd->ident_data.sectors; dp->cylinders = cgd->ident_data.cylinders; dp->sectors = cgd->ident_data.cylinders * dp->heads * dp->secs_per_track; } lbasize = (u_int32_t)cgd->ident_data.lba_size_1 | ((u_int32_t)cgd->ident_data.lba_size_2 << 16); /* use the 28bit LBA size if valid or bigger than the CHS mapping */ if (cgd->ident_data.cylinders == 16383 || dp->sectors < lbasize) dp->sectors = lbasize; /* use the 48bit LBA size if valid */ lbasize48 = ((u_int64_t)cgd->ident_data.lba_size48_1) | ((u_int64_t)cgd->ident_data.lba_size48_2 << 16) | ((u_int64_t)cgd->ident_data.lba_size48_3 << 32) | ((u_int64_t)cgd->ident_data.lba_size48_4 << 48); if ((cgd->ident_data.support.command2 & ATA_SUPPORT_ADDRESS48) && lbasize48 > ATA_MAX_28BIT_LBA) dp->sectors = lbasize48; } static void adasendorderedtag(void *arg) { struct ada_softc *softc = arg; if (ada_send_ordered) { if (softc->outstanding_cmds > 0) { if ((softc->flags & ADA_FLAG_WAS_OTAG) == 0) softc->flags |= ADA_FLAG_NEED_OTAG; softc->flags &= ~ADA_FLAG_WAS_OTAG; } } /* Queue us up again */ callout_reset(&softc->sendordered_c, (ada_default_timeout * hz) / ADA_ORDEREDTAG_INTERVAL, adasendorderedtag, softc); } /* * Step through all ADA peripheral drivers, and if the device is still open, * sync the disk cache to physical media. */ static void adaflush(void) { struct cam_periph *periph; struct ada_softc *softc; union ccb *ccb; int error; CAM_PERIPH_FOREACH(periph, &adadriver) { softc = (struct ada_softc *)periph->softc; if (SCHEDULER_STOPPED()) { /* If we paniced with the lock held, do not recurse. */ if (!cam_periph_owned(periph) && (softc->flags & ADA_FLAG_OPEN)) { adadump(softc->disk, NULL, 0, 0, 0); } continue; } cam_periph_lock(periph); /* * We only sync the cache if the drive is still open, and * if the drive is capable of it.. */ if (((softc->flags & ADA_FLAG_OPEN) == 0) || (softc->flags & ADA_FLAG_CAN_FLUSHCACHE) == 0) { cam_periph_unlock(periph); continue; } ccb = cam_periph_getccb(periph, CAM_PRIORITY_NORMAL); cam_fill_ataio(&ccb->ataio, 0, adadone, CAM_DIR_NONE, 0, NULL, 0, ada_default_timeout*1000); if (softc->flags & ADA_FLAG_CAN_48BIT) ata_48bit_cmd(&ccb->ataio, ATA_FLUSHCACHE48, 0, 0, 0); else ata_28bit_cmd(&ccb->ataio, ATA_FLUSHCACHE, 0, 0, 0); error = cam_periph_runccb(ccb, adaerror, /*cam_flags*/0, /*sense_flags*/ SF_NO_RECOVERY | SF_NO_RETRY, softc->disk->d_devstat); if (error != 0) xpt_print(periph->path, "Synchronize cache failed\n"); xpt_release_ccb(ccb); cam_periph_unlock(periph); } } static void adaspindown(uint8_t cmd, int flags) { struct cam_periph *periph; struct ada_softc *softc; union ccb *ccb; int error; CAM_PERIPH_FOREACH(periph, &adadriver) { /* If we paniced with lock held - not recurse here. */ if (cam_periph_owned(periph)) continue; cam_periph_lock(periph); softc = (struct ada_softc *)periph->softc; /* * We only spin-down the drive if it is capable of it.. */ if ((softc->flags & ADA_FLAG_CAN_POWERMGT) == 0) { cam_periph_unlock(periph); continue; } if (bootverbose) xpt_print(periph->path, "spin-down\n"); ccb = cam_periph_getccb(periph, CAM_PRIORITY_NORMAL); cam_fill_ataio(&ccb->ataio, 0, adadone, CAM_DIR_NONE | flags, 0, NULL, 0, ada_default_timeout*1000); ata_28bit_cmd(&ccb->ataio, cmd, 0, 0, 0); error = cam_periph_runccb(ccb, adaerror, /*cam_flags*/0, /*sense_flags*/ SF_NO_RECOVERY | SF_NO_RETRY, softc->disk->d_devstat); if (error != 0) xpt_print(periph->path, "Spin-down disk failed\n"); xpt_release_ccb(ccb); cam_periph_unlock(periph); } } static void adashutdown(void *arg, int howto) { adaflush(); if (ada_spindown_shutdown != 0 && (howto & (RB_HALT | RB_POWEROFF)) != 0) adaspindown(ATA_STANDBY_IMMEDIATE, 0); } static void adasuspend(void *arg) { adaflush(); if (ada_spindown_suspend != 0) adaspindown(ATA_SLEEP, CAM_DEV_QFREEZE); } static void adaresume(void *arg) { struct cam_periph *periph; struct ada_softc *softc; if (ada_spindown_suspend == 0) return; CAM_PERIPH_FOREACH(periph, &adadriver) { cam_periph_lock(periph); softc = (struct ada_softc *)periph->softc; /* * We only spin-down the drive if it is capable of it.. */ if ((softc->flags & ADA_FLAG_CAN_POWERMGT) == 0) { cam_periph_unlock(periph); continue; } if (bootverbose) xpt_print(periph->path, "resume\n"); /* * Drop freeze taken due to CAM_DEV_QFREEZE flag set on * sleep request. */ cam_release_devq(periph->path, /*relsim_flags*/0, /*openings*/0, /*timeout*/0, /*getcount_only*/0); cam_periph_unlock(periph); } } #endif /* _KERNEL */ Index: stable/10/sys/cam/scsi/scsi_all.h =================================================================== --- stable/10/sys/cam/scsi/scsi_all.h (revision 271237) +++ stable/10/sys/cam/scsi/scsi_all.h (revision 271238) @@ -1,3651 +1,3651 @@ /*- * Largely written by Julian Elischer (julian@tfs.com) * for TRW Financial Systems. * * TRW Financial Systems, in accordance with their agreement with Carnegie * Mellon University, makes this software available to CMU to distribute * or use in any manner that they see fit as long as this message is kept with * the software. For this reason TFS also grants any other persons or * organisations permission to use or modify this software. * * TFS supplies this software to be publicly redistributed * on the understanding that TFS is not responsible for the correct * functioning of this software in any circumstances. * * Ported to run under 386BSD by Julian Elischer (julian@tfs.com) Sept 1992 * * $FreeBSD$ */ /* * SCSI general interface description */ #ifndef _SCSI_SCSI_ALL_H #define _SCSI_SCSI_ALL_H 1 #include #include #ifdef _KERNEL /* * This is the number of seconds we wait for devices to settle after a SCSI * bus reset. */ extern int scsi_delay; #endif /* _KERNEL */ /* * SCSI command format */ /* * Define dome bits that are in ALL (or a lot of) scsi commands */ #define SCSI_CTL_LINK 0x01 #define SCSI_CTL_FLAG 0x02 #define SCSI_CTL_VENDOR 0xC0 #define SCSI_CMD_LUN 0xA0 /* these two should not be needed */ #define SCSI_CMD_LUN_SHIFT 5 /* LUN in the cmd is no longer SCSI */ #define SCSI_MAX_CDBLEN 16 /* * 16 byte commands are in the * SCSI-3 spec */ #if defined(CAM_MAX_CDBLEN) && (CAM_MAX_CDBLEN < SCSI_MAX_CDBLEN) #error "CAM_MAX_CDBLEN cannot be less than SCSI_MAX_CDBLEN" #endif /* 6byte CDBs special case 0 length to be 256 */ #define SCSI_CDB6_LEN(len) ((len) == 0 ? 256 : len) /* * This type defines actions to be taken when a particular sense code is * received. Right now, these flags are only defined to take up 16 bits, * but can be expanded in the future if necessary. */ typedef enum { SS_NOP = 0x000000, /* Do nothing */ SS_RETRY = 0x010000, /* Retry the command */ SS_FAIL = 0x020000, /* Bail out */ SS_START = 0x030000, /* Send a Start Unit command to the device, * then retry the original command. */ SS_TUR = 0x040000, /* Send a Test Unit Ready command to the * device, then retry the original command. */ SS_MASK = 0xff0000 } scsi_sense_action; typedef enum { SSQ_NONE = 0x0000, SSQ_DECREMENT_COUNT = 0x0100, /* Decrement the retry count */ SSQ_MANY = 0x0200, /* send lots of recovery commands */ SSQ_RANGE = 0x0400, /* * This table entry represents the * end of a range of ASCQs that * have identical error actions * and text. */ SSQ_PRINT_SENSE = 0x0800, SSQ_UA = 0x1000, /* Broadcast UA. */ SSQ_RESCAN = 0x2000, /* Rescan target for LUNs. */ SSQ_LOST = 0x4000, /* Destroy the LUNs. */ SSQ_MASK = 0xff00 } scsi_sense_action_qualifier; /* Mask for error status values */ #define SS_ERRMASK 0xff /* The default, retyable, error action */ #define SS_RDEF SS_RETRY|SSQ_DECREMENT_COUNT|SSQ_PRINT_SENSE|EIO /* The retyable, error action, with table specified error code */ #define SS_RET SS_RETRY|SSQ_DECREMENT_COUNT|SSQ_PRINT_SENSE /* Fatal error action, with table specified error code */ #define SS_FATAL SS_FAIL|SSQ_PRINT_SENSE struct scsi_generic { u_int8_t opcode; u_int8_t bytes[11]; }; struct scsi_request_sense { u_int8_t opcode; u_int8_t byte2; #define SRS_DESC 0x01 u_int8_t unused[2]; u_int8_t length; u_int8_t control; }; struct scsi_test_unit_ready { u_int8_t opcode; u_int8_t byte2; u_int8_t unused[3]; u_int8_t control; }; struct scsi_receive_diag { uint8_t opcode; uint8_t byte2; #define SRD_PCV 0x01 uint8_t page_code; uint8_t length[2]; uint8_t control; }; struct scsi_send_diag { uint8_t opcode; uint8_t byte2; #define SSD_UNITOFFL 0x01 #define SSD_DEVOFFL 0x02 #define SSD_SELFTEST 0x04 #define SSD_PF 0x10 #define SSD_SELF_TEST_CODE_MASK 0xE0 #define SSD_SELF_TEST_CODE_SHIFT 5 #define SSD_SELF_TEST_CODE_NONE 0x00 #define SSD_SELF_TEST_CODE_BG_SHORT 0x01 #define SSD_SELF_TEST_CODE_BG_EXTENDED 0x02 #define SSD_SELF_TEST_CODE_BG_ABORT 0x04 #define SSD_SELF_TEST_CODE_FG_SHORT 0x05 #define SSD_SELF_TEST_CODE_FG_EXTENDED 0x06 uint8_t reserved; uint8_t length[2]; uint8_t control; }; struct scsi_sense { u_int8_t opcode; u_int8_t byte2; u_int8_t unused[2]; u_int8_t length; u_int8_t control; }; struct scsi_inquiry { u_int8_t opcode; u_int8_t byte2; #define SI_EVPD 0x01 #define SI_CMDDT 0x02 u_int8_t page_code; u_int8_t length[2]; u_int8_t control; }; struct scsi_mode_sense_6 { u_int8_t opcode; u_int8_t byte2; #define SMS_DBD 0x08 u_int8_t page; #define SMS_PAGE_CODE 0x3F #define SMS_VENDOR_SPECIFIC_PAGE 0x00 #define SMS_DISCONNECT_RECONNECT_PAGE 0x02 #define SMS_FORMAT_DEVICE_PAGE 0x03 #define SMS_GEOMETRY_PAGE 0x04 #define SMS_CACHE_PAGE 0x08 #define SMS_PERIPHERAL_DEVICE_PAGE 0x09 #define SMS_CONTROL_MODE_PAGE 0x0A #define SMS_PROTO_SPECIFIC_PAGE 0x19 #define SMS_INFO_EXCEPTIONS_PAGE 0x1C #define SMS_ALL_PAGES_PAGE 0x3F #define SMS_PAGE_CTRL_MASK 0xC0 #define SMS_PAGE_CTRL_CURRENT 0x00 #define SMS_PAGE_CTRL_CHANGEABLE 0x40 #define SMS_PAGE_CTRL_DEFAULT 0x80 #define SMS_PAGE_CTRL_SAVED 0xC0 u_int8_t subpage; #define SMS_SUBPAGE_PAGE_0 0x00 #define SMS_SUBPAGE_ALL 0xff u_int8_t length; u_int8_t control; }; struct scsi_mode_sense_10 { u_int8_t opcode; u_int8_t byte2; /* same bits as small version */ #define SMS10_LLBAA 0x10 u_int8_t page; /* same bits as small version */ u_int8_t subpage; u_int8_t unused[3]; u_int8_t length[2]; u_int8_t control; }; struct scsi_mode_select_6 { u_int8_t opcode; u_int8_t byte2; #define SMS_SP 0x01 #define SMS_PF 0x10 u_int8_t unused[2]; u_int8_t length; u_int8_t control; }; struct scsi_mode_select_10 { u_int8_t opcode; u_int8_t byte2; /* same bits as small version */ u_int8_t unused[5]; u_int8_t length[2]; u_int8_t control; }; /* * When sending a mode select to a tape drive, the medium type must be 0. */ struct scsi_mode_hdr_6 { u_int8_t datalen; u_int8_t medium_type; u_int8_t dev_specific; u_int8_t block_descr_len; }; struct scsi_mode_hdr_10 { u_int8_t datalen[2]; u_int8_t medium_type; u_int8_t dev_specific; u_int8_t reserved[2]; u_int8_t block_descr_len[2]; }; struct scsi_mode_block_descr { u_int8_t density_code; u_int8_t num_blocks[3]; u_int8_t reserved; u_int8_t block_len[3]; }; struct scsi_per_res_in { u_int8_t opcode; u_int8_t action; #define SPRI_RK 0x00 #define SPRI_RR 0x01 #define SPRI_RC 0x02 #define SPRI_RS 0x03 u_int8_t reserved[5]; u_int8_t length[2]; #define SPRI_MAX_LEN 0xffff u_int8_t control; }; struct scsi_per_res_in_header { u_int8_t generation[4]; u_int8_t length[4]; }; struct scsi_per_res_key { u_int8_t key[8]; }; struct scsi_per_res_in_keys { struct scsi_per_res_in_header header; struct scsi_per_res_key keys[0]; }; struct scsi_per_res_cap { uint8_t length[2]; uint8_t flags1; #define SPRI_RLR_C 0x80 #define SPRI_CRH 0x10 #define SPRI_SIP_C 0x08 #define SPRI_ATP_C 0x04 #define SPRI_PTPL_C 0x01 uint8_t flags2; #define SPRI_TMV 0x80 #define SPRI_ALLOW_CMD_MASK 0x70 #define SPRI_ALLOW_CMD_SHIFT 4 #define SPRI_ALLOW_NA 0x00 #define SPRI_ALLOW_1 0x10 #define SPRI_ALLOW_2 0x20 #define SPRI_ALLOW_3 0x30 #define SPRI_ALLOW_4 0x40 #define SPRI_PTPL_A 0x01 uint8_t type_mask[2]; #define SPRI_TM_WR_EX_AR 0x8000 #define SPRI_TM_EX_AC_RO 0x4000 #define SPRI_TM_WR_EX_RO 0x2000 #define SPRI_TM_EX_AC 0x0800 #define SPRI_TM_WR_EX 0x0200 #define SPRI_TM_EX_AC_AR 0x0001 uint8_t reserved[2]; }; struct scsi_per_res_in_rsrv_data { uint8_t reservation[8]; uint8_t scope_addr[4]; uint8_t reserved; uint8_t scopetype; #define SPRT_WE 0x01 #define SPRT_EA 0x03 #define SPRT_WERO 0x05 #define SPRT_EARO 0x06 #define SPRT_WEAR 0x07 #define SPRT_EAAR 0x08 uint8_t extent_length[2]; }; struct scsi_per_res_in_rsrv { struct scsi_per_res_in_header header; struct scsi_per_res_in_rsrv_data data; }; struct scsi_per_res_in_full_desc { struct scsi_per_res_key res_key; uint8_t reserved1[4]; uint8_t flags; #define SPRI_FULL_ALL_TG_PT 0x02 #define SPRI_FULL_R_HOLDER 0x01 uint8_t scopetype; uint8_t reserved2[4]; uint8_t rel_trgt_port_id[2]; uint8_t additional_length[4]; uint8_t transport_id[]; }; struct scsi_per_res_in_full { struct scsi_per_res_in_header header; struct scsi_per_res_in_full_desc desc[]; }; struct scsi_per_res_out { u_int8_t opcode; u_int8_t action; #define SPRO_REGISTER 0x00 #define SPRO_RESERVE 0x01 #define SPRO_RELEASE 0x02 #define SPRO_CLEAR 0x03 #define SPRO_PREEMPT 0x04 #define SPRO_PRE_ABO 0x05 #define SPRO_REG_IGNO 0x06 #define SPRO_REG_MOVE 0x07 #define SPRO_REPL_LOST_RES 0x08 #define SPRO_ACTION_MASK 0x1f u_int8_t scope_type; #define SPR_SCOPE_MASK 0xf0 #define SPR_SCOPE_SHIFT 4 #define SPR_LU_SCOPE 0x00 #define SPR_EXTENT_SCOPE 0x10 #define SPR_ELEMENT_SCOPE 0x20 #define SPR_TYPE_MASK 0x0f #define SPR_TYPE_RD_SHARED 0x00 #define SPR_TYPE_WR_EX 0x01 #define SPR_TYPE_RD_EX 0x02 #define SPR_TYPE_EX_AC 0x03 #define SPR_TYPE_SHARED 0x04 #define SPR_TYPE_WR_EX_RO 0x05 #define SPR_TYPE_EX_AC_RO 0x06 #define SPR_TYPE_WR_EX_AR 0x07 #define SPR_TYPE_EX_AC_AR 0x08 u_int8_t reserved[2]; u_int8_t length[4]; u_int8_t control; }; struct scsi_per_res_out_parms { struct scsi_per_res_key res_key; u_int8_t serv_act_res_key[8]; u_int8_t scope_spec_address[4]; u_int8_t flags; #define SPR_SPEC_I_PT 0x08 #define SPR_ALL_TG_PT 0x04 #define SPR_APTPL 0x01 u_int8_t reserved1; u_int8_t extent_length[2]; u_int8_t transport_id_list[]; }; struct scsi_per_res_out_trans_ids { u_int8_t additional_length[4]; u_int8_t transport_ids[]; }; /* * Used with REGISTER AND MOVE serivce action of the PERSISTENT RESERVE OUT * command. */ struct scsi_per_res_reg_move { struct scsi_per_res_key res_key; u_int8_t serv_act_res_key[8]; u_int8_t reserved; u_int8_t flags; #define SPR_REG_MOVE_UNREG 0x02 #define SPR_REG_MOVE_APTPL 0x01 u_int8_t rel_trgt_port_id[2]; u_int8_t transport_id_length[4]; u_int8_t transport_id[]; }; struct scsi_transportid_header { uint8_t format_protocol; #define SCSI_TRN_FORMAT_MASK 0xc0 #define SCSI_TRN_FORMAT_SHIFT 6 #define SCSI_TRN_PROTO_MASK 0x0f }; struct scsi_transportid_fcp { uint8_t format_protocol; #define SCSI_TRN_FCP_FORMAT_DEFAULT 0x00 uint8_t reserved1[7]; uint8_t n_port_name[8]; uint8_t reserved2[8]; }; struct scsi_transportid_spi { uint8_t format_protocol; #define SCSI_TRN_SPI_FORMAT_DEFAULT 0x00 uint8_t reserved1; uint8_t scsi_addr[2]; uint8_t obsolete[2]; uint8_t rel_trgt_port_id[2]; uint8_t reserved2[16]; }; struct scsi_transportid_1394 { uint8_t format_protocol; #define SCSI_TRN_1394_FORMAT_DEFAULT 0x00 uint8_t reserved1[7]; uint8_t eui64[8]; uint8_t reserved2[8]; }; struct scsi_transportid_rdma { uint8_t format_protocol; #define SCSI_TRN_RDMA_FORMAT_DEFAULT 0x00 uint8_t reserved[7]; #define SCSI_TRN_RDMA_PORT_LEN 16 uint8_t initiator_port_id[SCSI_TRN_RDMA_PORT_LEN]; }; struct scsi_transportid_iscsi_device { uint8_t format_protocol; #define SCSI_TRN_ISCSI_FORMAT_DEVICE 0x00 uint8_t reserved; uint8_t additional_length[2]; uint8_t iscsi_name[]; }; struct scsi_transportid_iscsi_port { uint8_t format_protocol; #define SCSI_TRN_ISCSI_FORMAT_PORT 0x40 uint8_t reserved; uint8_t additional_length[2]; uint8_t iscsi_name[]; /* * Followed by a separator and iSCSI initiator session ID */ }; struct scsi_transportid_sas { uint8_t format_protocol; #define SCSI_TRN_SAS_FORMAT_DEFAULT 0x00 uint8_t reserved1[3]; uint8_t sas_address[8]; uint8_t reserved2[12]; }; struct scsi_sop_routing_id_norm { uint8_t bus; uint8_t devfunc; #define SCSI_TRN_SOP_BUS_MAX 0xff #define SCSI_TRN_SOP_DEV_MAX 0x1f #define SCSI_TRN_SOP_DEV_MASK 0xf8 #define SCSI_TRN_SOP_DEV_SHIFT 3 #define SCSI_TRN_SOP_FUNC_NORM_MASK 0x07 #define SCSI_TRN_SOP_FUNC_NORM_MAX 0x07 }; struct scsi_sop_routing_id_alt { uint8_t bus; uint8_t function; #define SCSI_TRN_SOP_FUNC_ALT_MAX 0xff }; struct scsi_transportid_sop { uint8_t format_protocol; #define SCSI_TRN_SOP_FORMAT_DEFAULT 0x00 uint8_t reserved1; uint8_t routing_id[2]; uint8_t reserved2[20]; }; struct scsi_log_sense { u_int8_t opcode; u_int8_t byte2; #define SLS_SP 0x01 #define SLS_PPC 0x02 u_int8_t page; #define SLS_PAGE_CODE 0x3F #define SLS_ALL_PAGES_PAGE 0x00 #define SLS_OVERRUN_PAGE 0x01 #define SLS_ERROR_WRITE_PAGE 0x02 #define SLS_ERROR_READ_PAGE 0x03 #define SLS_ERROR_READREVERSE_PAGE 0x04 #define SLS_ERROR_VERIFY_PAGE 0x05 #define SLS_ERROR_NONMEDIUM_PAGE 0x06 #define SLS_ERROR_LASTN_PAGE 0x07 #define SLS_SELF_TEST_PAGE 0x10 #define SLS_IE_PAGE 0x2f #define SLS_PAGE_CTRL_MASK 0xC0 #define SLS_PAGE_CTRL_THRESHOLD 0x00 #define SLS_PAGE_CTRL_CUMULATIVE 0x40 #define SLS_PAGE_CTRL_THRESH_DEFAULT 0x80 #define SLS_PAGE_CTRL_CUMUL_DEFAULT 0xC0 u_int8_t reserved[2]; u_int8_t paramptr[2]; u_int8_t length[2]; u_int8_t control; }; struct scsi_log_select { u_int8_t opcode; u_int8_t byte2; /* SLS_SP 0x01 */ #define SLS_PCR 0x02 u_int8_t page; /* SLS_PAGE_CTRL_MASK 0xC0 */ /* SLS_PAGE_CTRL_THRESHOLD 0x00 */ /* SLS_PAGE_CTRL_CUMULATIVE 0x40 */ /* SLS_PAGE_CTRL_THRESH_DEFAULT 0x80 */ /* SLS_PAGE_CTRL_CUMUL_DEFAULT 0xC0 */ u_int8_t reserved[4]; u_int8_t length[2]; u_int8_t control; }; struct scsi_log_header { u_int8_t page; u_int8_t reserved; u_int8_t datalen[2]; }; struct scsi_log_param_header { u_int8_t param_code[2]; u_int8_t param_control; #define SLP_LP 0x01 #define SLP_LBIN 0x02 #define SLP_TMC_MASK 0x0C #define SLP_TMC_ALWAYS 0x00 #define SLP_TMC_EQUAL 0x04 #define SLP_TMC_NOTEQUAL 0x08 #define SLP_TMC_GREATER 0x0C #define SLP_ETC 0x10 #define SLP_TSD 0x20 #define SLP_DS 0x40 #define SLP_DU 0x80 u_int8_t param_len; }; struct scsi_control_page { u_int8_t page_code; u_int8_t page_length; u_int8_t rlec; #define SCP_RLEC 0x01 /*Report Log Exception Cond*/ #define SCP_GLTSD 0x02 /*Global Logging target save disable */ #define SCP_DSENSE 0x04 /*Descriptor Sense */ #define SCP_DPICZ 0x08 /*Disable Prot. Info Check if Prot. Field is Zero */ #define SCP_TMF_ONLY 0x10 /*TM Functions Only*/ #define SCP_TST_MASK 0xE0 /*Task Set Type Mask*/ #define SCP_TST_ONE 0x00 /*One Task Set*/ #define SCP_TST_SEPARATE 0x20 /*Separate Task Sets*/ u_int8_t queue_flags; #define SCP_QUEUE_ALG_MASK 0xF0 #define SCP_QUEUE_ALG_RESTRICTED 0x00 #define SCP_QUEUE_ALG_UNRESTRICTED 0x10 #define SCP_NUAR 0x08 /*No UA on release*/ #define SCP_QUEUE_ERR 0x02 /*Queued I/O aborted for CACs*/ #define SCP_QUEUE_DQUE 0x01 /*Queued I/O disabled*/ u_int8_t eca_and_aen; #define SCP_EECA 0x80 /*Enable Extended CA*/ #define SCP_RAC 0x40 /*Report a check*/ #define SCP_SWP 0x08 /*Software Write Protect*/ #define SCP_RAENP 0x04 /*Ready AEN Permission*/ #define SCP_UAAENP 0x02 /*UA AEN Permission*/ #define SCP_EAENP 0x01 /*Error AEN Permission*/ u_int8_t flags4; #define SCP_ATO 0x80 /*Application tag owner*/ #define SCP_TAS 0x40 /*Task aborted status*/ #define SCP_ATMPE 0x20 /*Application tag mode page*/ #define SCP_RWWP 0x10 /*Reject write without prot*/ u_int8_t aen_holdoff_period[2]; u_int8_t busy_timeout_period[2]; u_int8_t extended_selftest_completion_time[2]; }; struct scsi_cache_page { u_int8_t page_code; #define SCHP_PAGE_SAVABLE 0x80 /* Page is savable */ u_int8_t page_length; u_int8_t cache_flags; #define SCHP_FLAGS_WCE 0x04 /* Write Cache Enable */ #define SCHP_FLAGS_MF 0x02 /* Multiplication factor */ #define SCHP_FLAGS_RCD 0x01 /* Read Cache Disable */ u_int8_t rw_cache_policy; u_int8_t dis_prefetch[2]; u_int8_t min_prefetch[2]; u_int8_t max_prefetch[2]; u_int8_t max_prefetch_ceil[2]; }; /* * XXX KDM * Updated version of the cache page, as of SBC. Update this to SBC-3 and * rationalize the two. */ struct scsi_caching_page { uint8_t page_code; #define SMS_CACHING_PAGE 0x08 uint8_t page_length; uint8_t flags1; #define SCP_IC 0x80 #define SCP_ABPF 0x40 #define SCP_CAP 0x20 #define SCP_DISC 0x10 #define SCP_SIZE 0x08 #define SCP_WCE 0x04 #define SCP_MF 0x02 #define SCP_RCD 0x01 uint8_t ret_priority; uint8_t disable_pf_transfer_len[2]; uint8_t min_prefetch[2]; uint8_t max_prefetch[2]; uint8_t max_pf_ceiling[2]; uint8_t flags2; #define SCP_FSW 0x80 #define SCP_LBCSS 0x40 #define SCP_DRA 0x20 #define SCP_VS1 0x10 #define SCP_VS2 0x08 uint8_t cache_segments; uint8_t cache_seg_size[2]; uint8_t reserved; uint8_t non_cache_seg_size[3]; }; /* * XXX KDM move this off to a vendor shim. */ struct copan_power_subpage { uint8_t page_code; #define PWR_PAGE_CODE 0x00 uint8_t subpage; #define PWR_SUBPAGE_CODE 0x02 uint8_t page_length[2]; uint8_t page_version; #define PWR_VERSION 0x01 uint8_t total_luns; uint8_t max_active_luns; #define PWR_DFLT_MAX_LUNS 0x07 uint8_t reserved[25]; }; /* * XXX KDM move this off to a vendor shim. */ struct copan_aps_subpage { uint8_t page_code; #define APS_PAGE_CODE 0x00 uint8_t subpage; #define APS_SUBPAGE_CODE 0x03 uint8_t page_length[2]; uint8_t page_version; #define APS_VERSION 0x00 uint8_t lock_active; #define APS_LOCK_ACTIVE 0x01 #define APS_LOCK_INACTIVE 0x00 uint8_t reserved[26]; }; /* * XXX KDM move this off to a vendor shim. */ struct copan_debugconf_subpage { uint8_t page_code; #define DBGCNF_PAGE_CODE 0x00 uint8_t subpage; #define DBGCNF_SUBPAGE_CODE 0xF0 uint8_t page_length[2]; uint8_t page_version; #define DBGCNF_VERSION 0x00 uint8_t ctl_time_io_secs[2]; }; struct scsi_info_exceptions_page { u_int8_t page_code; #define SIEP_PAGE_SAVABLE 0x80 /* Page is savable */ u_int8_t page_length; u_int8_t info_flags; #define SIEP_FLAGS_PERF 0x80 #define SIEP_FLAGS_EBF 0x20 #define SIEP_FLAGS_EWASC 0x10 #define SIEP_FLAGS_DEXCPT 0x08 #define SIEP_FLAGS_TEST 0x04 #define SIEP_FLAGS_EBACKERR 0x02 #define SIEP_FLAGS_LOGERR 0x01 u_int8_t mrie; u_int8_t interval_timer[4]; u_int8_t report_count[4]; }; /* * SCSI protocol identifier values, current as of SPC4r36l. */ #define SCSI_PROTO_FC 0x00 /* Fibre Channel */ #define SCSI_PROTO_SPI 0x01 /* Parallel SCSI */ #define SCSI_PROTO_SSA 0x02 /* Serial Storage Arch. */ #define SCSI_PROTO_1394 0x03 /* IEEE 1394 (Firewire) */ #define SCSI_PROTO_RDMA 0x04 /* SCSI RDMA Protocol */ #define SCSI_PROTO_ISCSI 0x05 /* Internet SCSI */ #define SCSI_PROTO_iSCSI 0x05 /* Internet SCSI */ #define SCSI_PROTO_SAS 0x06 /* SAS Serial SCSI Protocol */ #define SCSI_PROTO_ADT 0x07 /* Automation/Drive Int. Trans. Prot.*/ #define SCSI_PROTO_ADITP 0x07 /* Automation/Drive Int. Trans. Prot.*/ #define SCSI_PROTO_ATA 0x08 /* AT Attachment Interface */ #define SCSI_PROTO_UAS 0x09 /* USB Atached SCSI */ #define SCSI_PROTO_SOP 0x0a /* SCSI over PCI Express */ #define SCSI_PROTO_NONE 0x0f /* No specific protocol */ struct scsi_proto_specific_page { u_int8_t page_code; #define SPSP_PAGE_SAVABLE 0x80 /* Page is savable */ u_int8_t page_length; u_int8_t protocol; #define SPSP_PROTO_FC SCSI_PROTO_FC #define SPSP_PROTO_SPI SCSI_PROTO_SPI #define SPSP_PROTO_SSA SCSI_PROTO_SSA #define SPSP_PROTO_1394 SCSI_PROTO_1394 #define SPSP_PROTO_RDMA SCSI_PROTO_RDMA #define SPSP_PROTO_ISCSI SCSI_PROTO_ISCSI #define SPSP_PROTO_SAS SCSI_PROTO_SAS #define SPSP_PROTO_ADT SCSI_PROTO_ADITP #define SPSP_PROTO_ATA SCSI_PROTO_ATA #define SPSP_PROTO_UAS SCSI_PROTO_UAS #define SPSP_PROTO_SOP SCSI_PROTO_SOP #define SPSP_PROTO_NONE SCSI_PROTO_NONE }; struct scsi_reserve { u_int8_t opcode; u_int8_t byte2; #define SR_EXTENT 0x01 #define SR_ID_MASK 0x0e #define SR_3RDPTY 0x10 #define SR_LUN_MASK 0xe0 u_int8_t resv_id; u_int8_t length[2]; u_int8_t control; }; struct scsi_reserve_10 { uint8_t opcode; uint8_t byte2; #define SR10_3RDPTY 0x10 #define SR10_LONGID 0x02 #define SR10_EXTENT 0x01 uint8_t resv_id; uint8_t thirdparty_id; uint8_t reserved[3]; uint8_t length[2]; uint8_t control; }; struct scsi_release { u_int8_t opcode; u_int8_t byte2; u_int8_t resv_id; u_int8_t unused[1]; u_int8_t length; u_int8_t control; }; struct scsi_release_10 { uint8_t opcode; uint8_t byte2; uint8_t resv_id; uint8_t thirdparty_id; uint8_t reserved[3]; uint8_t length[2]; uint8_t control; }; struct scsi_prevent { u_int8_t opcode; u_int8_t byte2; u_int8_t unused[2]; u_int8_t how; u_int8_t control; }; #define PR_PREVENT 0x01 #define PR_ALLOW 0x00 struct scsi_sync_cache { u_int8_t opcode; u_int8_t byte2; #define SSC_IMMED 0x02 #define SSC_RELADR 0x01 u_int8_t begin_lba[4]; u_int8_t reserved; u_int8_t lb_count[2]; u_int8_t control; }; struct scsi_sync_cache_16 { uint8_t opcode; uint8_t byte2; uint8_t begin_lba[8]; uint8_t lb_count[4]; uint8_t reserved; uint8_t control; }; struct scsi_format { uint8_t opcode; uint8_t byte2; #define SF_LONGLIST 0x20 #define SF_FMTDATA 0x10 #define SF_CMPLIST 0x08 #define SF_FORMAT_MASK 0x07 #define SF_FORMAT_BLOCK 0x00 #define SF_FORMAT_LONG_BLOCK 0x03 #define SF_FORMAT_BFI 0x04 #define SF_FORMAT_PHYS 0x05 uint8_t vendor; uint8_t interleave[2]; uint8_t control; }; struct scsi_format_header_short { uint8_t reserved; #define SF_DATA_FOV 0x80 #define SF_DATA_DPRY 0x40 #define SF_DATA_DCRT 0x20 #define SF_DATA_STPF 0x10 #define SF_DATA_IP 0x08 #define SF_DATA_DSP 0x04 #define SF_DATA_IMMED 0x02 #define SF_DATA_VS 0x01 uint8_t byte2; uint8_t defect_list_len[2]; }; struct scsi_format_header_long { uint8_t reserved; uint8_t byte2; uint8_t reserved2[2]; uint8_t defect_list_len[4]; }; struct scsi_changedef { u_int8_t opcode; u_int8_t byte2; u_int8_t unused1; u_int8_t how; u_int8_t unused[4]; u_int8_t datalen; u_int8_t control; }; struct scsi_read_buffer { u_int8_t opcode; u_int8_t byte2; #define RWB_MODE 0x1F #define RWB_MODE_HDR_DATA 0x00 #define RWB_MODE_VENDOR 0x01 #define RWB_MODE_DATA 0x02 #define RWB_MODE_DESCR 0x03 #define RWB_MODE_DOWNLOAD 0x04 #define RWB_MODE_DOWNLOAD_SAVE 0x05 #define RWB_MODE_ECHO 0x0A #define RWB_MODE_ECHO_DESCR 0x0B #define RWB_MODE_ERROR_HISTORY 0x1C u_int8_t buffer_id; u_int8_t offset[3]; u_int8_t length[3]; u_int8_t control; }; struct scsi_write_buffer { u_int8_t opcode; u_int8_t byte2; u_int8_t buffer_id; u_int8_t offset[3]; u_int8_t length[3]; u_int8_t control; }; struct scsi_rw_6 { u_int8_t opcode; u_int8_t addr[3]; /* only 5 bits are valid in the MSB address byte */ #define SRW_TOPADDR 0x1F u_int8_t length; u_int8_t control; }; struct scsi_rw_10 { u_int8_t opcode; #define SRW10_RELADDR 0x01 /* EBP defined for WRITE(10) only */ #define SRW10_EBP 0x04 #define SRW10_FUA 0x08 #define SRW10_DPO 0x10 u_int8_t byte2; u_int8_t addr[4]; u_int8_t reserved; u_int8_t length[2]; u_int8_t control; }; struct scsi_rw_12 { u_int8_t opcode; #define SRW12_RELADDR 0x01 #define SRW12_FUA 0x08 #define SRW12_DPO 0x10 u_int8_t byte2; u_int8_t addr[4]; u_int8_t length[4]; u_int8_t reserved; u_int8_t control; }; struct scsi_rw_16 { u_int8_t opcode; #define SRW16_RELADDR 0x01 #define SRW16_FUA 0x08 #define SRW16_DPO 0x10 u_int8_t byte2; u_int8_t addr[8]; u_int8_t length[4]; u_int8_t reserved; u_int8_t control; }; struct scsi_write_same_10 { uint8_t opcode; uint8_t byte2; #define SWS_LBDATA 0x02 #define SWS_PBDATA 0x04 #define SWS_UNMAP 0x08 #define SWS_ANCHOR 0x10 uint8_t addr[4]; uint8_t group; uint8_t length[2]; uint8_t control; }; struct scsi_write_same_16 { uint8_t opcode; uint8_t byte2; uint8_t addr[8]; uint8_t length[4]; uint8_t group; uint8_t control; }; struct scsi_unmap { uint8_t opcode; uint8_t byte2; #define SU_ANCHOR 0x01 uint8_t reserved[4]; uint8_t group; uint8_t length[2]; uint8_t control; }; struct scsi_unmap_header { uint8_t length[2]; uint8_t desc_length[2]; uint8_t reserved[4]; }; struct scsi_unmap_desc { uint8_t lba[8]; uint8_t length[4]; uint8_t reserved[4]; }; struct scsi_write_verify_10 { uint8_t opcode; uint8_t byte2; #define SWV_BYTCHK 0x02 #define SWV_DPO 0x10 #define SWV_WRPROECT_MASK 0xe0 uint8_t addr[4]; uint8_t group; uint8_t length[2]; uint8_t control; }; struct scsi_write_verify_12 { uint8_t opcode; uint8_t byte2; uint8_t addr[4]; uint8_t length[4]; uint8_t group; uint8_t control; }; struct scsi_write_verify_16 { uint8_t opcode; uint8_t byte2; uint8_t addr[8]; uint8_t length[4]; uint8_t group; uint8_t control; }; struct scsi_start_stop_unit { u_int8_t opcode; u_int8_t byte2; #define SSS_IMMED 0x01 u_int8_t reserved[2]; u_int8_t how; #define SSS_START 0x01 #define SSS_LOEJ 0x02 #define SSS_PC_MASK 0xf0 #define SSS_PC_START_VALID 0x00 #define SSS_PC_ACTIVE 0x10 #define SSS_PC_IDLE 0x20 #define SSS_PC_STANDBY 0x30 #define SSS_PC_LU_CONTROL 0x70 #define SSS_PC_FORCE_IDLE_0 0xa0 #define SSS_PC_FORCE_STANDBY_0 0xb0 u_int8_t control; }; struct ata_pass_12 { u_int8_t opcode; u_int8_t protocol; #define AP_PROTO_HARD_RESET (0x00 << 1) #define AP_PROTO_SRST (0x01 << 1) #define AP_PROTO_NON_DATA (0x03 << 1) #define AP_PROTO_PIO_IN (0x04 << 1) #define AP_PROTO_PIO_OUT (0x05 << 1) #define AP_PROTO_DMA (0x06 << 1) #define AP_PROTO_DMA_QUEUED (0x07 << 1) #define AP_PROTO_DEVICE_DIAG (0x08 << 1) #define AP_PROTO_DEVICE_RESET (0x09 << 1) #define AP_PROTO_UDMA_IN (0x0a << 1) #define AP_PROTO_UDMA_OUT (0x0b << 1) #define AP_PROTO_FPDMA (0x0c << 1) #define AP_PROTO_RESP_INFO (0x0f << 1) #define AP_MULTI 0xe0 u_int8_t flags; #define AP_T_LEN 0x03 #define AP_BB 0x04 #define AP_T_DIR 0x08 #define AP_CK_COND 0x20 #define AP_OFFLINE 0x60 u_int8_t features; u_int8_t sector_count; u_int8_t lba_low; u_int8_t lba_mid; u_int8_t lba_high; u_int8_t device; u_int8_t command; u_int8_t reserved; u_int8_t control; }; struct scsi_maintenance_in { uint8_t opcode; uint8_t byte2; #define SERVICE_ACTION_MASK 0x1f #define SA_RPRT_TRGT_GRP 0x0a uint8_t reserved[4]; uint8_t length[4]; uint8_t reserved1; uint8_t control; }; struct scsi_report_supported_opcodes { uint8_t opcode; uint8_t service_action; uint8_t options; #define RSO_RCTD 0x80 #define RSO_OPTIONS_MASK 0x07 #define RSO_OPTIONS_ALL 0x00 #define RSO_OPTIONS_OC 0x01 #define RSO_OPTIONS_OC_SA 0x02 uint8_t requested_opcode; uint8_t requested_service_action[2]; uint8_t length[4]; uint8_t reserved1; uint8_t control; }; struct scsi_report_supported_opcodes_timeout { uint8_t length[2]; uint8_t reserved; uint8_t cmd_specific; uint8_t nominal_time[4]; uint8_t recommended_time[4]; }; struct scsi_report_supported_opcodes_descr { uint8_t opcode; uint8_t reserved; uint8_t service_action[2]; uint8_t reserved2; uint8_t flags; #define RSO_SERVACTV 0x01 #define RSO_CTDP 0x02 uint8_t cdb_length[2]; struct scsi_report_supported_opcodes_timeout timeout[0]; }; struct scsi_report_supported_opcodes_all { uint8_t length[4]; struct scsi_report_supported_opcodes_descr descr[0]; }; struct scsi_report_supported_opcodes_one { uint8_t reserved; uint8_t support; #define RSO_ONE_CTDP 0x80 uint8_t cdb_length[2]; uint8_t cdb_usage[]; }; struct scsi_report_supported_tmf { uint8_t opcode; uint8_t service_action; uint8_t reserved[4]; uint8_t length[4]; uint8_t reserved1; uint8_t control; }; struct scsi_report_supported_tmf_data { uint8_t byte1; #define RST_WAKES 0x01 #define RST_TRS 0x02 #define RST_QTS 0x04 #define RST_LURS 0x08 #define RST_CTSS 0x10 #define RST_CACAS 0x20 #define RST_ATSS 0x40 #define RST_ATS 0x80 uint8_t byte2; #define RST_ITNRS 0x01 #define RST_QTSS 0x02 #define RST_QAES 0x04 uint8_t reserved[2]; }; struct scsi_report_timestamp { uint8_t opcode; uint8_t service_action; uint8_t reserved[4]; uint8_t length[4]; uint8_t reserved1; uint8_t control; }; struct scsi_report_timestamp_data { uint8_t length[2]; uint8_t origin; #define RTS_ORIG_MASK 0x00 #define RTS_ORIG_ZERO 0x00 #define RTS_ORIG_SET 0x02 #define RTS_ORIG_OUTSIDE 0x03 uint8_t reserved; uint8_t timestamp[6]; uint8_t reserve2[2]; }; struct scsi_receive_copy_status_lid1 { uint8_t opcode; uint8_t service_action; #define RCS_RCS_LID1 0x00 uint8_t list_identifier; uint8_t reserved[7]; uint8_t length[4]; uint8_t reserved1; uint8_t control; }; struct scsi_receive_copy_status_lid1_data { uint8_t available_data[4]; uint8_t copy_command_status; #define RCS_CCS_INPROG 0x00 #define RCS_CCS_COMPLETED 0x01 #define RCS_CCS_ERROR 0x02 uint8_t segments_processed[2]; uint8_t transfer_count_units; #define RCS_TC_BYTES 0x00 #define RCS_TC_KBYTES 0x01 #define RCS_TC_MBYTES 0x02 #define RCS_TC_GBYTES 0x03 #define RCS_TC_TBYTES 0x04 #define RCS_TC_PBYTES 0x05 #define RCS_TC_EBYTES 0x06 #define RCS_TC_LBAS 0xf1 uint8_t transfer_count[4]; }; struct scsi_receive_copy_failure_details { uint8_t opcode; uint8_t service_action; #define RCS_RCFD 0x04 uint8_t list_identifier; uint8_t reserved[7]; uint8_t length[4]; uint8_t reserved1; uint8_t control; }; struct scsi_receive_copy_failure_details_data { uint8_t available_data[4]; uint8_t reserved[52]; uint8_t copy_command_status; uint8_t reserved2; uint8_t sense_data_length[2]; uint8_t sense_data[]; }; struct scsi_receive_copy_status_lid4 { uint8_t opcode; uint8_t service_action; #define RCS_RCS_LID4 0x05 uint8_t list_identifier[4]; uint8_t reserved[4]; uint8_t length[4]; uint8_t reserved1; uint8_t control; }; struct scsi_receive_copy_status_lid4_data { uint8_t available_data[4]; uint8_t response_to_service_action; uint8_t copy_command_status; #define RCS_CCS_COMPLETED_PROD 0x03 #define RCS_CCS_COMPLETED_RESID 0x04 #define RCS_CCS_INPROG_FGBG 0x10 #define RCS_CCS_INPROG_FG 0x11 #define RCS_CCS_INPROG_BG 0x12 #define RCS_CCS_ABORTED 0x60 uint8_t operation_counter[2]; uint8_t estimated_status_update_delay[4]; uint8_t extended_copy_completion_status; uint8_t length_of_the_sense_data_field; uint8_t sense_data_length; uint8_t transfer_count_units; uint8_t transfer_count[8]; uint8_t segments_processed[2]; uint8_t reserved[6]; uint8_t sense_data[]; }; struct scsi_receive_copy_operating_parameters { uint8_t opcode; uint8_t service_action; #define RCS_RCOP 0x03 uint8_t reserved[8]; uint8_t length[4]; uint8_t reserved1; uint8_t control; }; struct scsi_receive_copy_operating_parameters_data { uint8_t length[4]; uint8_t snlid; #define RCOP_SNLID 0x01 uint8_t reserved[3]; uint8_t maximum_cscd_descriptor_count[2]; uint8_t maximum_segment_descriptor_count[2]; uint8_t maximum_descriptor_list_length[4]; uint8_t maximum_segment_length[4]; uint8_t maximum_inline_data_length[4]; uint8_t held_data_limit[4]; uint8_t maximum_stream_device_transfer_size[4]; uint8_t reserved2[2]; uint8_t total_concurrent_copies[2]; uint8_t maximum_concurrent_copies; uint8_t data_segment_granularity; uint8_t inline_data_granularity; uint8_t held_data_granularity; uint8_t reserved3[3]; uint8_t implemented_descriptor_list_length; uint8_t list_of_implemented_descriptor_type_codes[0]; }; struct scsi_extended_copy { uint8_t opcode; uint8_t service_action; #define EC_EC_LID1 0x00 #define EC_EC_LID4 0x01 uint8_t reserved[8]; uint8_t length[4]; uint8_t reserved1; uint8_t control; }; struct scsi_ec_cscd_dtsp { uint8_t flags; #define EC_CSCD_FIXED 0x01 #define EC_CSCD_PAD 0x04 uint8_t block_length[3]; }; struct scsi_ec_cscd { uint8_t type_code; #define EC_CSCD_EXT 0xff uint8_t luidt_pdt; #define EC_LUIDT_MASK 0xc0 #define EC_LUIDT_LUN 0x00 #define EC_LUIDT_PROXY_TOKEN 0x40 uint8_t relative_initiator_port[2]; uint8_t cscd_params[24]; struct scsi_ec_cscd_dtsp dtsp; }; struct scsi_ec_cscd_id { uint8_t type_code; #define EC_CSCD_ID 0xe4 uint8_t luidt_pdt; uint8_t relative_initiator_port[2]; uint8_t codeset; uint8_t id_type; uint8_t reserved; uint8_t length; uint8_t designator[20]; struct scsi_ec_cscd_dtsp dtsp; }; struct scsi_ec_segment { uint8_t type_code; uint8_t flags; #define EC_SEG_DC 0x02 #define EC_SEG_CAT 0x01 uint8_t descr_length[2]; uint8_t params[]; }; struct scsi_ec_segment_b2b { uint8_t type_code; #define EC_SEG_B2B 0x02 uint8_t flags; uint8_t descr_length[2]; uint8_t src_cscd[2]; uint8_t dst_cscd[2]; uint8_t reserved[2]; uint8_t number_of_blocks[2]; uint8_t src_lba[8]; uint8_t dst_lba[8]; }; struct scsi_ec_segment_verify { uint8_t type_code; #define EC_SEG_VERIFY 0x07 uint8_t reserved; uint8_t descr_length[2]; uint8_t src_cscd[2]; uint8_t reserved2[2]; uint8_t tur; uint8_t reserved3[3]; }; struct scsi_ec_segment_register_key { uint8_t type_code; #define EC_SEG_REGISTER_KEY 0x14 uint8_t reserved; uint8_t descr_length[2]; uint8_t reserved2[2]; uint8_t dst_cscd[2]; uint8_t res_key[8]; uint8_t sa_res_key[8]; uint8_t reserved3[4]; }; struct scsi_extended_copy_lid1_data { uint8_t list_identifier; uint8_t flags; #define EC_PRIORITY 0x07 #define EC_LIST_ID_USAGE_MASK 0x18 #define EC_LIST_ID_USAGE_FULL 0x08 #define EC_LIST_ID_USAGE_NOHOLD 0x10 #define EC_LIST_ID_USAGE_NONE 0x18 #define EC_STR 0x20 uint8_t cscd_list_length[2]; uint8_t reserved[4]; uint8_t segment_list_length[4]; uint8_t inline_data_length[4]; uint8_t data[]; }; struct scsi_extended_copy_lid4_data { uint8_t list_format; #define EC_LIST_FORMAT 0x01 uint8_t flags; uint8_t header_cscd_list_length[2]; uint8_t reserved[11]; uint8_t flags2; #define EC_IMMED 0x01 #define EC_G_SENSE 0x02 uint8_t header_cscd_type_code; uint8_t reserved2[3]; uint8_t list_identifier[4]; uint8_t reserved3[18]; uint8_t cscd_list_length[2]; uint8_t segment_list_length[2]; uint8_t inline_data_length[2]; uint8_t data[]; }; struct scsi_copy_operation_abort { uint8_t opcode; uint8_t service_action; #define EC_COA 0x1c uint8_t list_identifier[4]; uint8_t reserved[9]; uint8_t control; }; struct scsi_populate_token { uint8_t opcode; uint8_t service_action; #define EC_PT 0x10 uint8_t reserved[4]; uint8_t list_identifier[4]; uint8_t length[4]; uint8_t group_number; uint8_t control; }; struct scsi_range_desc { uint8_t lba[8]; uint8_t length[4]; uint8_t reserved[4]; }; struct scsi_populate_token_data { uint8_t length[2]; uint8_t flags; #define EC_PT_IMMED 0x01 #define EC_PT_RTV 0x02 uint8_t reserved; uint8_t inactivity_timeout[4]; uint8_t rod_type[4]; uint8_t reserved2[2]; uint8_t range_descriptor_length[2]; struct scsi_range_desc desc[]; }; struct scsi_write_using_token { uint8_t opcode; uint8_t service_action; #define EC_WUT 0x11 uint8_t reserved[4]; uint8_t list_identifier[4]; uint8_t length[4]; uint8_t group_number; uint8_t control; }; struct scsi_write_using_token_data { uint8_t length[2]; uint8_t flags; #define EC_WUT_IMMED 0x01 #define EC_WUT_DEL_TKN 0x02 uint8_t reserved[5]; uint8_t offset_into_rod[8]; uint8_t rod_token[512]; uint8_t reserved2[6]; uint8_t range_descriptor_length[2]; struct scsi_range_desc desc[]; }; struct scsi_receive_rod_token_information { uint8_t opcode; uint8_t service_action; #define RCS_RRTI 0x07 uint8_t list_identifier[4]; uint8_t reserved[4]; uint8_t length[4]; uint8_t reserved2; uint8_t control; }; struct scsi_token { uint8_t type[4]; #define ROD_TYPE_INTERNAL 0x00000000 #define ROD_TYPE_AUR 0x00001000 #define ROD_TYPE_PIT_DEF 0x00080000 #define ROD_TYPE_PIT_VULN 0x00080001 #define ROD_TYPE_PIT_PERS 0x00080002 #define ROD_TYPE_PIT_ANY 0x0008FFFF #define ROD_TYPE_BLOCK_ZERO 0xFFFF0001 uint8_t reserved[2]; uint8_t length[2]; uint8_t body[0]; }; struct scsi_report_all_rod_tokens { uint8_t opcode; uint8_t service_action; #define RCS_RART 0x08 uint8_t reserved[8]; uint8_t length[4]; uint8_t reserved2; uint8_t control; }; struct scsi_report_all_rod_tokens_data { uint8_t available_data[4]; uint8_t reserved[4]; uint8_t rod_management_token_list[]; }; struct ata_pass_16 { u_int8_t opcode; u_int8_t protocol; #define AP_EXTEND 0x01 u_int8_t flags; #define AP_FLAG_TLEN_NO_DATA (0 << 0) #define AP_FLAG_TLEN_FEAT (1 << 0) #define AP_FLAG_TLEN_SECT_CNT (2 << 0) #define AP_FLAG_TLEN_STPSIU (3 << 0) #define AP_FLAG_BYT_BLOK_BYTES (0 << 2) #define AP_FLAG_BYT_BLOK_BLOCKS (1 << 2) #define AP_FLAG_TDIR_TO_DEV (0 << 3) #define AP_FLAG_TDIR_FROM_DEV (1 << 3) #define AP_FLAG_CHK_COND (1 << 5) u_int8_t features_ext; u_int8_t features; u_int8_t sector_count_ext; u_int8_t sector_count; u_int8_t lba_low_ext; u_int8_t lba_low; u_int8_t lba_mid_ext; u_int8_t lba_mid; u_int8_t lba_high_ext; u_int8_t lba_high; u_int8_t device; u_int8_t command; u_int8_t control; }; #define SC_SCSI_1 0x01 #define SC_SCSI_2 0x03 /* * Opcodes */ #define TEST_UNIT_READY 0x00 #define REQUEST_SENSE 0x03 #define READ_6 0x08 #define WRITE_6 0x0A #define INQUIRY 0x12 #define MODE_SELECT_6 0x15 #define MODE_SENSE_6 0x1A #define START_STOP_UNIT 0x1B #define START_STOP 0x1B #define RESERVE 0x16 #define RELEASE 0x17 #define RECEIVE_DIAGNOSTIC 0x1C #define SEND_DIAGNOSTIC 0x1D #define PREVENT_ALLOW 0x1E #define READ_CAPACITY 0x25 #define READ_10 0x28 #define WRITE_10 0x2A #define POSITION_TO_ELEMENT 0x2B #define WRITE_VERIFY_10 0x2E #define VERIFY_10 0x2F #define SYNCHRONIZE_CACHE 0x35 #define READ_DEFECT_DATA_10 0x37 #define WRITE_BUFFER 0x3B #define READ_BUFFER 0x3C #define CHANGE_DEFINITION 0x40 #define WRITE_SAME_10 0x41 #define UNMAP 0x42 #define LOG_SELECT 0x4C #define LOG_SENSE 0x4D #define MODE_SELECT_10 0x55 #define RESERVE_10 0x56 #define RELEASE_10 0x57 #define MODE_SENSE_10 0x5A #define PERSISTENT_RES_IN 0x5E #define PERSISTENT_RES_OUT 0x5F #define EXTENDED_COPY 0x83 #define RECEIVE_COPY_STATUS 0x84 #define ATA_PASS_16 0x85 #define READ_16 0x88 #define COMPARE_AND_WRITE 0x89 #define WRITE_16 0x8A #define WRITE_VERIFY_16 0x8E #define VERIFY_16 0x8F #define SYNCHRONIZE_CACHE_16 0x91 #define WRITE_SAME_16 0x93 #define SERVICE_ACTION_IN 0x9E #define REPORT_LUNS 0xA0 #define ATA_PASS_12 0xA1 #define MAINTENANCE_IN 0xA3 #define MAINTENANCE_OUT 0xA4 #define MOVE_MEDIUM 0xA5 #define READ_12 0xA8 #define WRITE_12 0xAA #define WRITE_VERIFY_12 0xAE #define VERIFY_12 0xAF #define READ_ELEMENT_STATUS 0xB8 #define READ_CD 0xBE /* Maintenance In Service Action Codes */ #define REPORT_IDENTIFYING_INFRMATION 0x05 #define REPORT_TARGET_PORT_GROUPS 0x0A #define REPORT_ALIASES 0x0B #define REPORT_SUPPORTED_OPERATION_CODES 0x0C #define REPORT_SUPPORTED_TASK_MANAGEMENT_FUNCTIONS 0x0D #define REPORT_PRIORITY 0x0E #define REPORT_TIMESTAMP 0x0F #define MANAGEMENT_PROTOCOL_IN 0x10 /* Maintenance Out Service Action Codes */ #define SET_IDENTIFY_INFORMATION 0x06 #define SET_TARGET_PORT_GROUPS 0x0A #define CHANGE_ALIASES 0x0B #define SET_PRIORITY 0x0E #define SET_TIMESTAMP 0x0F #define MANGAEMENT_PROTOCOL_OUT 0x10 /* * Device Types */ #define T_DIRECT 0x00 #define T_SEQUENTIAL 0x01 #define T_PRINTER 0x02 #define T_PROCESSOR 0x03 #define T_WORM 0x04 #define T_CDROM 0x05 #define T_SCANNER 0x06 #define T_OPTICAL 0x07 #define T_CHANGER 0x08 #define T_COMM 0x09 #define T_ASC0 0x0a #define T_ASC1 0x0b #define T_STORARRAY 0x0c #define T_ENCLOSURE 0x0d #define T_RBC 0x0e #define T_OCRW 0x0f #define T_OSD 0x11 #define T_ADC 0x12 #define T_NODEVICE 0x1f #define T_ANY 0xff /* Used in Quirk table matches */ #define T_REMOV 1 #define T_FIXED 0 /* * This length is the initial inquiry length used by the probe code, as * well as the length necessary for scsi_print_inquiry() to function * correctly. If either use requires a different length in the future, * the two values should be de-coupled. */ #define SHORT_INQUIRY_LENGTH 36 struct scsi_inquiry_data { u_int8_t device; #define SID_TYPE(inq_data) ((inq_data)->device & 0x1f) #define SID_QUAL(inq_data) (((inq_data)->device & 0xE0) >> 5) #define SID_QUAL_LU_CONNECTED 0x00 /* * The specified peripheral device * type is currently connected to * logical unit. If the target cannot * determine whether or not a physical * device is currently connected, it * shall also use this peripheral * qualifier when returning the INQUIRY * data. This peripheral qualifier * does not mean that the device is * ready for access by the initiator. */ #define SID_QUAL_LU_OFFLINE 0x01 /* * The target is capable of supporting * the specified peripheral device type * on this logical unit; however, the * physical device is not currently * connected to this logical unit. */ #define SID_QUAL_RSVD 0x02 #define SID_QUAL_BAD_LU 0x03 /* * The target is not capable of * supporting a physical device on * this logical unit. For this * peripheral qualifier the peripheral * device type shall be set to 1Fh to * provide compatibility with previous * versions of SCSI. All other * peripheral device type values are * reserved for this peripheral * qualifier. */ #define SID_QUAL_IS_VENDOR_UNIQUE(inq_data) ((SID_QUAL(inq_data) & 0x08) != 0) u_int8_t dev_qual2; #define SID_QUAL2 0x7F #define SID_LU_CONG 0x40 #define SID_RMB 0x80 #define SID_IS_REMOVABLE(inq_data) (((inq_data)->dev_qual2 & SID_RMB) != 0) u_int8_t version; #define SID_ANSI_REV(inq_data) ((inq_data)->version & 0x07) #define SCSI_REV_0 0 #define SCSI_REV_CCS 1 #define SCSI_REV_2 2 #define SCSI_REV_SPC 3 #define SCSI_REV_SPC2 4 #define SCSI_REV_SPC3 5 #define SCSI_REV_SPC4 6 #define SID_ECMA 0x38 #define SID_ISO 0xC0 u_int8_t response_format; #define SID_AENC 0x80 #define SID_TrmIOP 0x40 #define SID_NormACA 0x20 #define SID_HiSup 0x10 u_int8_t additional_length; #define SID_ADDITIONAL_LENGTH(iqd) \ ((iqd)->additional_length + \ __offsetof(struct scsi_inquiry_data, additional_length) + 1) u_int8_t spc3_flags; #define SPC3_SID_PROTECT 0x01 #define SPC3_SID_3PC 0x08 #define SPC3_SID_TPGS_MASK 0x30 #define SPC3_SID_TPGS_IMPLICIT 0x10 #define SPC3_SID_TPGS_EXPLICIT 0x20 #define SPC3_SID_ACC 0x40 #define SPC3_SID_SCCS 0x80 u_int8_t spc2_flags; #define SPC2_SID_ADDR16 0x01 #define SPC2_SID_MChngr 0x08 #define SPC2_SID_MultiP 0x10 #define SPC2_SID_EncServ 0x40 #define SPC2_SID_BQueue 0x80 #define INQ_DATA_TQ_ENABLED(iqd) \ ((SID_ANSI_REV(iqd) < SCSI_REV_SPC2)? ((iqd)->flags & SID_CmdQue) : \ (((iqd)->flags & SID_CmdQue) && !((iqd)->spc2_flags & SPC2_SID_BQueue)) || \ (!((iqd)->flags & SID_CmdQue) && ((iqd)->spc2_flags & SPC2_SID_BQueue))) u_int8_t flags; #define SID_SftRe 0x01 #define SID_CmdQue 0x02 #define SID_Linked 0x08 #define SID_Sync 0x10 #define SID_WBus16 0x20 #define SID_WBus32 0x40 #define SID_RelAdr 0x80 #define SID_VENDOR_SIZE 8 char vendor[SID_VENDOR_SIZE]; #define SID_PRODUCT_SIZE 16 char product[SID_PRODUCT_SIZE]; #define SID_REVISION_SIZE 4 char revision[SID_REVISION_SIZE]; /* * The following fields were taken from SCSI Primary Commands - 2 * (SPC-2) Revision 14, Dated 11 November 1999 */ #define SID_VENDOR_SPECIFIC_0_SIZE 20 u_int8_t vendor_specific0[SID_VENDOR_SPECIFIC_0_SIZE]; /* * An extension of SCSI Parallel Specific Values */ #define SID_SPI_IUS 0x01 #define SID_SPI_QAS 0x02 #define SID_SPI_CLOCK_ST 0x00 #define SID_SPI_CLOCK_DT 0x04 #define SID_SPI_CLOCK_DT_ST 0x0C #define SID_SPI_MASK 0x0F u_int8_t spi3data; u_int8_t reserved2; /* * Version Descriptors, stored 2 byte values. */ u_int8_t version1[2]; u_int8_t version2[2]; u_int8_t version3[2]; u_int8_t version4[2]; u_int8_t version5[2]; u_int8_t version6[2]; u_int8_t version7[2]; u_int8_t version8[2]; u_int8_t reserved3[22]; #define SID_VENDOR_SPECIFIC_1_SIZE 160 u_int8_t vendor_specific1[SID_VENDOR_SPECIFIC_1_SIZE]; }; /* * This structure is more suited to initiator operation, because the * maximum number of supported pages is already allocated. */ struct scsi_vpd_supported_page_list { u_int8_t device; u_int8_t page_code; #define SVPD_SUPPORTED_PAGE_LIST 0x00 #define SVPD_SUPPORTED_PAGES_HDR_LEN 4 u_int8_t reserved; u_int8_t length; /* number of VPD entries */ #define SVPD_SUPPORTED_PAGES_SIZE 251 u_int8_t list[SVPD_SUPPORTED_PAGES_SIZE]; }; /* * This structure is more suited to target operation, because the * number of supported pages is left to the user to allocate. */ struct scsi_vpd_supported_pages { u_int8_t device; u_int8_t page_code; u_int8_t reserved; #define SVPD_SUPPORTED_PAGES 0x00 u_int8_t length; u_int8_t page_list[0]; }; struct scsi_vpd_unit_serial_number { u_int8_t device; u_int8_t page_code; #define SVPD_UNIT_SERIAL_NUMBER 0x80 u_int8_t reserved; u_int8_t length; /* serial number length */ #define SVPD_SERIAL_NUM_SIZE 251 u_int8_t serial_num[SVPD_SERIAL_NUM_SIZE]; }; struct scsi_vpd_device_id { u_int8_t device; u_int8_t page_code; #define SVPD_DEVICE_ID 0x83 #define SVPD_DEVICE_ID_MAX_SIZE 252 #define SVPD_DEVICE_ID_HDR_LEN \ __offsetof(struct scsi_vpd_device_id, desc_list) u_int8_t length[2]; u_int8_t desc_list[]; }; struct scsi_vpd_id_descriptor { u_int8_t proto_codeset; /* * See the SCSI_PROTO definitions above for the protocols. */ #define SVPD_ID_PROTO_SHIFT 4 #define SVPD_ID_CODESET_BINARY 0x01 #define SVPD_ID_CODESET_ASCII 0x02 #define SVPD_ID_CODESET_UTF8 0x03 #define SVPD_ID_CODESET_MASK 0x0f u_int8_t id_type; #define SVPD_ID_PIV 0x80 #define SVPD_ID_ASSOC_LUN 0x00 #define SVPD_ID_ASSOC_PORT 0x10 #define SVPD_ID_ASSOC_TARGET 0x20 #define SVPD_ID_ASSOC_MASK 0x30 #define SVPD_ID_TYPE_VENDOR 0x00 #define SVPD_ID_TYPE_T10 0x01 #define SVPD_ID_TYPE_EUI64 0x02 #define SVPD_ID_TYPE_NAA 0x03 #define SVPD_ID_TYPE_RELTARG 0x04 #define SVPD_ID_TYPE_TPORTGRP 0x05 #define SVPD_ID_TYPE_LUNGRP 0x06 #define SVPD_ID_TYPE_MD5_LUN_ID 0x07 #define SVPD_ID_TYPE_SCSI_NAME 0x08 #define SVPD_ID_TYPE_MASK 0x0f u_int8_t reserved; u_int8_t length; #define SVPD_DEVICE_ID_DESC_HDR_LEN \ __offsetof(struct scsi_vpd_id_descriptor, identifier) u_int8_t identifier[]; }; struct scsi_vpd_id_t10 { u_int8_t vendor[8]; u_int8_t vendor_spec_id[0]; }; struct scsi_vpd_id_eui64 { u_int8_t ieee_company_id[3]; u_int8_t extension_id[5]; }; struct scsi_vpd_id_naa_basic { uint8_t naa; /* big endian, packed: uint8_t naa : 4; uint8_t naa_desig : 4; */ #define SVPD_ID_NAA_NAA_SHIFT 4 #define SVPD_ID_NAA_IEEE_EXT 0x02 #define SVPD_ID_NAA_LOCAL_REG 0x03 #define SVPD_ID_NAA_IEEE_REG 0x05 #define SVPD_ID_NAA_IEEE_REG_EXT 0x06 uint8_t naa_data[]; }; struct scsi_vpd_id_naa_ieee_extended_id { uint8_t naa; uint8_t vendor_specific_id_a; uint8_t ieee_company_id[3]; uint8_t vendor_specific_id_b[4]; }; struct scsi_vpd_id_naa_local_reg { uint8_t naa; uint8_t local_value[7]; }; struct scsi_vpd_id_naa_ieee_reg { uint8_t naa; uint8_t reg_value[7]; /* big endian, packed: uint8_t naa_basic : 4; uint8_t ieee_company_id_0 : 4; uint8_t ieee_company_id_1[2]; uint8_t ieee_company_id_2 : 4; uint8_t vendor_specific_id_0 : 4; uint8_t vendor_specific_id_1[4]; */ }; struct scsi_vpd_id_naa_ieee_reg_extended { uint8_t naa; uint8_t reg_value[15]; /* big endian, packed: uint8_t naa_basic : 4; uint8_t ieee_company_id_0 : 4; uint8_t ieee_company_id_1[2]; uint8_t ieee_company_id_2 : 4; uint8_t vendor_specific_id_0 : 4; uint8_t vendor_specific_id_1[4]; uint8_t vendor_specific_id_ext[8]; */ }; struct scsi_vpd_id_rel_trgt_port_id { uint8_t obsolete[2]; uint8_t rel_trgt_port_id[2]; }; struct scsi_vpd_id_trgt_port_grp_id { uint8_t reserved[2]; uint8_t trgt_port_grp[2]; }; struct scsi_vpd_id_lun_grp_id { uint8_t reserved[2]; uint8_t log_unit_grp[2]; }; struct scsi_vpd_id_md5_lun_id { uint8_t lun_id[16]; }; struct scsi_vpd_id_scsi_name { uint8_t name_string[256]; }; struct scsi_service_action_in { uint8_t opcode; uint8_t service_action; uint8_t action_dependent[13]; uint8_t control; }; struct scsi_diag_page { uint8_t page_code; uint8_t page_specific_flags; uint8_t length[2]; uint8_t params[0]; }; struct scsi_vpd_port_designation { uint8_t reserved[2]; uint8_t relative_port_id[2]; uint8_t reserved2[2]; uint8_t initiator_transportid_length[2]; uint8_t initiator_transportid[0]; }; struct scsi_vpd_port_designation_cont { uint8_t reserved[2]; uint8_t target_port_descriptors_length[2]; struct scsi_vpd_id_descriptor target_port_descriptors[0]; }; struct scsi_vpd_scsi_ports { u_int8_t device; u_int8_t page_code; #define SVPD_SCSI_PORTS 0x88 u_int8_t page_length[2]; struct scsi_vpd_port_designation design[]; }; /* * ATA Information VPD Page based on * T10/2126-D Revision 04 */ #define SVPD_ATA_INFORMATION 0x89 struct scsi_vpd_tpc_descriptor { uint8_t desc_type[2]; uint8_t desc_length[2]; uint8_t parameters[]; }; struct scsi_vpd_tpc_descriptor_bdrl { uint8_t desc_type[2]; #define SVPD_TPC_BDRL 0x0000 uint8_t desc_length[2]; uint8_t vendor_specific[6]; uint8_t maximum_ranges[2]; uint8_t maximum_inactivity_timeout[4]; uint8_t default_inactivity_timeout[4]; uint8_t maximum_token_transfer_size[8]; uint8_t optimal_transfer_count[8]; }; struct scsi_vpd_tpc_descriptor_sc_descr { uint8_t opcode; uint8_t sa_length; uint8_t supported_service_actions[0]; }; struct scsi_vpd_tpc_descriptor_sc { uint8_t desc_type[2]; #define SVPD_TPC_SC 0x0001 uint8_t desc_length[2]; uint8_t list_length; struct scsi_vpd_tpc_descriptor_sc_descr descr[]; }; struct scsi_vpd_tpc_descriptor_pd { uint8_t desc_type[2]; #define SVPD_TPC_PD 0x0004 uint8_t desc_length[2]; uint8_t reserved[4]; uint8_t maximum_cscd_descriptor_count[2]; uint8_t maximum_segment_descriptor_count[2]; uint8_t maximum_descriptor_list_length[4]; uint8_t maximum_inline_data_length[4]; uint8_t reserved2[12]; }; struct scsi_vpd_tpc_descriptor_sd { uint8_t desc_type[2]; #define SVPD_TPC_SD 0x0008 uint8_t desc_length[2]; uint8_t list_length; uint8_t supported_descriptor_codes[]; }; struct scsi_vpd_tpc_descriptor_sdid { uint8_t desc_type[2]; #define SVPD_TPC_SDID 0x000C uint8_t desc_length[2]; uint8_t list_length[2]; uint8_t supported_descriptor_ids[]; }; struct scsi_vpd_tpc_descriptor_rtf_block { uint8_t type_format; #define SVPD_TPC_RTF_BLOCK 0x00 uint8_t reserved; uint8_t desc_length[2]; uint8_t reserved2[2]; uint8_t optimal_length_granularity[2]; uint8_t maximum_bytes[8]; uint8_t optimal_bytes[8]; uint8_t optimal_bytes_to_token_per_segment[8]; uint8_t optimal_bytes_from_token_per_segment[8]; uint8_t reserved3[8]; }; struct scsi_vpd_tpc_descriptor_rtf { uint8_t desc_type[2]; #define SVPD_TPC_RTF 0x0106 uint8_t desc_length[2]; uint8_t remote_tokens; uint8_t reserved[11]; uint8_t minimum_token_lifetime[4]; uint8_t maximum_token_lifetime[4]; uint8_t maximum_token_inactivity_timeout[4]; uint8_t reserved2[18]; uint8_t type_specific_features_length[2]; uint8_t type_specific_features[0]; }; struct scsi_vpd_tpc_descriptor_srtd { uint8_t rod_type[4]; uint8_t flags; #define SVPD_TPC_SRTD_TOUT 0x01 #define SVPD_TPC_SRTD_TIN 0x02 #define SVPD_TPC_SRTD_ECPY 0x80 uint8_t reserved; uint8_t preference_indicator[2]; uint8_t reserved2[56]; }; struct scsi_vpd_tpc_descriptor_srt { uint8_t desc_type[2]; #define SVPD_TPC_SRT 0x0108 uint8_t desc_length[2]; uint8_t reserved[2]; uint8_t rod_type_descriptors_length[2]; uint8_t rod_type_descriptors[0]; }; struct scsi_vpd_tpc_descriptor_gco { uint8_t desc_type[2]; #define SVPD_TPC_GCO 0x8001 uint8_t desc_length[2]; uint8_t total_concurrent_copies[4]; uint8_t maximum_identified_concurrent_copies[4]; uint8_t maximum_segment_length[4]; uint8_t data_segment_granularity; uint8_t inline_data_granularity; uint8_t reserved[18]; }; struct scsi_vpd_tpc { uint8_t device; uint8_t page_code; #define SVPD_SCSI_TPC 0x8F uint8_t page_length[2]; struct scsi_vpd_tpc_descriptor descr[]; }; /* * Block Device Characteristics VPD Page based on * T10/1799-D Revision 31 */ struct scsi_vpd_block_characteristics { u_int8_t device; u_int8_t page_code; #define SVPD_BDC 0xB1 u_int8_t page_length[2]; u_int8_t medium_rotation_rate[2]; #define SVPD_BDC_RATE_NOT_REPORTED 0x00 -#define SVPD_BDC_RATE_NONE_ROTATING 0x01 +#define SVPD_BDC_RATE_NON_ROTATING 0x01 u_int8_t reserved1; u_int8_t nominal_form_factor; #define SVPD_BDC_FORM_NOT_REPORTED 0x00 #define SVPD_BDC_FORM_5_25INCH 0x01 #define SVPD_BDC_FORM_3_5INCH 0x02 #define SVPD_BDC_FORM_2_5INCH 0x03 #define SVPD_BDC_FORM_1_5INCH 0x04 #define SVPD_BDC_FORM_LESSTHAN_1_5INCH 0x05 u_int8_t reserved2[56]; }; /* * Block Device Characteristics VPD Page */ struct scsi_vpd_block_device_characteristics { uint8_t device; uint8_t page_code; #define SVPD_BDC 0xB1 uint8_t page_length[2]; uint8_t medium_rotation_rate[2]; #define SVPD_NOT_REPORTED 0x0000 #define SVPD_NON_ROTATING 0x0001 uint8_t product_type; uint8_t wab_wac_ff; uint8_t flags; #define SVPD_VBULS 0x01 #define SVPD_FUAB 0x02 #define SVPD_HAW_ZBC 0x10 uint8_t reserved[55]; }; /* * Logical Block Provisioning VPD Page based on * T10/1799-D Revision 31 */ struct scsi_vpd_logical_block_prov { u_int8_t device; u_int8_t page_code; #define SVPD_LBP 0xB2 u_int8_t page_length[2]; #define SVPD_LBP_PL_BASIC 0x04 u_int8_t threshold_exponent; u_int8_t flags; #define SVPD_LBP_UNMAP 0x80 #define SVPD_LBP_WS16 0x40 #define SVPD_LBP_WS10 0x20 #define SVPD_LBP_RZ 0x04 #define SVPD_LBP_ANC_SUP 0x02 #define SVPD_LBP_DP 0x01 u_int8_t prov_type; #define SVPD_LBP_RESOURCE 0x01 #define SVPD_LBP_THIN 0x02 u_int8_t reserved; /* * Provisioning Group Descriptor can be here if SVPD_LBP_DP is set * Its size can be determined from page_length - 4 */ }; /* * Block Limits VDP Page based on * T10/1799-D Revision 31 */ struct scsi_vpd_block_limits { u_int8_t device; u_int8_t page_code; #define SVPD_BLOCK_LIMITS 0xB0 u_int8_t page_length[2]; #define SVPD_BL_PL_BASIC 0x10 #define SVPD_BL_PL_TP 0x3C u_int8_t reserved1; u_int8_t max_cmp_write_len; u_int8_t opt_txfer_len_grain[2]; u_int8_t max_txfer_len[4]; u_int8_t opt_txfer_len[4]; u_int8_t max_prefetch[4]; u_int8_t max_unmap_lba_cnt[4]; u_int8_t max_unmap_blk_cnt[4]; u_int8_t opt_unmap_grain[4]; u_int8_t unmap_grain_align[4]; u_int8_t max_write_same_length[8]; u_int8_t reserved2[20]; }; struct scsi_read_capacity { u_int8_t opcode; u_int8_t byte2; #define SRC_RELADR 0x01 u_int8_t addr[4]; u_int8_t unused[2]; u_int8_t pmi; #define SRC_PMI 0x01 u_int8_t control; }; struct scsi_read_capacity_16 { uint8_t opcode; #define SRC16_SERVICE_ACTION 0x10 uint8_t service_action; uint8_t addr[8]; uint8_t alloc_len[4]; #define SRC16_PMI 0x01 #define SRC16_RELADR 0x02 uint8_t reladr; uint8_t control; }; struct scsi_read_capacity_data { u_int8_t addr[4]; u_int8_t length[4]; }; struct scsi_read_capacity_data_long { uint8_t addr[8]; uint8_t length[4]; #define SRC16_PROT_EN 0x01 #define SRC16_P_TYPE 0x0e #define SRC16_PTYPE_1 0x00 #define SRC16_PTYPE_2 0x02 #define SRC16_PTYPE_3 0x04 uint8_t prot; #define SRC16_LBPPBE 0x0f #define SRC16_PI_EXPONENT 0xf0 #define SRC16_PI_EXPONENT_SHIFT 4 uint8_t prot_lbppbe; #define SRC16_LALBA 0x3f #define SRC16_LBPRZ 0x40 #define SRC16_LBPME 0x80 /* * Alternate versions of these macros that are intended for use on a 16-bit * version of the lalba_lbp field instead of the array of 2 8 bit numbers. */ #define SRC16_LALBA_A 0x3fff #define SRC16_LBPRZ_A 0x4000 #define SRC16_LBPME_A 0x8000 uint8_t lalba_lbp[2]; uint8_t reserved[16]; }; struct scsi_report_luns { uint8_t opcode; uint8_t reserved1; #define RPL_REPORT_DEFAULT 0x00 #define RPL_REPORT_WELLKNOWN 0x01 #define RPL_REPORT_ALL 0x02 uint8_t select_report; uint8_t reserved2[3]; uint8_t length[4]; uint8_t reserved3; uint8_t control; }; struct scsi_report_luns_lundata { uint8_t lundata[8]; #define RPL_LUNDATA_PERIPH_BUS_MASK 0x3f #define RPL_LUNDATA_FLAT_LUN_MASK 0x3f #define RPL_LUNDATA_FLAT_LUN_BITS 0x06 #define RPL_LUNDATA_LUN_TARG_MASK 0x3f #define RPL_LUNDATA_LUN_BUS_MASK 0xe0 #define RPL_LUNDATA_LUN_LUN_MASK 0x1f #define RPL_LUNDATA_EXT_LEN_MASK 0x30 #define RPL_LUNDATA_EXT_EAM_MASK 0x0f #define RPL_LUNDATA_EXT_EAM_WK 0x01 #define RPL_LUNDATA_EXT_EAM_NOT_SPEC 0x0f #define RPL_LUNDATA_ATYP_MASK 0xc0 /* MBZ for type 0 lun */ #define RPL_LUNDATA_ATYP_PERIPH 0x00 #define RPL_LUNDATA_ATYP_FLAT 0x40 #define RPL_LUNDATA_ATYP_LUN 0x80 #define RPL_LUNDATA_ATYP_EXTLUN 0xc0 }; struct scsi_report_luns_data { u_int8_t length[4]; /* length of LUN inventory, in bytes */ u_int8_t reserved[4]; /* unused */ /* * LUN inventory- we only support the type zero form for now. */ struct scsi_report_luns_lundata luns[0]; }; struct scsi_target_group { uint8_t opcode; uint8_t service_action; #define STG_PDF_MASK 0xe0 #define STG_PDF_LENGTH 0x00 #define STG_PDF_EXTENDED 0x20 uint8_t reserved1[4]; uint8_t length[4]; uint8_t reserved2; uint8_t control; }; struct scsi_target_port_descriptor { uint8_t reserved[2]; uint8_t relative_target_port_identifier[2]; uint8_t desc_list[]; }; struct scsi_target_port_group_descriptor { uint8_t pref_state; #define TPG_PRIMARY 0x80 #define TPG_ASYMMETRIC_ACCESS_STATE_MASK 0xf #define TPG_ASYMMETRIC_ACCESS_OPTIMIZED 0x0 #define TPG_ASYMMETRIC_ACCESS_NONOPTIMIZED 0x1 #define TPG_ASYMMETRIC_ACCESS_STANDBY 0x2 #define TPG_ASYMMETRIC_ACCESS_UNAVAILABLE 0x3 #define TPG_ASYMMETRIC_ACCESS_LBA_DEPENDENT 0x4 #define TPG_ASYMMETRIC_ACCESS_OFFLINE 0xE #define TPG_ASYMMETRIC_ACCESS_TRANSITIONING 0xF uint8_t support; #define TPG_AO_SUP 0x01 #define TPG_AN_SUP 0x02 #define TPG_S_SUP 0x04 #define TPG_U_SUP 0x08 #define TPG_LBD_SUP 0x10 #define TPG_O_SUP 0x40 #define TPG_T_SUP 0x80 uint8_t target_port_group[2]; uint8_t reserved; uint8_t status; #define TPG_UNAVLBL 0 #define TPG_SET_BY_STPG 0x01 #define TPG_IMPLICIT 0x02 uint8_t vendor_specific; uint8_t target_port_count; struct scsi_target_port_descriptor descriptors[]; }; struct scsi_target_group_data { uint8_t length[4]; /* length of returned data, in bytes */ struct scsi_target_port_group_descriptor groups[]; }; struct scsi_target_group_data_extended { uint8_t length[4]; /* length of returned data, in bytes */ uint8_t format_type; /* STG_PDF_LENGTH or STG_PDF_EXTENDED */ uint8_t implicit_transition_time; uint8_t reserved[2]; struct scsi_target_port_group_descriptor groups[]; }; typedef enum { SSD_TYPE_NONE, SSD_TYPE_FIXED, SSD_TYPE_DESC } scsi_sense_data_type; typedef enum { SSD_ELEM_NONE, SSD_ELEM_SKIP, SSD_ELEM_DESC, SSD_ELEM_SKS, SSD_ELEM_COMMAND, SSD_ELEM_INFO, SSD_ELEM_FRU, SSD_ELEM_STREAM, SSD_ELEM_MAX } scsi_sense_elem_type; struct scsi_sense_data { uint8_t error_code; /* * SPC-4 says that the maximum length of sense data is 252 bytes. * So this structure is exactly 252 bytes log. */ #define SSD_FULL_SIZE 252 uint8_t sense_buf[SSD_FULL_SIZE - 1]; /* * XXX KDM is this still a reasonable minimum size? */ #define SSD_MIN_SIZE 18 /* * Maximum value for the extra_len field in the sense data. */ #define SSD_EXTRA_MAX 244 }; /* * Fixed format sense data. */ struct scsi_sense_data_fixed { u_int8_t error_code; #define SSD_ERRCODE 0x7F #define SSD_CURRENT_ERROR 0x70 #define SSD_DEFERRED_ERROR 0x71 #define SSD_ERRCODE_VALID 0x80 u_int8_t segment; u_int8_t flags; #define SSD_KEY 0x0F #define SSD_KEY_NO_SENSE 0x00 #define SSD_KEY_RECOVERED_ERROR 0x01 #define SSD_KEY_NOT_READY 0x02 #define SSD_KEY_MEDIUM_ERROR 0x03 #define SSD_KEY_HARDWARE_ERROR 0x04 #define SSD_KEY_ILLEGAL_REQUEST 0x05 #define SSD_KEY_UNIT_ATTENTION 0x06 #define SSD_KEY_DATA_PROTECT 0x07 #define SSD_KEY_BLANK_CHECK 0x08 #define SSD_KEY_Vendor_Specific 0x09 #define SSD_KEY_COPY_ABORTED 0x0a #define SSD_KEY_ABORTED_COMMAND 0x0b #define SSD_KEY_EQUAL 0x0c #define SSD_KEY_VOLUME_OVERFLOW 0x0d #define SSD_KEY_MISCOMPARE 0x0e #define SSD_KEY_COMPLETED 0x0f #define SSD_ILI 0x20 #define SSD_EOM 0x40 #define SSD_FILEMARK 0x80 u_int8_t info[4]; u_int8_t extra_len; u_int8_t cmd_spec_info[4]; u_int8_t add_sense_code; u_int8_t add_sense_code_qual; u_int8_t fru; u_int8_t sense_key_spec[3]; #define SSD_SCS_VALID 0x80 #define SSD_FIELDPTR_CMD 0x40 #define SSD_BITPTR_VALID 0x08 #define SSD_BITPTR_VALUE 0x07 u_int8_t extra_bytes[14]; #define SSD_FIXED_IS_PRESENT(sense, length, field) \ ((length >= (offsetof(struct scsi_sense_data_fixed, field) + \ sizeof(sense->field))) ? 1 :0) #define SSD_FIXED_IS_FILLED(sense, field) \ ((((offsetof(struct scsi_sense_data_fixed, field) + \ sizeof(sense->field)) - \ (offsetof(struct scsi_sense_data_fixed, extra_len) + \ sizeof(sense->extra_len))) <= sense->extra_len) ? 1 : 0) }; /* * Descriptor format sense data definitions. * Introduced in SPC-3. */ struct scsi_sense_data_desc { uint8_t error_code; #define SSD_DESC_CURRENT_ERROR 0x72 #define SSD_DESC_DEFERRED_ERROR 0x73 uint8_t sense_key; uint8_t add_sense_code; uint8_t add_sense_code_qual; uint8_t reserved[3]; /* * Note that SPC-4, section 4.5.2.1 says that the extra_len field * must be less than or equal to 244. */ uint8_t extra_len; uint8_t sense_desc[0]; #define SSD_DESC_IS_PRESENT(sense, length, field) \ ((length >= (offsetof(struct scsi_sense_data_desc, field) + \ sizeof(sense->field))) ? 1 :0) }; struct scsi_sense_desc_header { uint8_t desc_type; uint8_t length; }; /* * The information provide in the Information descriptor is device type or * command specific information, and defined in a command standard. * * Note that any changes to the field names or positions in this structure, * even reserved fields, should be accompanied by an examination of the * code in ctl_set_sense() that uses them. * * Maximum descriptors allowed: 1 (as of SPC-4) */ struct scsi_sense_info { uint8_t desc_type; #define SSD_DESC_INFO 0x00 uint8_t length; uint8_t byte2; #define SSD_INFO_VALID 0x80 uint8_t reserved; uint8_t info[8]; }; /* * Command-specific information depends on the command for which the * reported condition occured. * * Note that any changes to the field names or positions in this structure, * even reserved fields, should be accompanied by an examination of the * code in ctl_set_sense() that uses them. * * Maximum descriptors allowed: 1 (as of SPC-4) */ struct scsi_sense_command { uint8_t desc_type; #define SSD_DESC_COMMAND 0x01 uint8_t length; uint8_t reserved[2]; uint8_t command_info[8]; }; /* * Sense key specific descriptor. The sense key specific data format * depends on the sense key in question. * * Maximum descriptors allowed: 1 (as of SPC-4) */ struct scsi_sense_sks { uint8_t desc_type; #define SSD_DESC_SKS 0x02 uint8_t length; uint8_t reserved1[2]; uint8_t sense_key_spec[3]; #define SSD_SKS_VALID 0x80 uint8_t reserved2; }; /* * This is used for the Illegal Request sense key (0x05) only. */ struct scsi_sense_sks_field { uint8_t byte0; #define SSD_SKS_FIELD_VALID 0x80 #define SSD_SKS_FIELD_CMD 0x40 #define SSD_SKS_BPV 0x08 #define SSD_SKS_BIT_VALUE 0x07 uint8_t field[2]; }; /* * This is used for the Hardware Error (0x04), Medium Error (0x03) and * Recovered Error (0x01) sense keys. */ struct scsi_sense_sks_retry { uint8_t byte0; #define SSD_SKS_RETRY_VALID 0x80 uint8_t actual_retry_count[2]; }; /* * Used with the NO Sense (0x00) or Not Ready (0x02) sense keys. */ struct scsi_sense_sks_progress { uint8_t byte0; #define SSD_SKS_PROGRESS_VALID 0x80 uint8_t progress[2]; #define SSD_SKS_PROGRESS_DENOM 0x10000 }; /* * Used with the Copy Aborted (0x0a) sense key. */ struct scsi_sense_sks_segment { uint8_t byte0; #define SSD_SKS_SEGMENT_VALID 0x80 #define SSD_SKS_SEGMENT_SD 0x20 #define SSD_SKS_SEGMENT_BPV 0x08 #define SSD_SKS_SEGMENT_BITPTR 0x07 uint8_t field[2]; }; /* * Used with the Unit Attention (0x06) sense key. * * This is currently used to indicate that the unit attention condition * queue has overflowed (when the overflow bit is set). */ struct scsi_sense_sks_overflow { uint8_t byte0; #define SSD_SKS_OVERFLOW_VALID 0x80 #define SSD_SKS_OVERFLOW_SET 0x01 uint8_t reserved[2]; }; /* * This specifies which component is associated with the sense data. There * is no standard meaning for the fru value. * * Maximum descriptors allowed: 1 (as of SPC-4) */ struct scsi_sense_fru { uint8_t desc_type; #define SSD_DESC_FRU 0x03 uint8_t length; uint8_t reserved; uint8_t fru; }; /* * Used for Stream commands, defined in SSC-4. * * Maximum descriptors allowed: 1 (as of SPC-4) */ struct scsi_sense_stream { uint8_t desc_type; #define SSD_DESC_STREAM 0x04 uint8_t length; uint8_t reserved; uint8_t byte3; #define SSD_DESC_STREAM_FM 0x80 #define SSD_DESC_STREAM_EOM 0x40 #define SSD_DESC_STREAM_ILI 0x20 }; /* * Used for Block commands, defined in SBC-3. * * This is currently (as of SBC-3) only used for the Incorrect Length * Indication (ILI) bit, which says that the data length requested in the * READ LONG or WRITE LONG command did not match the length of the logical * block. * * Maximum descriptors allowed: 1 (as of SPC-4) */ struct scsi_sense_block { uint8_t desc_type; #define SSD_DESC_BLOCK 0x05 uint8_t length; uint8_t reserved; uint8_t byte3; #define SSD_DESC_BLOCK_ILI 0x20 }; /* * Used for Object-Based Storage Devices (OSD-3). * * Maximum descriptors allowed: 1 (as of SPC-4) */ struct scsi_sense_osd_objid { uint8_t desc_type; #define SSD_DESC_OSD_OBJID 0x06 uint8_t length; uint8_t reserved[6]; /* * XXX KDM provide the bit definitions here? There are a lot of * them, and we don't have an OSD driver yet. */ uint8_t not_init_cmds[4]; uint8_t completed_cmds[4]; uint8_t partition_id[8]; uint8_t object_id[8]; }; /* * Used for Object-Based Storage Devices (OSD-3). * * Maximum descriptors allowed: 1 (as of SPC-4) */ struct scsi_sense_osd_integrity { uint8_t desc_type; #define SSD_DESC_OSD_INTEGRITY 0x07 uint8_t length; uint8_t integ_check_val[32]; }; /* * Used for Object-Based Storage Devices (OSD-3). * * Maximum descriptors allowed: 1 (as of SPC-4) */ struct scsi_sense_osd_attr_id { uint8_t desc_type; #define SSD_DESC_OSD_ATTR_ID 0x08 uint8_t length; uint8_t reserved[2]; uint8_t attr_desc[0]; }; /* * Used with Sense keys No Sense (0x00) and Not Ready (0x02). * * Maximum descriptors allowed: 32 (as of SPC-4) */ struct scsi_sense_progress { uint8_t desc_type; #define SSD_DESC_PROGRESS 0x0a uint8_t length; uint8_t sense_key; uint8_t add_sense_code; uint8_t add_sense_code_qual; uint8_t reserved; uint8_t progress[2]; }; /* * This is typically forwarded as the result of an EXTENDED COPY command. * * Maximum descriptors allowed: 2 (as of SPC-4) */ struct scsi_sense_forwarded { uint8_t desc_type; #define SSD_DESC_FORWARDED 0x0c uint8_t length; uint8_t byte2; #define SSD_FORWARDED_FSDT 0x80 #define SSD_FORWARDED_SDS_MASK 0x0f #define SSD_FORWARDED_SDS_UNK 0x00 #define SSD_FORWARDED_SDS_EXSRC 0x01 #define SSD_FORWARDED_SDS_EXDST 0x02 }; /* * Vendor-specific sense descriptor. The desc_type field will be in the * range bewteen MIN and MAX inclusive. */ struct scsi_sense_vendor { uint8_t desc_type; #define SSD_DESC_VENDOR_MIN 0x80 #define SSD_DESC_VENDOR_MAX 0xff uint8_t length; uint8_t data[0]; }; struct scsi_mode_header_6 { u_int8_t data_length; /* Sense data length */ u_int8_t medium_type; u_int8_t dev_spec; u_int8_t blk_desc_len; }; struct scsi_mode_header_10 { u_int8_t data_length[2];/* Sense data length */ u_int8_t medium_type; u_int8_t dev_spec; u_int8_t unused[2]; u_int8_t blk_desc_len[2]; }; struct scsi_mode_page_header { u_int8_t page_code; #define SMPH_PS 0x80 #define SMPH_SPF 0x40 #define SMPH_PC_MASK 0x3f u_int8_t page_length; }; struct scsi_mode_page_header_sp { uint8_t page_code; uint8_t subpage; uint8_t page_length[2]; }; struct scsi_mode_blk_desc { u_int8_t density; u_int8_t nblocks[3]; u_int8_t reserved; u_int8_t blklen[3]; }; #define SCSI_DEFAULT_DENSITY 0x00 /* use 'default' density */ #define SCSI_SAME_DENSITY 0x7f /* use 'same' density- >= SCSI-2 only */ /* * Status Byte */ #define SCSI_STATUS_OK 0x00 #define SCSI_STATUS_CHECK_COND 0x02 #define SCSI_STATUS_COND_MET 0x04 #define SCSI_STATUS_BUSY 0x08 #define SCSI_STATUS_INTERMED 0x10 #define SCSI_STATUS_INTERMED_COND_MET 0x14 #define SCSI_STATUS_RESERV_CONFLICT 0x18 #define SCSI_STATUS_CMD_TERMINATED 0x22 /* Obsolete in SAM-2 */ #define SCSI_STATUS_QUEUE_FULL 0x28 #define SCSI_STATUS_ACA_ACTIVE 0x30 #define SCSI_STATUS_TASK_ABORTED 0x40 struct scsi_inquiry_pattern { u_int8_t type; u_int8_t media_type; #define SIP_MEDIA_REMOVABLE 0x01 #define SIP_MEDIA_FIXED 0x02 const char *vendor; const char *product; const char *revision; }; struct scsi_static_inquiry_pattern { u_int8_t type; u_int8_t media_type; char vendor[SID_VENDOR_SIZE+1]; char product[SID_PRODUCT_SIZE+1]; char revision[SID_REVISION_SIZE+1]; }; struct scsi_sense_quirk_entry { struct scsi_inquiry_pattern inq_pat; int num_sense_keys; int num_ascs; struct sense_key_table_entry *sense_key_info; struct asc_table_entry *asc_info; }; struct sense_key_table_entry { u_int8_t sense_key; u_int32_t action; const char *desc; }; struct asc_table_entry { u_int8_t asc; u_int8_t ascq; u_int32_t action; const char *desc; }; struct op_table_entry { u_int8_t opcode; u_int32_t opmask; const char *desc; }; struct scsi_op_quirk_entry { struct scsi_inquiry_pattern inq_pat; int num_ops; struct op_table_entry *op_table; }; typedef enum { SSS_FLAG_NONE = 0x00, SSS_FLAG_PRINT_COMMAND = 0x01 } scsi_sense_string_flags; struct scsi_nv { const char *name; uint64_t value; }; typedef enum { SCSI_NV_FOUND, SCSI_NV_AMBIGUOUS, SCSI_NV_NOT_FOUND } scsi_nv_status; typedef enum { SCSI_NV_FLAG_NONE = 0x00, SCSI_NV_FLAG_IG_CASE = 0x01 /* Case insensitive comparison */ } scsi_nv_flags; struct ccb_scsiio; struct cam_periph; union ccb; #ifndef _KERNEL struct cam_device; #endif extern const char *scsi_sense_key_text[]; struct sbuf; __BEGIN_DECLS void scsi_sense_desc(int sense_key, int asc, int ascq, struct scsi_inquiry_data *inq_data, const char **sense_key_desc, const char **asc_desc); scsi_sense_action scsi_error_action(struct ccb_scsiio* csio, struct scsi_inquiry_data *inq_data, u_int32_t sense_flags); const char * scsi_status_string(struct ccb_scsiio *csio); void scsi_desc_iterate(struct scsi_sense_data_desc *sense, u_int sense_len, int (*iter_func)(struct scsi_sense_data_desc *sense, u_int, struct scsi_sense_desc_header *, void *), void *arg); uint8_t *scsi_find_desc(struct scsi_sense_data_desc *sense, u_int sense_len, uint8_t desc_type); void scsi_set_sense_data(struct scsi_sense_data *sense_data, scsi_sense_data_type sense_format, int current_error, int sense_key, int asc, int ascq, ...) ; void scsi_set_sense_data_va(struct scsi_sense_data *sense_data, scsi_sense_data_type sense_format, int current_error, int sense_key, int asc, int ascq, va_list ap); int scsi_get_sense_info(struct scsi_sense_data *sense_data, u_int sense_len, uint8_t info_type, uint64_t *info, int64_t *signed_info); int scsi_get_sks(struct scsi_sense_data *sense_data, u_int sense_len, uint8_t *sks); int scsi_get_block_info(struct scsi_sense_data *sense_data, u_int sense_len, struct scsi_inquiry_data *inq_data, uint8_t *block_bits); int scsi_get_stream_info(struct scsi_sense_data *sense_data, u_int sense_len, struct scsi_inquiry_data *inq_data, uint8_t *stream_bits); void scsi_info_sbuf(struct sbuf *sb, uint8_t *cdb, int cdb_len, struct scsi_inquiry_data *inq_data, uint64_t info); void scsi_command_sbuf(struct sbuf *sb, uint8_t *cdb, int cdb_len, struct scsi_inquiry_data *inq_data, uint64_t csi); void scsi_progress_sbuf(struct sbuf *sb, uint16_t progress); int scsi_sks_sbuf(struct sbuf *sb, int sense_key, uint8_t *sks); void scsi_fru_sbuf(struct sbuf *sb, uint64_t fru); void scsi_stream_sbuf(struct sbuf *sb, uint8_t stream_bits, uint64_t info); void scsi_block_sbuf(struct sbuf *sb, uint8_t block_bits, uint64_t info); void scsi_sense_info_sbuf(struct sbuf *sb, struct scsi_sense_data *sense, u_int sense_len, uint8_t *cdb, int cdb_len, struct scsi_inquiry_data *inq_data, struct scsi_sense_desc_header *header); void scsi_sense_command_sbuf(struct sbuf *sb, struct scsi_sense_data *sense, u_int sense_len, uint8_t *cdb, int cdb_len, struct scsi_inquiry_data *inq_data, struct scsi_sense_desc_header *header); void scsi_sense_sks_sbuf(struct sbuf *sb, struct scsi_sense_data *sense, u_int sense_len, uint8_t *cdb, int cdb_len, struct scsi_inquiry_data *inq_data, struct scsi_sense_desc_header *header); void scsi_sense_fru_sbuf(struct sbuf *sb, struct scsi_sense_data *sense, u_int sense_len, uint8_t *cdb, int cdb_len, struct scsi_inquiry_data *inq_data, struct scsi_sense_desc_header *header); void scsi_sense_stream_sbuf(struct sbuf *sb, struct scsi_sense_data *sense, u_int sense_len, uint8_t *cdb, int cdb_len, struct scsi_inquiry_data *inq_data, struct scsi_sense_desc_header *header); void scsi_sense_block_sbuf(struct sbuf *sb, struct scsi_sense_data *sense, u_int sense_len, uint8_t *cdb, int cdb_len, struct scsi_inquiry_data *inq_data, struct scsi_sense_desc_header *header); void scsi_sense_progress_sbuf(struct sbuf *sb, struct scsi_sense_data *sense, u_int sense_len, uint8_t *cdb, int cdb_len, struct scsi_inquiry_data *inq_data, struct scsi_sense_desc_header *header); void scsi_sense_generic_sbuf(struct sbuf *sb, struct scsi_sense_data *sense, u_int sense_len, uint8_t *cdb, int cdb_len, struct scsi_inquiry_data *inq_data, struct scsi_sense_desc_header *header); void scsi_sense_desc_sbuf(struct sbuf *sb, struct scsi_sense_data *sense, u_int sense_len, uint8_t *cdb, int cdb_len, struct scsi_inquiry_data *inq_data, struct scsi_sense_desc_header *header); scsi_sense_data_type scsi_sense_type(struct scsi_sense_data *sense_data); void scsi_sense_only_sbuf(struct scsi_sense_data *sense, u_int sense_len, struct sbuf *sb, char *path_str, struct scsi_inquiry_data *inq_data, uint8_t *cdb, int cdb_len); #ifdef _KERNEL int scsi_command_string(struct ccb_scsiio *csio, struct sbuf *sb); int scsi_sense_sbuf(struct ccb_scsiio *csio, struct sbuf *sb, scsi_sense_string_flags flags); char * scsi_sense_string(struct ccb_scsiio *csio, char *str, int str_len); void scsi_sense_print(struct ccb_scsiio *csio); int scsi_vpd_supported_page(struct cam_periph *periph, uint8_t page_id); #else /* _KERNEL */ int scsi_command_string(struct cam_device *device, struct ccb_scsiio *csio, struct sbuf *sb); int scsi_sense_sbuf(struct cam_device *device, struct ccb_scsiio *csio, struct sbuf *sb, scsi_sense_string_flags flags); char * scsi_sense_string(struct cam_device *device, struct ccb_scsiio *csio, char *str, int str_len); void scsi_sense_print(struct cam_device *device, struct ccb_scsiio *csio, FILE *ofile); #endif /* _KERNEL */ const char * scsi_op_desc(u_int16_t opcode, struct scsi_inquiry_data *inq_data); char * scsi_cdb_string(u_int8_t *cdb_ptr, char *cdb_string, size_t len); void scsi_print_inquiry(struct scsi_inquiry_data *inq_data); void scsi_print_inquiry_short(struct scsi_inquiry_data *inq_data); u_int scsi_calc_syncsrate(u_int period_factor); u_int scsi_calc_syncparam(u_int period); typedef int (*scsi_devid_checkfn_t)(uint8_t *); int scsi_devid_is_naa_ieee_reg(uint8_t *bufp); int scsi_devid_is_sas_target(uint8_t *bufp); int scsi_devid_is_lun_eui64(uint8_t *bufp); int scsi_devid_is_lun_naa(uint8_t *bufp); int scsi_devid_is_lun_name(uint8_t *bufp); int scsi_devid_is_lun_t10(uint8_t *bufp); struct scsi_vpd_id_descriptor * scsi_get_devid(struct scsi_vpd_device_id *id, uint32_t len, scsi_devid_checkfn_t ck_fn); struct scsi_vpd_id_descriptor * scsi_get_devid_desc(struct scsi_vpd_id_descriptor *desc, uint32_t len, scsi_devid_checkfn_t ck_fn); int scsi_transportid_sbuf(struct sbuf *sb, struct scsi_transportid_header *hdr, uint32_t valid_len); const char * scsi_nv_to_str(struct scsi_nv *table, int num_table_entries, uint64_t value); scsi_nv_status scsi_get_nv(struct scsi_nv *table, int num_table_entries, char *name, int *table_entry, scsi_nv_flags flags); int scsi_parse_transportid_64bit(int proto_id, char *id_str, struct scsi_transportid_header **hdr, unsigned int *alloc_len, #ifdef _KERNEL struct malloc_type *type, int flags, #endif char *error_str, int error_str_len); int scsi_parse_transportid_spi(char *id_str, struct scsi_transportid_header **hdr, unsigned int *alloc_len, #ifdef _KERNEL struct malloc_type *type, int flags, #endif char *error_str, int error_str_len); int scsi_parse_transportid_rdma(char *id_str, struct scsi_transportid_header **hdr, unsigned int *alloc_len, #ifdef _KERNEL struct malloc_type *type, int flags, #endif char *error_str, int error_str_len); int scsi_parse_transportid_iscsi(char *id_str, struct scsi_transportid_header **hdr, unsigned int *alloc_len, #ifdef _KERNEL struct malloc_type *type, int flags, #endif char *error_str,int error_str_len); int scsi_parse_transportid_sop(char *id_str, struct scsi_transportid_header **hdr, unsigned int *alloc_len, #ifdef _KERNEL struct malloc_type *type, int flags, #endif char *error_str,int error_str_len); int scsi_parse_transportid(char *transportid_str, struct scsi_transportid_header **hdr, unsigned int *alloc_len, #ifdef _KERNEL struct malloc_type *type, int flags, #endif char *error_str, int error_str_len); void scsi_test_unit_ready(struct ccb_scsiio *csio, u_int32_t retries, void (*cbfcnp)(struct cam_periph *, union ccb *), u_int8_t tag_action, u_int8_t sense_len, u_int32_t timeout); void scsi_request_sense(struct ccb_scsiio *csio, u_int32_t retries, void (*cbfcnp)(struct cam_periph *, union ccb *), void *data_ptr, u_int8_t dxfer_len, u_int8_t tag_action, u_int8_t sense_len, u_int32_t timeout); void scsi_inquiry(struct ccb_scsiio *csio, u_int32_t retries, void (*cbfcnp)(struct cam_periph *, union ccb *), u_int8_t tag_action, u_int8_t *inq_buf, u_int32_t inq_len, int evpd, u_int8_t page_code, u_int8_t sense_len, u_int32_t timeout); void scsi_mode_sense(struct ccb_scsiio *csio, u_int32_t retries, void (*cbfcnp)(struct cam_periph *, union ccb *), u_int8_t tag_action, int dbd, u_int8_t page_code, u_int8_t page, u_int8_t *param_buf, u_int32_t param_len, u_int8_t sense_len, u_int32_t timeout); void scsi_mode_sense_len(struct ccb_scsiio *csio, u_int32_t retries, void (*cbfcnp)(struct cam_periph *, union ccb *), u_int8_t tag_action, int dbd, u_int8_t page_code, u_int8_t page, u_int8_t *param_buf, u_int32_t param_len, int minimum_cmd_size, u_int8_t sense_len, u_int32_t timeout); void scsi_mode_select(struct ccb_scsiio *csio, u_int32_t retries, void (*cbfcnp)(struct cam_periph *, union ccb *), u_int8_t tag_action, int scsi_page_fmt, int save_pages, u_int8_t *param_buf, u_int32_t param_len, u_int8_t sense_len, u_int32_t timeout); void scsi_mode_select_len(struct ccb_scsiio *csio, u_int32_t retries, void (*cbfcnp)(struct cam_periph *, union ccb *), u_int8_t tag_action, int scsi_page_fmt, int save_pages, u_int8_t *param_buf, u_int32_t param_len, int minimum_cmd_size, u_int8_t sense_len, u_int32_t timeout); void scsi_log_sense(struct ccb_scsiio *csio, u_int32_t retries, void (*cbfcnp)(struct cam_periph *, union ccb *), u_int8_t tag_action, u_int8_t page_code, u_int8_t page, int save_pages, int ppc, u_int32_t paramptr, u_int8_t *param_buf, u_int32_t param_len, u_int8_t sense_len, u_int32_t timeout); void scsi_log_select(struct ccb_scsiio *csio, u_int32_t retries, void (*cbfcnp)(struct cam_periph *, union ccb *), u_int8_t tag_action, u_int8_t page_code, int save_pages, int pc_reset, u_int8_t *param_buf, u_int32_t param_len, u_int8_t sense_len, u_int32_t timeout); void scsi_prevent(struct ccb_scsiio *csio, u_int32_t retries, void (*cbfcnp)(struct cam_periph *, union ccb *), u_int8_t tag_action, u_int8_t action, u_int8_t sense_len, u_int32_t timeout); void scsi_read_capacity(struct ccb_scsiio *csio, u_int32_t retries, void (*cbfcnp)(struct cam_periph *, union ccb *), u_int8_t tag_action, struct scsi_read_capacity_data *, u_int8_t sense_len, u_int32_t timeout); void scsi_read_capacity_16(struct ccb_scsiio *csio, uint32_t retries, void (*cbfcnp)(struct cam_periph *, union ccb *), uint8_t tag_action, uint64_t lba, int reladr, int pmi, uint8_t *rcap_buf, int rcap_buf_len, uint8_t sense_len, uint32_t timeout); void scsi_report_luns(struct ccb_scsiio *csio, u_int32_t retries, void (*cbfcnp)(struct cam_periph *, union ccb *), u_int8_t tag_action, u_int8_t select_report, struct scsi_report_luns_data *rpl_buf, u_int32_t alloc_len, u_int8_t sense_len, u_int32_t timeout); void scsi_report_target_group(struct ccb_scsiio *csio, u_int32_t retries, void (*cbfcnp)(struct cam_periph *, union ccb *), u_int8_t tag_action, u_int8_t pdf, void *buf, u_int32_t alloc_len, u_int8_t sense_len, u_int32_t timeout); void scsi_set_target_group(struct ccb_scsiio *csio, u_int32_t retries, void (*cbfcnp)(struct cam_periph *, union ccb *), u_int8_t tag_action, void *buf, u_int32_t alloc_len, u_int8_t sense_len, u_int32_t timeout); void scsi_synchronize_cache(struct ccb_scsiio *csio, u_int32_t retries, void (*cbfcnp)(struct cam_periph *, union ccb *), u_int8_t tag_action, u_int32_t begin_lba, u_int16_t lb_count, u_int8_t sense_len, u_int32_t timeout); void scsi_receive_diagnostic_results(struct ccb_scsiio *csio, u_int32_t retries, void (*cbfcnp)(struct cam_periph *, union ccb*), uint8_t tag_action, int pcv, uint8_t page_code, uint8_t *data_ptr, uint16_t allocation_length, uint8_t sense_len, uint32_t timeout); void scsi_send_diagnostic(struct ccb_scsiio *csio, u_int32_t retries, void (*cbfcnp)(struct cam_periph *, union ccb *), uint8_t tag_action, int unit_offline, int device_offline, int self_test, int page_format, int self_test_code, uint8_t *data_ptr, uint16_t param_list_length, uint8_t sense_len, uint32_t timeout); void scsi_read_buffer(struct ccb_scsiio *csio, u_int32_t retries, void (*cbfcnp)(struct cam_periph *, union ccb*), uint8_t tag_action, int mode, uint8_t buffer_id, u_int32_t offset, uint8_t *data_ptr, uint32_t allocation_length, uint8_t sense_len, uint32_t timeout); void scsi_write_buffer(struct ccb_scsiio *csio, u_int32_t retries, void (*cbfcnp)(struct cam_periph *, union ccb *), uint8_t tag_action, int mode, uint8_t buffer_id, u_int32_t offset, uint8_t *data_ptr, uint32_t param_list_length, uint8_t sense_len, uint32_t timeout); #define SCSI_RW_READ 0x0001 #define SCSI_RW_WRITE 0x0002 #define SCSI_RW_DIRMASK 0x0003 #define SCSI_RW_BIO 0x1000 void scsi_read_write(struct ccb_scsiio *csio, u_int32_t retries, void (*cbfcnp)(struct cam_periph *, union ccb *), u_int8_t tag_action, int readop, u_int8_t byte2, int minimum_cmd_size, u_int64_t lba, u_int32_t block_count, u_int8_t *data_ptr, u_int32_t dxfer_len, u_int8_t sense_len, u_int32_t timeout); void scsi_write_same(struct ccb_scsiio *csio, u_int32_t retries, void (*cbfcnp)(struct cam_periph *, union ccb *), u_int8_t tag_action, u_int8_t byte2, int minimum_cmd_size, u_int64_t lba, u_int32_t block_count, u_int8_t *data_ptr, u_int32_t dxfer_len, u_int8_t sense_len, u_int32_t timeout); void scsi_ata_identify(struct ccb_scsiio *csio, u_int32_t retries, void (*cbfcnp)(struct cam_periph *, union ccb *), u_int8_t tag_action, u_int8_t *data_ptr, u_int16_t dxfer_len, u_int8_t sense_len, u_int32_t timeout); void scsi_ata_trim(struct ccb_scsiio *csio, u_int32_t retries, void (*cbfcnp)(struct cam_periph *, union ccb *), u_int8_t tag_action, u_int16_t block_count, u_int8_t *data_ptr, u_int16_t dxfer_len, u_int8_t sense_len, u_int32_t timeout); void scsi_ata_pass_16(struct ccb_scsiio *csio, u_int32_t retries, void (*cbfcnp)(struct cam_periph *, union ccb *), u_int32_t flags, u_int8_t tag_action, u_int8_t protocol, u_int8_t ata_flags, u_int16_t features, u_int16_t sector_count, uint64_t lba, u_int8_t command, u_int8_t control, u_int8_t *data_ptr, u_int16_t dxfer_len, u_int8_t sense_len, u_int32_t timeout); void scsi_unmap(struct ccb_scsiio *csio, u_int32_t retries, void (*cbfcnp)(struct cam_periph *, union ccb *), u_int8_t tag_action, u_int8_t byte2, u_int8_t *data_ptr, u_int16_t dxfer_len, u_int8_t sense_len, u_int32_t timeout); void scsi_start_stop(struct ccb_scsiio *csio, u_int32_t retries, void (*cbfcnp)(struct cam_periph *, union ccb *), u_int8_t tag_action, int start, int load_eject, int immediate, u_int8_t sense_len, u_int32_t timeout); void scsi_persistent_reserve_in(struct ccb_scsiio *csio, uint32_t retries, void (*cbfcnp)(struct cam_periph *,union ccb *), uint8_t tag_action, int service_action, uint8_t *data_ptr, uint32_t dxfer_len, int sense_len, int timeout); void scsi_persistent_reserve_out(struct ccb_scsiio *csio, uint32_t retries, void (*cbfcnp)(struct cam_periph *, union ccb *), uint8_t tag_action, int service_action, int scope, int res_type, uint8_t *data_ptr, uint32_t dxfer_len, int sense_len, int timeout); int scsi_inquiry_match(caddr_t inqbuffer, caddr_t table_entry); int scsi_static_inquiry_match(caddr_t inqbuffer, caddr_t table_entry); int scsi_devid_match(uint8_t *rhs, size_t rhs_len, uint8_t *lhs, size_t lhs_len); void scsi_extract_sense(struct scsi_sense_data *sense, int *error_code, int *sense_key, int *asc, int *ascq); int scsi_extract_sense_ccb(union ccb *ccb, int *error_code, int *sense_key, int *asc, int *ascq); void scsi_extract_sense_len(struct scsi_sense_data *sense, u_int sense_len, int *error_code, int *sense_key, int *asc, int *ascq, int show_errors); int scsi_get_sense_key(struct scsi_sense_data *sense, u_int sense_len, int show_errors); int scsi_get_asc(struct scsi_sense_data *sense, u_int sense_len, int show_errors); int scsi_get_ascq(struct scsi_sense_data *sense, u_int sense_len, int show_errors); static __inline void scsi_ulto2b(u_int32_t val, u_int8_t *bytes); static __inline void scsi_ulto3b(u_int32_t val, u_int8_t *bytes); static __inline void scsi_ulto4b(u_int32_t val, u_int8_t *bytes); static __inline void scsi_u64to8b(u_int64_t val, u_int8_t *bytes); static __inline uint32_t scsi_2btoul(const uint8_t *bytes); static __inline uint32_t scsi_3btoul(const uint8_t *bytes); static __inline int32_t scsi_3btol(const uint8_t *bytes); static __inline uint32_t scsi_4btoul(const uint8_t *bytes); static __inline uint64_t scsi_8btou64(const uint8_t *bytes); static __inline void *find_mode_page_6(struct scsi_mode_header_6 *mode_header); static __inline void *find_mode_page_10(struct scsi_mode_header_10 *mode_header); static __inline void scsi_ulto2b(u_int32_t val, u_int8_t *bytes) { bytes[0] = (val >> 8) & 0xff; bytes[1] = val & 0xff; } static __inline void scsi_ulto3b(u_int32_t val, u_int8_t *bytes) { bytes[0] = (val >> 16) & 0xff; bytes[1] = (val >> 8) & 0xff; bytes[2] = val & 0xff; } static __inline void scsi_ulto4b(u_int32_t val, u_int8_t *bytes) { bytes[0] = (val >> 24) & 0xff; bytes[1] = (val >> 16) & 0xff; bytes[2] = (val >> 8) & 0xff; bytes[3] = val & 0xff; } static __inline void scsi_u64to8b(u_int64_t val, u_int8_t *bytes) { bytes[0] = (val >> 56) & 0xff; bytes[1] = (val >> 48) & 0xff; bytes[2] = (val >> 40) & 0xff; bytes[3] = (val >> 32) & 0xff; bytes[4] = (val >> 24) & 0xff; bytes[5] = (val >> 16) & 0xff; bytes[6] = (val >> 8) & 0xff; bytes[7] = val & 0xff; } static __inline uint32_t scsi_2btoul(const uint8_t *bytes) { uint32_t rv; rv = (bytes[0] << 8) | bytes[1]; return (rv); } static __inline uint32_t scsi_3btoul(const uint8_t *bytes) { uint32_t rv; rv = (bytes[0] << 16) | (bytes[1] << 8) | bytes[2]; return (rv); } static __inline int32_t scsi_3btol(const uint8_t *bytes) { uint32_t rc = scsi_3btoul(bytes); if (rc & 0x00800000) rc |= 0xff000000; return (int32_t) rc; } static __inline uint32_t scsi_4btoul(const uint8_t *bytes) { uint32_t rv; rv = (bytes[0] << 24) | (bytes[1] << 16) | (bytes[2] << 8) | bytes[3]; return (rv); } static __inline uint64_t scsi_8btou64(const uint8_t *bytes) { uint64_t rv; rv = (((uint64_t)bytes[0]) << 56) | (((uint64_t)bytes[1]) << 48) | (((uint64_t)bytes[2]) << 40) | (((uint64_t)bytes[3]) << 32) | (((uint64_t)bytes[4]) << 24) | (((uint64_t)bytes[5]) << 16) | (((uint64_t)bytes[6]) << 8) | bytes[7]; return (rv); } /* * Given the pointer to a returned mode sense buffer, return a pointer to * the start of the first mode page. */ static __inline void * find_mode_page_6(struct scsi_mode_header_6 *mode_header) { void *page_start; page_start = (void *)((u_int8_t *)&mode_header[1] + mode_header->blk_desc_len); return(page_start); } static __inline void * find_mode_page_10(struct scsi_mode_header_10 *mode_header) { void *page_start; page_start = (void *)((u_int8_t *)&mode_header[1] + scsi_2btoul(mode_header->blk_desc_len)); return(page_start); } __END_DECLS #endif /*_SCSI_SCSI_ALL_H*/ Index: stable/10/sys/cam/scsi/scsi_da.c =================================================================== --- stable/10/sys/cam/scsi/scsi_da.c (revision 271237) +++ stable/10/sys/cam/scsi/scsi_da.c (revision 271238) @@ -1,3885 +1,3906 @@ /*- * Implementation of SCSI Direct Access Peripheral driver for CAM. * * Copyright (c) 1997 Justin T. Gibbs. * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions, and the following disclaimer, * without modification, immediately at the beginning of the file. * 2. The name of the author may not be used to endorse or promote products * derived from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE FOR * ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include #ifdef _KERNEL #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #endif /* _KERNEL */ #ifndef _KERNEL #include #include #endif /* _KERNEL */ #include #include #include #include #include #include #ifndef _KERNEL #include #endif /* !_KERNEL */ #ifdef _KERNEL typedef enum { DA_STATE_PROBE_RC, DA_STATE_PROBE_RC16, DA_STATE_PROBE_LBP, DA_STATE_PROBE_BLK_LIMITS, DA_STATE_PROBE_BDC, DA_STATE_PROBE_ATA, DA_STATE_NORMAL } da_state; typedef enum { DA_FLAG_PACK_INVALID = 0x001, DA_FLAG_NEW_PACK = 0x002, DA_FLAG_PACK_LOCKED = 0x004, DA_FLAG_PACK_REMOVABLE = 0x008, DA_FLAG_NEED_OTAG = 0x020, DA_FLAG_WAS_OTAG = 0x040, DA_FLAG_RETRY_UA = 0x080, DA_FLAG_OPEN = 0x100, DA_FLAG_SCTX_INIT = 0x200, DA_FLAG_CAN_RC16 = 0x400, DA_FLAG_PROBED = 0x800, DA_FLAG_DIRTY = 0x1000, DA_FLAG_ANNOUNCED = 0x2000 } da_flags; typedef enum { DA_Q_NONE = 0x00, DA_Q_NO_SYNC_CACHE = 0x01, DA_Q_NO_6_BYTE = 0x02, DA_Q_NO_PREVENT = 0x04, DA_Q_4K = 0x08, DA_Q_NO_RC16 = 0x10, DA_Q_NO_UNMAP = 0x20 } da_quirks; #define DA_Q_BIT_STRING \ "\020" \ "\001NO_SYNC_CACHE" \ "\002NO_6_BYTE" \ "\003NO_PREVENT" \ "\0044K" \ "\005NO_RC16" typedef enum { DA_CCB_PROBE_RC = 0x01, DA_CCB_PROBE_RC16 = 0x02, DA_CCB_PROBE_LBP = 0x03, DA_CCB_PROBE_BLK_LIMITS = 0x04, DA_CCB_PROBE_BDC = 0x05, DA_CCB_PROBE_ATA = 0x06, DA_CCB_BUFFER_IO = 0x07, DA_CCB_DUMP = 0x0A, DA_CCB_DELETE = 0x0B, DA_CCB_TUR = 0x0C, DA_CCB_TYPE_MASK = 0x0F, DA_CCB_RETRY_UA = 0x10 } da_ccb_state; /* * Order here is important for method choice * * We prefer ATA_TRIM as tests run against a Sandforce 2281 SSD attached to * LSI 2008 (mps) controller (FW: v12, Drv: v14) resulted 20% quicker deletes * using ATA_TRIM than the corresponding UNMAP results for a real world mysql * import taking 5mins. * */ typedef enum { DA_DELETE_NONE, DA_DELETE_DISABLE, DA_DELETE_ATA_TRIM, DA_DELETE_UNMAP, DA_DELETE_WS16, DA_DELETE_WS10, DA_DELETE_ZERO, DA_DELETE_MIN = DA_DELETE_ATA_TRIM, DA_DELETE_MAX = DA_DELETE_ZERO } da_delete_methods; typedef void da_delete_func_t (struct cam_periph *periph, union ccb *ccb, struct bio *bp); static da_delete_func_t da_delete_trim; static da_delete_func_t da_delete_unmap; static da_delete_func_t da_delete_ws; static const void * da_delete_functions[] = { NULL, NULL, da_delete_trim, da_delete_unmap, da_delete_ws, da_delete_ws, da_delete_ws }; static const char *da_delete_method_names[] = { "NONE", "DISABLE", "ATA_TRIM", "UNMAP", "WS16", "WS10", "ZERO" }; static const char *da_delete_method_desc[] = { "NONE", "DISABLED", "ATA TRIM", "UNMAP", "WRITE SAME(16) with UNMAP", "WRITE SAME(10) with UNMAP", "ZERO" }; /* Offsets into our private area for storing information */ #define ccb_state ppriv_field0 #define ccb_bp ppriv_ptr1 struct disk_params { u_int8_t heads; u_int32_t cylinders; u_int8_t secs_per_track; u_int32_t secsize; /* Number of bytes/sector */ u_int64_t sectors; /* total number sectors */ u_int stripesize; u_int stripeoffset; }; #define UNMAP_RANGE_MAX 0xffffffff #define UNMAP_HEAD_SIZE 8 #define UNMAP_RANGE_SIZE 16 #define UNMAP_MAX_RANGES 2048 /* Protocol Max is 4095 */ #define UNMAP_BUF_SIZE ((UNMAP_MAX_RANGES * UNMAP_RANGE_SIZE) + \ UNMAP_HEAD_SIZE) #define WS10_MAX_BLKS 0xffff #define WS16_MAX_BLKS 0xffffffff #define ATA_TRIM_MAX_RANGES ((UNMAP_BUF_SIZE / \ (ATA_DSM_RANGE_SIZE * ATA_DSM_BLK_SIZE)) * ATA_DSM_BLK_SIZE) struct da_softc { struct bio_queue_head bio_queue; struct bio_queue_head delete_queue; struct bio_queue_head delete_run_queue; LIST_HEAD(, ccb_hdr) pending_ccbs; int tur; /* TEST UNIT READY should be sent */ int refcount; /* Active xpt_action() calls */ da_state state; da_flags flags; da_quirks quirks; int sort_io_queue; int minimum_cmd_size; int error_inject; int trim_max_ranges; int delete_running; int delete_available; /* Delete methods possibly available */ u_int maxio; uint32_t unmap_max_ranges; uint32_t unmap_max_lba; /* Max LBAs in UNMAP req */ uint64_t ws_max_blks; da_delete_methods delete_method; da_delete_func_t *delete_func; struct disk_params params; struct disk *disk; union ccb saved_ccb; struct task sysctl_task; struct sysctl_ctx_list sysctl_ctx; struct sysctl_oid *sysctl_tree; struct callout sendordered_c; uint64_t wwpn; uint8_t unmap_buf[UNMAP_BUF_SIZE]; struct scsi_read_capacity_data_long rcaplong; struct callout mediapoll_c; }; #define dadeleteflag(softc, delete_method, enable) \ if (enable) { \ softc->delete_available |= (1 << delete_method); \ } else { \ softc->delete_available &= ~(1 << delete_method); \ } struct da_quirk_entry { struct scsi_inquiry_pattern inq_pat; da_quirks quirks; }; static const char quantum[] = "QUANTUM"; static const char microp[] = "MICROP"; static struct da_quirk_entry da_quirk_table[] = { /* SPI, FC devices */ { /* * Fujitsu M2513A MO drives. * Tested devices: M2513A2 firmware versions 1200 & 1300. * (dip switch selects whether T_DIRECT or T_OPTICAL device) * Reported by: W.Scholten */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "FUJITSU", "M2513A", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* See above. */ {T_OPTICAL, SIP_MEDIA_REMOVABLE, "FUJITSU", "M2513A", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * This particular Fujitsu drive doesn't like the * synchronize cache command. * Reported by: Tom Jackson */ {T_DIRECT, SIP_MEDIA_FIXED, "FUJITSU", "M2954*", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * This drive doesn't like the synchronize cache command * either. Reported by: Matthew Jacob * in NetBSD PR kern/6027, August 24, 1998. */ {T_DIRECT, SIP_MEDIA_FIXED, microp, "2217*", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * This drive doesn't like the synchronize cache command * either. Reported by: Hellmuth Michaelis (hm@kts.org) * (PR 8882). */ {T_DIRECT, SIP_MEDIA_FIXED, microp, "2112*", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * Doesn't like the synchronize cache command. * Reported by: Blaz Zupan */ {T_DIRECT, SIP_MEDIA_FIXED, "NEC", "D3847*", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * Doesn't like the synchronize cache command. * Reported by: Blaz Zupan */ {T_DIRECT, SIP_MEDIA_FIXED, quantum, "MAVERICK 540S", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * Doesn't like the synchronize cache command. */ {T_DIRECT, SIP_MEDIA_FIXED, quantum, "LPS525S", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * Doesn't like the synchronize cache command. * Reported by: walter@pelissero.de */ {T_DIRECT, SIP_MEDIA_FIXED, quantum, "LPS540S", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * Doesn't work correctly with 6 byte reads/writes. * Returns illegal request, and points to byte 9 of the * 6-byte CDB. * Reported by: Adam McDougall */ {T_DIRECT, SIP_MEDIA_FIXED, quantum, "VIKING 4*", "*"}, /*quirks*/ DA_Q_NO_6_BYTE }, { /* See above. */ {T_DIRECT, SIP_MEDIA_FIXED, quantum, "VIKING 2*", "*"}, /*quirks*/ DA_Q_NO_6_BYTE }, { /* * Doesn't like the synchronize cache command. * Reported by: walter@pelissero.de */ {T_DIRECT, SIP_MEDIA_FIXED, "CONNER", "CP3500*", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * The CISS RAID controllers do not support SYNC_CACHE */ {T_DIRECT, SIP_MEDIA_FIXED, "COMPAQ", "RAID*", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * The STEC 842 sometimes hang on UNMAP. */ {T_DIRECT, SIP_MEDIA_FIXED, "STEC", "S842E800M2", "*"}, /*quirks*/ DA_Q_NO_UNMAP }, /* USB mass storage devices supported by umass(4) */ { /* * EXATELECOM (Sigmatel) i-Bead 100/105 USB Flash MP3 Player * PR: kern/51675 */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "EXATEL", "i-BEAD10*", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * Power Quotient Int. (PQI) USB flash key * PR: kern/53067 */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "Generic*", "USB Flash Disk*", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * Creative Nomad MUVO mp3 player (USB) * PR: kern/53094 */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "CREATIVE", "NOMAD_MUVO", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE|DA_Q_NO_PREVENT }, { /* * Jungsoft NEXDISK USB flash key * PR: kern/54737 */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "JUNGSOFT", "NEXDISK*", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * FreeDik USB Mini Data Drive * PR: kern/54786 */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "FreeDik*", "Mini Data Drive", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * Sigmatel USB Flash MP3 Player * PR: kern/57046 */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "SigmaTel", "MSCN", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE|DA_Q_NO_PREVENT }, { /* * Neuros USB Digital Audio Computer * PR: kern/63645 */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "NEUROS", "dig. audio comp.", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * SEAGRAND NP-900 MP3 Player * PR: kern/64563 */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "SEAGRAND", "NP-900*", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE|DA_Q_NO_PREVENT }, { /* * iRiver iFP MP3 player (with UMS Firmware) * PR: kern/54881, i386/63941, kern/66124 */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "iRiver", "iFP*", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * Frontier Labs NEX IA+ Digital Audio Player, rev 1.10/0.01 * PR: kern/70158 */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "FL" , "Nex*", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * ZICPlay USB MP3 Player with FM * PR: kern/75057 */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "ACTIONS*" , "USB DISK*", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * TEAC USB floppy mechanisms */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "TEAC" , "FD-05*", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * Kingston DataTraveler II+ USB Pen-Drive. * Reported by: Pawel Jakub Dawidek */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "Kingston" , "DataTraveler II+", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * USB DISK Pro PMAP * Reported by: jhs * PR: usb/96381 */ {T_DIRECT, SIP_MEDIA_REMOVABLE, " ", "USB DISK Pro", "PMAP"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * Motorola E398 Mobile Phone (TransFlash memory card). * Reported by: Wojciech A. Koszek * PR: usb/89889 */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "Motorola" , "Motorola Phone", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * Qware BeatZkey! Pro * PR: usb/79164 */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "GENERIC", "USB DISK DEVICE", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * Time DPA20B 1GB MP3 Player * PR: usb/81846 */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "USB2.0*", "(FS) FLASH DISK*", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * Samsung USB key 128Mb * PR: usb/90081 */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "USB-DISK", "FreeDik-FlashUsb", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * Kingston DataTraveler 2.0 USB Flash memory. * PR: usb/89196 */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "Kingston", "DataTraveler 2.0", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * Creative MUVO Slim mp3 player (USB) * PR: usb/86131 */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "CREATIVE", "MuVo Slim", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE|DA_Q_NO_PREVENT }, { /* * United MP5512 Portable MP3 Player (2-in-1 USB DISK/MP3) * PR: usb/80487 */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "Generic*", "MUSIC DISK", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * SanDisk Micro Cruzer 128MB * PR: usb/75970 */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "SanDisk" , "Micro Cruzer", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * TOSHIBA TransMemory USB sticks * PR: kern/94660 */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "TOSHIBA", "TransMemory", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * PNY USB Flash keys * PR: usb/75578, usb/72344, usb/65436 */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "*" , "USB DISK*", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * Genesys 6-in-1 Card Reader * PR: usb/94647 */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "Generic*", "STORAGE DEVICE*", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * Rekam Digital CAMERA * PR: usb/98713 */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "CAMERA*", "4MP-9J6*", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * iRiver H10 MP3 player * PR: usb/102547 */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "iriver", "H10*", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * iRiver U10 MP3 player * PR: usb/92306 */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "iriver", "U10*", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * X-Micro Flash Disk * PR: usb/96901 */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "X-Micro", "Flash Disk", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * EasyMP3 EM732X USB 2.0 Flash MP3 Player * PR: usb/96546 */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "EM732X", "MP3 Player*", "1.00"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * Denver MP3 player * PR: usb/107101 */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "DENVER", "MP3 PLAYER", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * Philips USB Key Audio KEY013 * PR: usb/68412 */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "PHILIPS", "Key*", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE | DA_Q_NO_PREVENT }, { /* * JNC MP3 Player * PR: usb/94439 */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "JNC*" , "MP3 Player*", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * SAMSUNG MP0402H * PR: usb/108427 */ {T_DIRECT, SIP_MEDIA_FIXED, "SAMSUNG", "MP0402H", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * I/O Magic USB flash - Giga Bank * PR: usb/108810 */ {T_DIRECT, SIP_MEDIA_FIXED, "GS-Magic", "stor*", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * JoyFly 128mb USB Flash Drive * PR: 96133 */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "USB 2.0", "Flash Disk*", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * ChipsBnk usb stick * PR: 103702 */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "ChipsBnk", "USB*", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * Storcase (Kingston) InfoStation IFS FC2/SATA-R 201A * PR: 129858 */ {T_DIRECT, SIP_MEDIA_FIXED, "IFS", "FC2/SATA-R*", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * Samsung YP-U3 mp3-player * PR: 125398 */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "Samsung", "YP-U3", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { {T_DIRECT, SIP_MEDIA_REMOVABLE, "Netac", "OnlyDisk*", "2000"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * Sony Cyber-Shot DSC cameras * PR: usb/137035 */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "Sony", "Sony DSC", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE | DA_Q_NO_PREVENT }, { {T_DIRECT, SIP_MEDIA_REMOVABLE, "Kingston", "DataTraveler G3", "1.00"}, /*quirks*/ DA_Q_NO_PREVENT }, { /* At least several Transcent USB sticks lie on RC16. */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "JetFlash", "Transcend*", "*"}, /*quirks*/ DA_Q_NO_RC16 }, /* ATA/SATA devices over SAS/USB/... */ { /* Hitachi Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "Hitachi", "H??????????E3*", "*" }, /*quirks*/DA_Q_4K }, { /* Samsung Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "SAMSUNG HD155UI*", "*" }, /*quirks*/DA_Q_4K }, { /* Samsung Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "SAMSUNG", "HD155UI*", "*" }, /*quirks*/DA_Q_4K }, { /* Samsung Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "SAMSUNG HD204UI*", "*" }, /*quirks*/DA_Q_4K }, { /* Samsung Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "SAMSUNG", "HD204UI*", "*" }, /*quirks*/DA_Q_4K }, { /* Seagate Barracuda Green Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "ST????DL*", "*" }, /*quirks*/DA_Q_4K }, { /* Seagate Barracuda Green Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "ST????DL", "*", "*" }, /*quirks*/DA_Q_4K }, { /* Seagate Barracuda Green Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "ST???DM*", "*" }, /*quirks*/DA_Q_4K }, { /* Seagate Barracuda Green Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "ST???DM*", "*", "*" }, /*quirks*/DA_Q_4K }, { /* Seagate Barracuda Green Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "ST????DM*", "*" }, /*quirks*/DA_Q_4K }, { /* Seagate Barracuda Green Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "ST????DM", "*", "*" }, /*quirks*/DA_Q_4K }, { /* Seagate Momentus Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "ST9500423AS*", "*" }, /*quirks*/DA_Q_4K }, { /* Seagate Momentus Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "ST950042", "3AS*", "*" }, /*quirks*/DA_Q_4K }, { /* Seagate Momentus Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "ST9500424AS*", "*" }, /*quirks*/DA_Q_4K }, { /* Seagate Momentus Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "ST950042", "4AS*", "*" }, /*quirks*/DA_Q_4K }, { /* Seagate Momentus Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "ST9640423AS*", "*" }, /*quirks*/DA_Q_4K }, { /* Seagate Momentus Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "ST964042", "3AS*", "*" }, /*quirks*/DA_Q_4K }, { /* Seagate Momentus Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "ST9640424AS*", "*" }, /*quirks*/DA_Q_4K }, { /* Seagate Momentus Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "ST964042", "4AS*", "*" }, /*quirks*/DA_Q_4K }, { /* Seagate Momentus Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "ST9750420AS*", "*" }, /*quirks*/DA_Q_4K }, { /* Seagate Momentus Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "ST975042", "0AS*", "*" }, /*quirks*/DA_Q_4K }, { /* Seagate Momentus Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "ST9750422AS*", "*" }, /*quirks*/DA_Q_4K }, { /* Seagate Momentus Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "ST975042", "2AS*", "*" }, /*quirks*/DA_Q_4K }, { /* Seagate Momentus Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "ST9750423AS*", "*" }, /*quirks*/DA_Q_4K }, { /* Seagate Momentus Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "ST975042", "3AS*", "*" }, /*quirks*/DA_Q_4K }, { /* Seagate Momentus Thin Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "ST???LT*", "*" }, /*quirks*/DA_Q_4K }, { /* Seagate Momentus Thin Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "ST???LT*", "*", "*" }, /*quirks*/DA_Q_4K }, { /* WDC Caviar Green Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "WDC WD????RS*", "*" }, /*quirks*/DA_Q_4K }, { /* WDC Caviar Green Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "WDC WD??", "??RS*", "*" }, /*quirks*/DA_Q_4K }, { /* WDC Caviar Green Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "WDC WD????RX*", "*" }, /*quirks*/DA_Q_4K }, { /* WDC Caviar Green Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "WDC WD??", "??RX*", "*" }, /*quirks*/DA_Q_4K }, { /* WDC Caviar Green Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "WDC WD??????RS*", "*" }, /*quirks*/DA_Q_4K }, { /* WDC Caviar Green Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "WDC WD??", "????RS*", "*" }, /*quirks*/DA_Q_4K }, { /* WDC Caviar Green Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "WDC WD??????RX*", "*" }, /*quirks*/DA_Q_4K }, { /* WDC Caviar Green Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "WDC WD??", "????RX*", "*" }, /*quirks*/DA_Q_4K }, { /* WDC Scorpio Black Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "WDC WD???PKT*", "*" }, /*quirks*/DA_Q_4K }, { /* WDC Scorpio Black Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "WDC WD??", "?PKT*", "*" }, /*quirks*/DA_Q_4K }, { /* WDC Scorpio Black Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "WDC WD?????PKT*", "*" }, /*quirks*/DA_Q_4K }, { /* WDC Scorpio Black Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "WDC WD??", "???PKT*", "*" }, /*quirks*/DA_Q_4K }, { /* WDC Scorpio Blue Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "WDC WD???PVT*", "*" }, /*quirks*/DA_Q_4K }, { /* WDC Scorpio Blue Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "WDC WD??", "?PVT*", "*" }, /*quirks*/DA_Q_4K }, { /* WDC Scorpio Blue Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "WDC WD?????PVT*", "*" }, /*quirks*/DA_Q_4K }, { /* WDC Scorpio Blue Advanced Format (4k) drives */ { T_DIRECT, SIP_MEDIA_FIXED, "WDC WD??", "???PVT*", "*" }, /*quirks*/DA_Q_4K }, { /* * Olympus FE-210 camera */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "OLYMPUS", "FE210*", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * LG UP3S MP3 player */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "LG", "UP3S", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * Laser MP3-2GA13 MP3 player */ {T_DIRECT, SIP_MEDIA_REMOVABLE, "USB 2.0", "(HS) Flash Disk", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, { /* * LaCie external 250GB Hard drive des by Porsche * Submitted by: Ben Stuyts * PR: 121474 */ {T_DIRECT, SIP_MEDIA_FIXED, "SAMSUNG", "HM250JI", "*"}, /*quirks*/ DA_Q_NO_SYNC_CACHE }, /* SATA SSDs */ { /* * Corsair Force 2 SSDs * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "Corsair CSSD-F*", "*" }, /*quirks*/DA_Q_4K }, { /* * Corsair Force 3 SSDs * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "Corsair Force 3*", "*" }, /*quirks*/DA_Q_4K }, { /* * Corsair Neutron GTX SSDs * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "Corsair Neutron GTX*", "*" }, /*quirks*/DA_Q_4K }, { /* * Corsair Force GT & GS SSDs * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "Corsair Force G*", "*" }, /*quirks*/DA_Q_4K }, { /* * Crucial M4 SSDs * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "M4-CT???M4SSD2*", "*" }, /*quirks*/DA_Q_4K }, { /* * Crucial RealSSD C300 SSDs * 4k optimised */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "C300-CTFDDAC???MAG*", "*" }, /*quirks*/DA_Q_4K }, { /* * Intel 320 Series SSDs * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "INTEL SSDSA2CW*", "*" }, /*quirks*/DA_Q_4K }, { /* * Intel 330 Series SSDs * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "INTEL SSDSC2CT*", "*" }, /*quirks*/DA_Q_4K }, { /* * Intel 510 Series SSDs * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "INTEL SSDSC2MH*", "*" }, /*quirks*/DA_Q_4K }, { /* * Intel 520 Series SSDs * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "INTEL SSDSC2BW*", "*" }, /*quirks*/DA_Q_4K }, { /* * Intel X25-M Series SSDs * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "INTEL SSDSA2M*", "*" }, /*quirks*/DA_Q_4K }, { /* * Kingston E100 Series SSDs * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "KINGSTON SE100S3*", "*" }, /*quirks*/DA_Q_4K }, { /* * Kingston HyperX 3k SSDs * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "KINGSTON SH103S3*", "*" }, /*quirks*/DA_Q_4K }, { /* * Marvell SSDs (entry taken from OpenSolaris) * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "MARVELL SD88SA02*", "*" }, /*quirks*/DA_Q_4K }, { /* * OCZ Agility 2 SSDs * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "*", "OCZ-AGILITY2*", "*" }, /*quirks*/DA_Q_4K }, { /* * OCZ Agility 3 SSDs * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "OCZ-AGILITY3*", "*" }, /*quirks*/DA_Q_4K }, { /* * OCZ Deneva R Series SSDs * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "DENRSTE251M45*", "*" }, /*quirks*/DA_Q_4K }, { /* * OCZ Vertex 2 SSDs (inc pro series) * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "OCZ?VERTEX2*", "*" }, /*quirks*/DA_Q_4K }, { /* * OCZ Vertex 3 SSDs * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "OCZ-VERTEX3*", "*" }, /*quirks*/DA_Q_4K }, { /* * OCZ Vertex 4 SSDs * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "OCZ-VERTEX4*", "*" }, /*quirks*/DA_Q_4K }, { /* * Samsung 830 Series SSDs * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "SAMSUNG SSD 830 Series*", "*" }, /*quirks*/DA_Q_4K }, { /* * Samsung 840 SSDs * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "Samsung SSD 840*", "*" }, /*quirks*/DA_Q_4K }, { /* * SuperTalent TeraDrive CT SSDs * 4k optimised & trim only works in 4k requests + 4k aligned */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "FTM??CT25H*", "*" }, /*quirks*/DA_Q_4K }, { /* * XceedIOPS SATA SSDs * 4k optimised */ { T_DIRECT, SIP_MEDIA_FIXED, "ATA", "SG9XCS2D*", "*" }, /*quirks*/DA_Q_4K }, }; static disk_strategy_t dastrategy; static dumper_t dadump; static periph_init_t dainit; static void daasync(void *callback_arg, u_int32_t code, struct cam_path *path, void *arg); static void dasysctlinit(void *context, int pending); static int dacmdsizesysctl(SYSCTL_HANDLER_ARGS); static int dadeletemethodsysctl(SYSCTL_HANDLER_ARGS); static int dadeletemaxsysctl(SYSCTL_HANDLER_ARGS); static void dadeletemethodset(struct da_softc *softc, da_delete_methods delete_method); static off_t dadeletemaxsize(struct da_softc *softc, da_delete_methods delete_method); static void dadeletemethodchoose(struct da_softc *softc, da_delete_methods default_method); static void daprobedone(struct cam_periph *periph, union ccb *ccb); static periph_ctor_t daregister; static periph_dtor_t dacleanup; static periph_start_t dastart; static periph_oninv_t daoninvalidate; static void dadone(struct cam_periph *periph, union ccb *done_ccb); static int daerror(union ccb *ccb, u_int32_t cam_flags, u_int32_t sense_flags); static void daprevent(struct cam_periph *periph, int action); static void dareprobe(struct cam_periph *periph); static void dasetgeom(struct cam_periph *periph, uint32_t block_len, uint64_t maxsector, struct scsi_read_capacity_data_long *rcaplong, size_t rcap_size); static timeout_t dasendorderedtag; static void dashutdown(void *arg, int howto); static timeout_t damediapoll; #ifndef DA_DEFAULT_POLL_PERIOD #define DA_DEFAULT_POLL_PERIOD 3 #endif #ifndef DA_DEFAULT_TIMEOUT #define DA_DEFAULT_TIMEOUT 60 /* Timeout in seconds */ #endif #ifndef DA_DEFAULT_RETRY #define DA_DEFAULT_RETRY 4 #endif #ifndef DA_DEFAULT_SEND_ORDERED #define DA_DEFAULT_SEND_ORDERED 1 #endif #define DA_SIO (softc->sort_io_queue >= 0 ? \ softc->sort_io_queue : cam_sort_io_queues) static int da_poll_period = DA_DEFAULT_POLL_PERIOD; static int da_retry_count = DA_DEFAULT_RETRY; static int da_default_timeout = DA_DEFAULT_TIMEOUT; static int da_send_ordered = DA_DEFAULT_SEND_ORDERED; static SYSCTL_NODE(_kern_cam, OID_AUTO, da, CTLFLAG_RD, 0, "CAM Direct Access Disk driver"); SYSCTL_INT(_kern_cam_da, OID_AUTO, poll_period, CTLFLAG_RW, &da_poll_period, 0, "Media polling period in seconds"); TUNABLE_INT("kern.cam.da.poll_period", &da_poll_period); SYSCTL_INT(_kern_cam_da, OID_AUTO, retry_count, CTLFLAG_RW, &da_retry_count, 0, "Normal I/O retry count"); TUNABLE_INT("kern.cam.da.retry_count", &da_retry_count); SYSCTL_INT(_kern_cam_da, OID_AUTO, default_timeout, CTLFLAG_RW, &da_default_timeout, 0, "Normal I/O timeout (in seconds)"); TUNABLE_INT("kern.cam.da.default_timeout", &da_default_timeout); SYSCTL_INT(_kern_cam_da, OID_AUTO, send_ordered, CTLFLAG_RW, &da_send_ordered, 0, "Send Ordered Tags"); TUNABLE_INT("kern.cam.da.send_ordered", &da_send_ordered); /* * DA_ORDEREDTAG_INTERVAL determines how often, relative * to the default timeout, we check to see whether an ordered * tagged transaction is appropriate to prevent simple tag * starvation. Since we'd like to ensure that there is at least * 1/2 of the timeout length left for a starved transaction to * complete after we've sent an ordered tag, we must poll at least * four times in every timeout period. This takes care of the worst * case where a starved transaction starts during an interval that * meets the requirement "don't send an ordered tag" test so it takes * us two intervals to determine that a tag must be sent. */ #ifndef DA_ORDEREDTAG_INTERVAL #define DA_ORDEREDTAG_INTERVAL 4 #endif static struct periph_driver dadriver = { dainit, "da", TAILQ_HEAD_INITIALIZER(dadriver.units), /* generation */ 0 }; PERIPHDRIVER_DECLARE(da, dadriver); static MALLOC_DEFINE(M_SCSIDA, "scsi_da", "scsi_da buffers"); static int daopen(struct disk *dp) { struct cam_periph *periph; struct da_softc *softc; int error; periph = (struct cam_periph *)dp->d_drv1; if (cam_periph_acquire(periph) != CAM_REQ_CMP) { return (ENXIO); } cam_periph_lock(periph); if ((error = cam_periph_hold(periph, PRIBIO|PCATCH)) != 0) { cam_periph_unlock(periph); cam_periph_release(periph); return (error); } CAM_DEBUG(periph->path, CAM_DEBUG_TRACE | CAM_DEBUG_PERIPH, ("daopen\n")); softc = (struct da_softc *)periph->softc; dareprobe(periph); /* Wait for the disk size update. */ error = cam_periph_sleep(periph, &softc->disk->d_mediasize, PRIBIO, "dareprobe", 0); if (error != 0) xpt_print(periph->path, "unable to retrieve capacity data\n"); if (periph->flags & CAM_PERIPH_INVALID) error = ENXIO; if (error == 0 && (softc->flags & DA_FLAG_PACK_REMOVABLE) != 0 && (softc->quirks & DA_Q_NO_PREVENT) == 0) daprevent(periph, PR_PREVENT); if (error == 0) { softc->flags &= ~DA_FLAG_PACK_INVALID; softc->flags |= DA_FLAG_OPEN; } cam_periph_unhold(periph); cam_periph_unlock(periph); if (error != 0) cam_periph_release(periph); return (error); } static int daclose(struct disk *dp) { struct cam_periph *periph; struct da_softc *softc; union ccb *ccb; int error; periph = (struct cam_periph *)dp->d_drv1; softc = (struct da_softc *)periph->softc; cam_periph_lock(periph); CAM_DEBUG(periph->path, CAM_DEBUG_TRACE | CAM_DEBUG_PERIPH, ("daclose\n")); if (cam_periph_hold(periph, PRIBIO) == 0) { /* Flush disk cache. */ if ((softc->flags & DA_FLAG_DIRTY) != 0 && (softc->quirks & DA_Q_NO_SYNC_CACHE) == 0 && (softc->flags & DA_FLAG_PACK_INVALID) == 0) { ccb = cam_periph_getccb(periph, CAM_PRIORITY_NORMAL); scsi_synchronize_cache(&ccb->csio, /*retries*/1, /*cbfcnp*/dadone, MSG_SIMPLE_Q_TAG, /*begin_lba*/0, /*lb_count*/0, SSD_FULL_SIZE, 5 * 60 * 1000); error = cam_periph_runccb(ccb, daerror, /*cam_flags*/0, /*sense_flags*/SF_RETRY_UA | SF_QUIET_IR, softc->disk->d_devstat); if (error == 0) softc->flags &= ~DA_FLAG_DIRTY; xpt_release_ccb(ccb); } /* Allow medium removal. */ if ((softc->flags & DA_FLAG_PACK_REMOVABLE) != 0 && (softc->quirks & DA_Q_NO_PREVENT) == 0) daprevent(periph, PR_ALLOW); cam_periph_unhold(periph); } /* * If we've got removeable media, mark the blocksize as * unavailable, since it could change when new media is * inserted. */ if ((softc->flags & DA_FLAG_PACK_REMOVABLE) != 0) softc->disk->d_devstat->flags |= DEVSTAT_BS_UNAVAILABLE; softc->flags &= ~DA_FLAG_OPEN; while (softc->refcount != 0) cam_periph_sleep(periph, &softc->refcount, PRIBIO, "daclose", 1); cam_periph_unlock(periph); cam_periph_release(periph); return (0); } static void daschedule(struct cam_periph *periph) { struct da_softc *softc = (struct da_softc *)periph->softc; if (softc->state != DA_STATE_NORMAL) return; /* Check if we have more work to do. */ if (bioq_first(&softc->bio_queue) || (!softc->delete_running && bioq_first(&softc->delete_queue)) || softc->tur) { xpt_schedule(periph, CAM_PRIORITY_NORMAL); } } /* * Actually translate the requested transfer into one the physical driver * can understand. The transfer is described by a buf and will include * only one physical transfer. */ static void dastrategy(struct bio *bp) { struct cam_periph *periph; struct da_softc *softc; periph = (struct cam_periph *)bp->bio_disk->d_drv1; softc = (struct da_softc *)periph->softc; cam_periph_lock(periph); /* * If the device has been made invalid, error out */ if ((softc->flags & DA_FLAG_PACK_INVALID)) { cam_periph_unlock(periph); biofinish(bp, NULL, ENXIO); return; } CAM_DEBUG(periph->path, CAM_DEBUG_TRACE, ("dastrategy(%p)\n", bp)); /* * Place it in the queue of disk activities for this disk */ if (bp->bio_cmd == BIO_DELETE) { bioq_disksort(&softc->delete_queue, bp); } else if (DA_SIO) { bioq_disksort(&softc->bio_queue, bp); } else { bioq_insert_tail(&softc->bio_queue, bp); } /* * Schedule ourselves for performing the work. */ daschedule(periph); cam_periph_unlock(periph); return; } static int dadump(void *arg, void *virtual, vm_offset_t physical, off_t offset, size_t length) { struct cam_periph *periph; struct da_softc *softc; u_int secsize; struct ccb_scsiio csio; struct disk *dp; int error = 0; dp = arg; periph = dp->d_drv1; softc = (struct da_softc *)periph->softc; cam_periph_lock(periph); secsize = softc->params.secsize; if ((softc->flags & DA_FLAG_PACK_INVALID) != 0) { cam_periph_unlock(periph); return (ENXIO); } if (length > 0) { xpt_setup_ccb(&csio.ccb_h, periph->path, CAM_PRIORITY_NORMAL); csio.ccb_h.ccb_state = DA_CCB_DUMP; scsi_read_write(&csio, /*retries*/0, dadone, MSG_ORDERED_Q_TAG, /*read*/SCSI_RW_WRITE, /*byte2*/0, /*minimum_cmd_size*/ softc->minimum_cmd_size, offset / secsize, length / secsize, /*data_ptr*/(u_int8_t *) virtual, /*dxfer_len*/length, /*sense_len*/SSD_FULL_SIZE, da_default_timeout * 1000); xpt_polled_action((union ccb *)&csio); error = cam_periph_error((union ccb *)&csio, 0, SF_NO_RECOVERY | SF_NO_RETRY, NULL); if ((csio.ccb_h.status & CAM_DEV_QFRZN) != 0) cam_release_devq(csio.ccb_h.path, /*relsim_flags*/0, /*reduction*/0, /*timeout*/0, /*getcount_only*/0); if (error != 0) printf("Aborting dump due to I/O error.\n"); cam_periph_unlock(periph); return (error); } /* * Sync the disk cache contents to the physical media. */ if ((softc->quirks & DA_Q_NO_SYNC_CACHE) == 0) { xpt_setup_ccb(&csio.ccb_h, periph->path, CAM_PRIORITY_NORMAL); csio.ccb_h.ccb_state = DA_CCB_DUMP; scsi_synchronize_cache(&csio, /*retries*/0, /*cbfcnp*/dadone, MSG_SIMPLE_Q_TAG, /*begin_lba*/0,/* Cover the whole disk */ /*lb_count*/0, SSD_FULL_SIZE, 5 * 60 * 1000); xpt_polled_action((union ccb *)&csio); error = cam_periph_error((union ccb *)&csio, 0, SF_NO_RECOVERY | SF_NO_RETRY | SF_QUIET_IR, NULL); if ((csio.ccb_h.status & CAM_DEV_QFRZN) != 0) cam_release_devq(csio.ccb_h.path, /*relsim_flags*/0, /*reduction*/0, /*timeout*/0, /*getcount_only*/0); if (error != 0) xpt_print(periph->path, "Synchronize cache failed\n"); } cam_periph_unlock(periph); return (error); } static int dagetattr(struct bio *bp) { int ret; struct cam_periph *periph; periph = (struct cam_periph *)bp->bio_disk->d_drv1; cam_periph_lock(periph); ret = xpt_getattr(bp->bio_data, bp->bio_length, bp->bio_attribute, periph->path); cam_periph_unlock(periph); if (ret == 0) bp->bio_completed = bp->bio_length; return ret; } static void dainit(void) { cam_status status; /* * Install a global async callback. This callback will * receive async callbacks like "new device found". */ status = xpt_register_async(AC_FOUND_DEVICE, daasync, NULL, NULL); if (status != CAM_REQ_CMP) { printf("da: Failed to attach master async callback " "due to status 0x%x!\n", status); } else if (da_send_ordered) { /* Register our shutdown event handler */ if ((EVENTHANDLER_REGISTER(shutdown_post_sync, dashutdown, NULL, SHUTDOWN_PRI_DEFAULT)) == NULL) printf("dainit: shutdown event registration failed!\n"); } } /* * Callback from GEOM, called when it has finished cleaning up its * resources. */ static void dadiskgonecb(struct disk *dp) { struct cam_periph *periph; periph = (struct cam_periph *)dp->d_drv1; cam_periph_release(periph); } static void daoninvalidate(struct cam_periph *periph) { struct da_softc *softc; softc = (struct da_softc *)periph->softc; /* * De-register any async callbacks. */ xpt_register_async(0, daasync, periph, periph->path); softc->flags |= DA_FLAG_PACK_INVALID; /* * Return all queued I/O with ENXIO. * XXX Handle any transactions queued to the card * with XPT_ABORT_CCB. */ bioq_flush(&softc->bio_queue, NULL, ENXIO); bioq_flush(&softc->delete_queue, NULL, ENXIO); /* * Tell GEOM that we've gone away, we'll get a callback when it is * done cleaning up its resources. */ disk_gone(softc->disk); } static void dacleanup(struct cam_periph *periph) { struct da_softc *softc; softc = (struct da_softc *)periph->softc; cam_periph_unlock(periph); /* * If we can't free the sysctl tree, oh well... */ if ((softc->flags & DA_FLAG_SCTX_INIT) != 0 && sysctl_ctx_free(&softc->sysctl_ctx) != 0) { xpt_print(periph->path, "can't remove sysctl context\n"); } callout_drain(&softc->mediapoll_c); disk_destroy(softc->disk); callout_drain(&softc->sendordered_c); free(softc, M_DEVBUF); cam_periph_lock(periph); } static void daasync(void *callback_arg, u_int32_t code, struct cam_path *path, void *arg) { struct cam_periph *periph; struct da_softc *softc; periph = (struct cam_periph *)callback_arg; switch (code) { case AC_FOUND_DEVICE: { struct ccb_getdev *cgd; cam_status status; cgd = (struct ccb_getdev *)arg; if (cgd == NULL) break; if (cgd->protocol != PROTO_SCSI) break; if (SID_TYPE(&cgd->inq_data) != T_DIRECT && SID_TYPE(&cgd->inq_data) != T_RBC && SID_TYPE(&cgd->inq_data) != T_OPTICAL) break; /* * Allocate a peripheral instance for * this device and start the probe * process. */ status = cam_periph_alloc(daregister, daoninvalidate, dacleanup, dastart, "da", CAM_PERIPH_BIO, path, daasync, AC_FOUND_DEVICE, cgd); if (status != CAM_REQ_CMP && status != CAM_REQ_INPROG) printf("daasync: Unable to attach to new device " "due to status 0x%x\n", status); return; } case AC_ADVINFO_CHANGED: { uintptr_t buftype; buftype = (uintptr_t)arg; if (buftype == CDAI_TYPE_PHYS_PATH) { struct da_softc *softc; softc = periph->softc; disk_attr_changed(softc->disk, "GEOM::physpath", M_NOWAIT); } break; } case AC_UNIT_ATTENTION: { union ccb *ccb; int error_code, sense_key, asc, ascq; softc = (struct da_softc *)periph->softc; ccb = (union ccb *)arg; /* * Handle all UNIT ATTENTIONs except our own, * as they will be handled by daerror(). */ if (xpt_path_periph(ccb->ccb_h.path) != periph && scsi_extract_sense_ccb(ccb, &error_code, &sense_key, &asc, &ascq)) { if (asc == 0x2A && ascq == 0x09) { xpt_print(ccb->ccb_h.path, "Capacity data has changed\n"); softc->flags &= ~DA_FLAG_PROBED; dareprobe(periph); } else if (asc == 0x28 && ascq == 0x00) { softc->flags &= ~DA_FLAG_PROBED; disk_media_changed(softc->disk, M_NOWAIT); } else if (asc == 0x3F && ascq == 0x03) { xpt_print(ccb->ccb_h.path, "INQUIRY data has changed\n"); softc->flags &= ~DA_FLAG_PROBED; dareprobe(periph); } } cam_periph_async(periph, code, path, arg); break; } case AC_SCSI_AEN: softc = (struct da_softc *)periph->softc; if (!softc->tur) { if (cam_periph_acquire(periph) == CAM_REQ_CMP) { softc->tur = 1; daschedule(periph); } } /* FALLTHROUGH */ case AC_SENT_BDR: case AC_BUS_RESET: { struct ccb_hdr *ccbh; softc = (struct da_softc *)periph->softc; /* * Don't fail on the expected unit attention * that will occur. */ softc->flags |= DA_FLAG_RETRY_UA; LIST_FOREACH(ccbh, &softc->pending_ccbs, periph_links.le) ccbh->ccb_state |= DA_CCB_RETRY_UA; break; } default: break; } cam_periph_async(periph, code, path, arg); } static void dasysctlinit(void *context, int pending) { struct cam_periph *periph; struct da_softc *softc; char tmpstr[80], tmpstr2[80]; struct ccb_trans_settings cts; periph = (struct cam_periph *)context; /* * periph was held for us when this task was enqueued */ if (periph->flags & CAM_PERIPH_INVALID) { cam_periph_release(periph); return; } softc = (struct da_softc *)periph->softc; snprintf(tmpstr, sizeof(tmpstr), "CAM DA unit %d", periph->unit_number); snprintf(tmpstr2, sizeof(tmpstr2), "%d", periph->unit_number); sysctl_ctx_init(&softc->sysctl_ctx); softc->flags |= DA_FLAG_SCTX_INIT; softc->sysctl_tree = SYSCTL_ADD_NODE(&softc->sysctl_ctx, SYSCTL_STATIC_CHILDREN(_kern_cam_da), OID_AUTO, tmpstr2, CTLFLAG_RD, 0, tmpstr); if (softc->sysctl_tree == NULL) { printf("dasysctlinit: unable to allocate sysctl tree\n"); cam_periph_release(periph); return; } /* * Now register the sysctl handler, so the user can change the value on * the fly. */ SYSCTL_ADD_PROC(&softc->sysctl_ctx, SYSCTL_CHILDREN(softc->sysctl_tree), OID_AUTO, "delete_method", CTLTYPE_STRING | CTLFLAG_RW, softc, 0, dadeletemethodsysctl, "A", "BIO_DELETE execution method"); SYSCTL_ADD_PROC(&softc->sysctl_ctx, SYSCTL_CHILDREN(softc->sysctl_tree), OID_AUTO, "delete_max", CTLTYPE_U64 | CTLFLAG_RW, softc, 0, dadeletemaxsysctl, "Q", "Maximum BIO_DELETE size"); SYSCTL_ADD_PROC(&softc->sysctl_ctx, SYSCTL_CHILDREN(softc->sysctl_tree), OID_AUTO, "minimum_cmd_size", CTLTYPE_INT | CTLFLAG_RW, &softc->minimum_cmd_size, 0, dacmdsizesysctl, "I", "Minimum CDB size"); SYSCTL_ADD_INT(&softc->sysctl_ctx, SYSCTL_CHILDREN(softc->sysctl_tree), OID_AUTO, "sort_io_queue", CTLFLAG_RW, &softc->sort_io_queue, 0, "Sort IO queue to try and optimise disk access patterns"); SYSCTL_ADD_INT(&softc->sysctl_ctx, SYSCTL_CHILDREN(softc->sysctl_tree), OID_AUTO, "error_inject", CTLFLAG_RW, &softc->error_inject, 0, "error_inject leaf"); /* * Add some addressing info. */ memset(&cts, 0, sizeof (cts)); xpt_setup_ccb(&cts.ccb_h, periph->path, CAM_PRIORITY_NONE); cts.ccb_h.func_code = XPT_GET_TRAN_SETTINGS; cts.type = CTS_TYPE_CURRENT_SETTINGS; cam_periph_lock(periph); xpt_action((union ccb *)&cts); cam_periph_unlock(periph); if (cts.ccb_h.status != CAM_REQ_CMP) { cam_periph_release(periph); return; } if (cts.protocol == PROTO_SCSI && cts.transport == XPORT_FC) { struct ccb_trans_settings_fc *fc = &cts.xport_specific.fc; if (fc->valid & CTS_FC_VALID_WWPN) { softc->wwpn = fc->wwpn; SYSCTL_ADD_UQUAD(&softc->sysctl_ctx, SYSCTL_CHILDREN(softc->sysctl_tree), OID_AUTO, "wwpn", CTLFLAG_RD, &softc->wwpn, "World Wide Port Name"); } } cam_periph_release(periph); } static int dadeletemaxsysctl(SYSCTL_HANDLER_ARGS) { int error; uint64_t value; struct da_softc *softc; softc = (struct da_softc *)arg1; value = softc->disk->d_delmaxsize; error = sysctl_handle_64(oidp, &value, 0, req); if ((error != 0) || (req->newptr == NULL)) return (error); /* only accept values smaller than the calculated value */ if (value > dadeletemaxsize(softc, softc->delete_method)) { return (EINVAL); } softc->disk->d_delmaxsize = value; return (0); } static int dacmdsizesysctl(SYSCTL_HANDLER_ARGS) { int error, value; value = *(int *)arg1; error = sysctl_handle_int(oidp, &value, 0, req); if ((error != 0) || (req->newptr == NULL)) return (error); /* * Acceptable values here are 6, 10, 12 or 16. */ if (value < 6) value = 6; else if ((value > 6) && (value <= 10)) value = 10; else if ((value > 10) && (value <= 12)) value = 12; else if (value > 12) value = 16; *(int *)arg1 = value; return (0); } static void dadeletemethodset(struct da_softc *softc, da_delete_methods delete_method) { softc->delete_method = delete_method; softc->disk->d_delmaxsize = dadeletemaxsize(softc, delete_method); softc->delete_func = da_delete_functions[delete_method]; if (softc->delete_method > DA_DELETE_DISABLE) softc->disk->d_flags |= DISKFLAG_CANDELETE; else softc->disk->d_flags &= ~DISKFLAG_CANDELETE; } static off_t dadeletemaxsize(struct da_softc *softc, da_delete_methods delete_method) { off_t sectors; switch(delete_method) { case DA_DELETE_UNMAP: sectors = (off_t)softc->unmap_max_lba; break; case DA_DELETE_ATA_TRIM: sectors = (off_t)ATA_DSM_RANGE_MAX * softc->trim_max_ranges; break; case DA_DELETE_WS16: sectors = (off_t)min(softc->ws_max_blks, WS16_MAX_BLKS); break; case DA_DELETE_ZERO: case DA_DELETE_WS10: sectors = (off_t)min(softc->ws_max_blks, WS10_MAX_BLKS); break; default: return 0; } return (off_t)softc->params.secsize * min(sectors, (off_t)softc->params.sectors); } static void daprobedone(struct cam_periph *periph, union ccb *ccb) { struct da_softc *softc; softc = (struct da_softc *)periph->softc; dadeletemethodchoose(softc, DA_DELETE_NONE); if (bootverbose && (softc->flags & DA_FLAG_ANNOUNCED) == 0) { char buf[80]; int i, sep; snprintf(buf, sizeof(buf), "Delete methods: <"); sep = 0; for (i = DA_DELETE_MIN; i <= DA_DELETE_MAX; i++) { if (softc->delete_available & (1 << i)) { if (sep) { strlcat(buf, ",", sizeof(buf)); } else { sep = 1; } strlcat(buf, da_delete_method_names[i], sizeof(buf)); if (i == softc->delete_method) { strlcat(buf, "(*)", sizeof(buf)); } } } if (sep == 0) { if (softc->delete_method == DA_DELETE_NONE) strlcat(buf, "NONE(*)", sizeof(buf)); else strlcat(buf, "DISABLED(*)", sizeof(buf)); } strlcat(buf, ">", sizeof(buf)); printf("%s%d: %s\n", periph->periph_name, periph->unit_number, buf); } /* * Since our peripheral may be invalidated by an error * above or an external event, we must release our CCB * before releasing the probe lock on the peripheral. * The peripheral will only go away once the last lock * is removed, and we need it around for the CCB release * operation. */ xpt_release_ccb(ccb); softc->state = DA_STATE_NORMAL; softc->flags |= DA_FLAG_PROBED; daschedule(periph); wakeup(&softc->disk->d_mediasize); if ((softc->flags & DA_FLAG_ANNOUNCED) == 0) { softc->flags |= DA_FLAG_ANNOUNCED; cam_periph_unhold(periph); } else cam_periph_release_locked(periph); } static void dadeletemethodchoose(struct da_softc *softc, da_delete_methods default_method) { int i, delete_method; delete_method = default_method; /* * Use the pre-defined order to choose the best * performing delete. */ for (i = DA_DELETE_MIN; i <= DA_DELETE_MAX; i++) { if (softc->delete_available & (1 << i)) { dadeletemethodset(softc, i); return; } } dadeletemethodset(softc, delete_method); } static int dadeletemethodsysctl(SYSCTL_HANDLER_ARGS) { char buf[16]; const char *p; struct da_softc *softc; int i, error, methods, value; softc = (struct da_softc *)arg1; value = softc->delete_method; if (value < 0 || value > DA_DELETE_MAX) p = "UNKNOWN"; else p = da_delete_method_names[value]; strncpy(buf, p, sizeof(buf)); error = sysctl_handle_string(oidp, buf, sizeof(buf), req); if (error != 0 || req->newptr == NULL) return (error); methods = softc->delete_available | (1 << DA_DELETE_DISABLE); for (i = 0; i <= DA_DELETE_MAX; i++) { if (!(methods & (1 << i)) || strcmp(buf, da_delete_method_names[i]) != 0) continue; dadeletemethodset(softc, i); return (0); } return (EINVAL); } static cam_status daregister(struct cam_periph *periph, void *arg) { struct da_softc *softc; struct ccb_pathinq cpi; struct ccb_getdev *cgd; char tmpstr[80]; caddr_t match; cgd = (struct ccb_getdev *)arg; if (cgd == NULL) { printf("daregister: no getdev CCB, can't register device\n"); return(CAM_REQ_CMP_ERR); } softc = (struct da_softc *)malloc(sizeof(*softc), M_DEVBUF, M_NOWAIT|M_ZERO); if (softc == NULL) { printf("daregister: Unable to probe new device. " "Unable to allocate softc\n"); return(CAM_REQ_CMP_ERR); } LIST_INIT(&softc->pending_ccbs); softc->state = DA_STATE_PROBE_RC; bioq_init(&softc->bio_queue); bioq_init(&softc->delete_queue); bioq_init(&softc->delete_run_queue); if (SID_IS_REMOVABLE(&cgd->inq_data)) softc->flags |= DA_FLAG_PACK_REMOVABLE; softc->unmap_max_ranges = UNMAP_MAX_RANGES; softc->unmap_max_lba = UNMAP_RANGE_MAX; softc->ws_max_blks = WS16_MAX_BLKS; softc->trim_max_ranges = ATA_TRIM_MAX_RANGES; softc->sort_io_queue = -1; periph->softc = softc; /* * See if this device has any quirks. */ match = cam_quirkmatch((caddr_t)&cgd->inq_data, (caddr_t)da_quirk_table, sizeof(da_quirk_table)/sizeof(*da_quirk_table), sizeof(*da_quirk_table), scsi_inquiry_match); if (match != NULL) softc->quirks = ((struct da_quirk_entry *)match)->quirks; else softc->quirks = DA_Q_NONE; /* Check if the SIM does not want 6 byte commands */ bzero(&cpi, sizeof(cpi)); xpt_setup_ccb(&cpi.ccb_h, periph->path, CAM_PRIORITY_NORMAL); cpi.ccb_h.func_code = XPT_PATH_INQ; xpt_action((union ccb *)&cpi); if (cpi.ccb_h.status == CAM_REQ_CMP && (cpi.hba_misc & PIM_NO_6_BYTE)) softc->quirks |= DA_Q_NO_6_BYTE; TASK_INIT(&softc->sysctl_task, 0, dasysctlinit, periph); /* * Take an exclusive refcount on the periph while dastart is called * to finish the probe. The reference will be dropped in dadone at * the end of probe. */ (void)cam_periph_hold(periph, PRIBIO); /* * Schedule a periodic event to occasionally send an * ordered tag to a device. */ callout_init_mtx(&softc->sendordered_c, cam_periph_mtx(periph), 0); callout_reset(&softc->sendordered_c, (da_default_timeout * hz) / DA_ORDEREDTAG_INTERVAL, dasendorderedtag, softc); cam_periph_unlock(periph); /* * RBC devices don't have to support READ(6), only READ(10). */ if (softc->quirks & DA_Q_NO_6_BYTE || SID_TYPE(&cgd->inq_data) == T_RBC) softc->minimum_cmd_size = 10; else softc->minimum_cmd_size = 6; /* * Load the user's default, if any. */ snprintf(tmpstr, sizeof(tmpstr), "kern.cam.da.%d.minimum_cmd_size", periph->unit_number); TUNABLE_INT_FETCH(tmpstr, &softc->minimum_cmd_size); /* * 6, 10, 12 and 16 are the currently permissible values. */ if (softc->minimum_cmd_size < 6) softc->minimum_cmd_size = 6; else if ((softc->minimum_cmd_size > 6) && (softc->minimum_cmd_size <= 10)) softc->minimum_cmd_size = 10; else if ((softc->minimum_cmd_size > 10) && (softc->minimum_cmd_size <= 12)) softc->minimum_cmd_size = 12; else if (softc->minimum_cmd_size > 12) softc->minimum_cmd_size = 16; /* Predict whether device may support READ CAPACITY(16). */ if (SID_ANSI_REV(&cgd->inq_data) >= SCSI_REV_SPC3 && (softc->quirks & DA_Q_NO_RC16) == 0) { softc->flags |= DA_FLAG_CAN_RC16; softc->state = DA_STATE_PROBE_RC16; } /* * Register this media as a disk. */ softc->disk = disk_alloc(); softc->disk->d_devstat = devstat_new_entry(periph->periph_name, periph->unit_number, 0, DEVSTAT_BS_UNAVAILABLE, SID_TYPE(&cgd->inq_data) | XPORT_DEVSTAT_TYPE(cpi.transport), DEVSTAT_PRIORITY_DISK); softc->disk->d_open = daopen; softc->disk->d_close = daclose; softc->disk->d_strategy = dastrategy; softc->disk->d_dump = dadump; softc->disk->d_getattr = dagetattr; softc->disk->d_gone = dadiskgonecb; softc->disk->d_name = "da"; softc->disk->d_drv1 = periph; if (cpi.maxio == 0) softc->maxio = DFLTPHYS; /* traditional default */ else if (cpi.maxio > MAXPHYS) softc->maxio = MAXPHYS; /* for safety */ else softc->maxio = cpi.maxio; softc->disk->d_maxsize = softc->maxio; softc->disk->d_unit = periph->unit_number; softc->disk->d_flags = DISKFLAG_DIRECT_COMPLETION; if ((softc->quirks & DA_Q_NO_SYNC_CACHE) == 0) softc->disk->d_flags |= DISKFLAG_CANFLUSHCACHE; if ((cpi.hba_misc & PIM_UNMAPPED) != 0) softc->disk->d_flags |= DISKFLAG_UNMAPPED_BIO; cam_strvis(softc->disk->d_descr, cgd->inq_data.vendor, sizeof(cgd->inq_data.vendor), sizeof(softc->disk->d_descr)); strlcat(softc->disk->d_descr, " ", sizeof(softc->disk->d_descr)); cam_strvis(&softc->disk->d_descr[strlen(softc->disk->d_descr)], cgd->inq_data.product, sizeof(cgd->inq_data.product), sizeof(softc->disk->d_descr) - strlen(softc->disk->d_descr)); softc->disk->d_hba_vendor = cpi.hba_vendor; softc->disk->d_hba_device = cpi.hba_device; softc->disk->d_hba_subvendor = cpi.hba_subvendor; softc->disk->d_hba_subdevice = cpi.hba_subdevice; /* * Acquire a reference to the periph before we register with GEOM. * We'll release this reference once GEOM calls us back (via * dadiskgonecb()) telling us that our provider has been freed. */ if (cam_periph_acquire(periph) != CAM_REQ_CMP) { xpt_print(periph->path, "%s: lost periph during " "registration!\n", __func__); cam_periph_lock(periph); return (CAM_REQ_CMP_ERR); } disk_create(softc->disk, DISK_VERSION); cam_periph_lock(periph); /* * Add async callbacks for events of interest. * I don't bother checking if this fails as, * in most cases, the system will function just * fine without them and the only alternative * would be to not attach the device on failure. */ xpt_register_async(AC_SENT_BDR | AC_BUS_RESET | AC_LOST_DEVICE | AC_ADVINFO_CHANGED | AC_SCSI_AEN | AC_UNIT_ATTENTION, daasync, periph, periph->path); /* * Emit an attribute changed notification just in case * physical path information arrived before our async * event handler was registered, but after anyone attaching * to our disk device polled it. */ disk_attr_changed(softc->disk, "GEOM::physpath", M_NOWAIT); /* * Schedule a periodic media polling events. */ callout_init_mtx(&softc->mediapoll_c, cam_periph_mtx(periph), 0); if ((softc->flags & DA_FLAG_PACK_REMOVABLE) && (cgd->inq_flags & SID_AEN) == 0 && da_poll_period != 0) callout_reset(&softc->mediapoll_c, da_poll_period * hz, damediapoll, periph); xpt_schedule(periph, CAM_PRIORITY_DEV); return(CAM_REQ_CMP); } static void dastart(struct cam_periph *periph, union ccb *start_ccb) { struct da_softc *softc; softc = (struct da_softc *)periph->softc; CAM_DEBUG(periph->path, CAM_DEBUG_TRACE, ("dastart\n")); skipstate: switch (softc->state) { case DA_STATE_NORMAL: { struct bio *bp; uint8_t tag_code; /* Run BIO_DELETE if not running yet. */ if (!softc->delete_running && (bp = bioq_first(&softc->delete_queue)) != NULL) { if (softc->delete_func != NULL) { softc->delete_func(periph, start_ccb, bp); goto out; } else { bioq_flush(&softc->delete_queue, NULL, 0); /* FALLTHROUGH */ } } /* Run regular command. */ bp = bioq_takefirst(&softc->bio_queue); if (bp == NULL) { if (softc->tur) { softc->tur = 0; scsi_test_unit_ready(&start_ccb->csio, /*retries*/ da_retry_count, dadone, MSG_SIMPLE_Q_TAG, SSD_FULL_SIZE, da_default_timeout * 1000); start_ccb->ccb_h.ccb_bp = NULL; start_ccb->ccb_h.ccb_state = DA_CCB_TUR; xpt_action(start_ccb); } else xpt_release_ccb(start_ccb); break; } if (softc->tur) { softc->tur = 0; cam_periph_release_locked(periph); } if ((bp->bio_flags & BIO_ORDERED) != 0 || (softc->flags & DA_FLAG_NEED_OTAG) != 0) { softc->flags &= ~DA_FLAG_NEED_OTAG; softc->flags |= DA_FLAG_WAS_OTAG; tag_code = MSG_ORDERED_Q_TAG; } else { tag_code = MSG_SIMPLE_Q_TAG; } switch (bp->bio_cmd) { case BIO_WRITE: softc->flags |= DA_FLAG_DIRTY; /* FALLTHROUGH */ case BIO_READ: scsi_read_write(&start_ccb->csio, /*retries*/da_retry_count, /*cbfcnp*/dadone, /*tag_action*/tag_code, /*read_op*/(bp->bio_cmd == BIO_READ ? SCSI_RW_READ : SCSI_RW_WRITE) | ((bp->bio_flags & BIO_UNMAPPED) != 0 ? SCSI_RW_BIO : 0), /*byte2*/0, softc->minimum_cmd_size, /*lba*/bp->bio_pblkno, /*block_count*/bp->bio_bcount / softc->params.secsize, /*data_ptr*/ (bp->bio_flags & BIO_UNMAPPED) != 0 ? (void *)bp : bp->bio_data, /*dxfer_len*/ bp->bio_bcount, /*sense_len*/SSD_FULL_SIZE, da_default_timeout * 1000); break; case BIO_FLUSH: /* * BIO_FLUSH doesn't currently communicate * range data, so we synchronize the cache * over the whole disk. We also force * ordered tag semantics the flush applies * to all previously queued I/O. */ scsi_synchronize_cache(&start_ccb->csio, /*retries*/1, /*cbfcnp*/dadone, MSG_ORDERED_Q_TAG, /*begin_lba*/0, /*lb_count*/0, SSD_FULL_SIZE, da_default_timeout*1000); break; } start_ccb->ccb_h.ccb_state = DA_CCB_BUFFER_IO; start_ccb->ccb_h.flags |= CAM_UNLOCKED; out: LIST_INSERT_HEAD(&softc->pending_ccbs, &start_ccb->ccb_h, periph_links.le); /* We expect a unit attention from this device */ if ((softc->flags & DA_FLAG_RETRY_UA) != 0) { start_ccb->ccb_h.ccb_state |= DA_CCB_RETRY_UA; softc->flags &= ~DA_FLAG_RETRY_UA; } start_ccb->ccb_h.ccb_bp = bp; softc->refcount++; cam_periph_unlock(periph); xpt_action(start_ccb); cam_periph_lock(periph); softc->refcount--; /* May have more work to do, so ensure we stay scheduled */ daschedule(periph); break; } case DA_STATE_PROBE_RC: { struct scsi_read_capacity_data *rcap; rcap = (struct scsi_read_capacity_data *) malloc(sizeof(*rcap), M_SCSIDA, M_NOWAIT|M_ZERO); if (rcap == NULL) { printf("dastart: Couldn't malloc read_capacity data\n"); /* da_free_periph??? */ break; } scsi_read_capacity(&start_ccb->csio, /*retries*/da_retry_count, dadone, MSG_SIMPLE_Q_TAG, rcap, SSD_FULL_SIZE, /*timeout*/5000); start_ccb->ccb_h.ccb_bp = NULL; start_ccb->ccb_h.ccb_state = DA_CCB_PROBE_RC; xpt_action(start_ccb); break; } case DA_STATE_PROBE_RC16: { struct scsi_read_capacity_data_long *rcaplong; rcaplong = (struct scsi_read_capacity_data_long *) malloc(sizeof(*rcaplong), M_SCSIDA, M_NOWAIT|M_ZERO); if (rcaplong == NULL) { printf("dastart: Couldn't malloc read_capacity data\n"); /* da_free_periph??? */ break; } scsi_read_capacity_16(&start_ccb->csio, /*retries*/ da_retry_count, /*cbfcnp*/ dadone, /*tag_action*/ MSG_SIMPLE_Q_TAG, /*lba*/ 0, /*reladr*/ 0, /*pmi*/ 0, /*rcap_buf*/ (uint8_t *)rcaplong, /*rcap_buf_len*/ sizeof(*rcaplong), /*sense_len*/ SSD_FULL_SIZE, /*timeout*/ da_default_timeout * 1000); start_ccb->ccb_h.ccb_bp = NULL; start_ccb->ccb_h.ccb_state = DA_CCB_PROBE_RC16; xpt_action(start_ccb); break; } case DA_STATE_PROBE_LBP: { struct scsi_vpd_logical_block_prov *lbp; if (!scsi_vpd_supported_page(periph, SVPD_LBP)) { /* * If we get here we don't support any SBC-3 delete * methods with UNMAP as the Logical Block Provisioning * VPD page support is required for devices which * support it according to T10/1799-D Revision 31 * however older revisions of the spec don't mandate * this so we currently don't remove these methods * from the available set. */ softc->state = DA_STATE_PROBE_BLK_LIMITS; goto skipstate; } lbp = (struct scsi_vpd_logical_block_prov *) malloc(sizeof(*lbp), M_SCSIDA, M_NOWAIT|M_ZERO); if (lbp == NULL) { printf("dastart: Couldn't malloc lbp data\n"); /* da_free_periph??? */ break; } scsi_inquiry(&start_ccb->csio, /*retries*/da_retry_count, /*cbfcnp*/dadone, /*tag_action*/MSG_SIMPLE_Q_TAG, /*inq_buf*/(u_int8_t *)lbp, /*inq_len*/sizeof(*lbp), /*evpd*/TRUE, /*page_code*/SVPD_LBP, /*sense_len*/SSD_MIN_SIZE, /*timeout*/da_default_timeout * 1000); start_ccb->ccb_h.ccb_bp = NULL; start_ccb->ccb_h.ccb_state = DA_CCB_PROBE_LBP; xpt_action(start_ccb); break; } case DA_STATE_PROBE_BLK_LIMITS: { struct scsi_vpd_block_limits *block_limits; if (!scsi_vpd_supported_page(periph, SVPD_BLOCK_LIMITS)) { /* Not supported skip to next probe */ softc->state = DA_STATE_PROBE_BDC; goto skipstate; } block_limits = (struct scsi_vpd_block_limits *) malloc(sizeof(*block_limits), M_SCSIDA, M_NOWAIT|M_ZERO); if (block_limits == NULL) { printf("dastart: Couldn't malloc block_limits data\n"); /* da_free_periph??? */ break; } scsi_inquiry(&start_ccb->csio, /*retries*/da_retry_count, /*cbfcnp*/dadone, /*tag_action*/MSG_SIMPLE_Q_TAG, /*inq_buf*/(u_int8_t *)block_limits, /*inq_len*/sizeof(*block_limits), /*evpd*/TRUE, /*page_code*/SVPD_BLOCK_LIMITS, /*sense_len*/SSD_MIN_SIZE, /*timeout*/da_default_timeout * 1000); start_ccb->ccb_h.ccb_bp = NULL; start_ccb->ccb_h.ccb_state = DA_CCB_PROBE_BLK_LIMITS; xpt_action(start_ccb); break; } case DA_STATE_PROBE_BDC: { struct scsi_vpd_block_characteristics *bdc; if (!scsi_vpd_supported_page(periph, SVPD_BDC)) { softc->state = DA_STATE_PROBE_ATA; goto skipstate; } bdc = (struct scsi_vpd_block_characteristics *) malloc(sizeof(*bdc), M_SCSIDA, M_NOWAIT|M_ZERO); if (bdc == NULL) { printf("dastart: Couldn't malloc bdc data\n"); /* da_free_periph??? */ break; } scsi_inquiry(&start_ccb->csio, /*retries*/da_retry_count, /*cbfcnp*/dadone, /*tag_action*/MSG_SIMPLE_Q_TAG, /*inq_buf*/(u_int8_t *)bdc, /*inq_len*/sizeof(*bdc), /*evpd*/TRUE, /*page_code*/SVPD_BDC, /*sense_len*/SSD_MIN_SIZE, /*timeout*/da_default_timeout * 1000); start_ccb->ccb_h.ccb_bp = NULL; start_ccb->ccb_h.ccb_state = DA_CCB_PROBE_BDC; xpt_action(start_ccb); break; } case DA_STATE_PROBE_ATA: { struct ata_params *ata_params; if (!scsi_vpd_supported_page(periph, SVPD_ATA_INFORMATION)) { daprobedone(periph, start_ccb); break; } ata_params = (struct ata_params*) malloc(sizeof(*ata_params), M_SCSIDA, M_NOWAIT|M_ZERO); if (ata_params == NULL) { printf("dastart: Couldn't malloc ata_params data\n"); /* da_free_periph??? */ break; } scsi_ata_identify(&start_ccb->csio, /*retries*/da_retry_count, /*cbfcnp*/dadone, /*tag_action*/MSG_SIMPLE_Q_TAG, /*data_ptr*/(u_int8_t *)ata_params, /*dxfer_len*/sizeof(*ata_params), /*sense_len*/SSD_FULL_SIZE, /*timeout*/da_default_timeout * 1000); start_ccb->ccb_h.ccb_bp = NULL; start_ccb->ccb_h.ccb_state = DA_CCB_PROBE_ATA; xpt_action(start_ccb); break; } } } /* * In each of the methods below, while its the caller's * responsibility to ensure the request will fit into a * single device request, we might have changed the delete * method due to the device incorrectly advertising either * its supported methods or limits. * * To prevent this causing further issues we validate the * against the methods limits, and warn which would * otherwise be unnecessary. */ static void da_delete_unmap(struct cam_periph *periph, union ccb *ccb, struct bio *bp) { struct da_softc *softc = (struct da_softc *)periph->softc;; struct bio *bp1; uint8_t *buf = softc->unmap_buf; uint64_t lba, lastlba = (uint64_t)-1; uint64_t totalcount = 0; uint64_t count; uint32_t lastcount = 0, c; uint32_t off, ranges = 0; /* * Currently this doesn't take the UNMAP * Granularity and Granularity Alignment * fields into account. * * This could result in both unoptimal unmap * requests as as well as UNMAP calls unmapping * fewer LBA's than requested. */ softc->delete_running = 1; bzero(softc->unmap_buf, sizeof(softc->unmap_buf)); bp1 = bp; do { bioq_remove(&softc->delete_queue, bp1); if (bp1 != bp) bioq_insert_tail(&softc->delete_run_queue, bp1); lba = bp1->bio_pblkno; count = bp1->bio_bcount / softc->params.secsize; /* Try to extend the previous range. */ if (lba == lastlba) { c = omin(count, UNMAP_RANGE_MAX - lastcount); lastcount += c; off = ((ranges - 1) * UNMAP_RANGE_SIZE) + UNMAP_HEAD_SIZE; scsi_ulto4b(lastcount, &buf[off + 8]); count -= c; lba +=c; totalcount += c; } while (count > 0) { c = omin(count, UNMAP_RANGE_MAX); if (totalcount + c > softc->unmap_max_lba || ranges >= softc->unmap_max_ranges) { xpt_print(periph->path, "%s issuing short delete %ld > %ld" "|| %d >= %d", da_delete_method_desc[softc->delete_method], totalcount + c, softc->unmap_max_lba, ranges, softc->unmap_max_ranges); break; } off = (ranges * UNMAP_RANGE_SIZE) + UNMAP_HEAD_SIZE; scsi_u64to8b(lba, &buf[off + 0]); scsi_ulto4b(c, &buf[off + 8]); lba += c; totalcount += c; ranges++; count -= c; lastcount = c; } lastlba = lba; bp1 = bioq_first(&softc->delete_queue); if (bp1 == NULL || ranges >= softc->unmap_max_ranges || totalcount + bp1->bio_bcount / softc->params.secsize > softc->unmap_max_lba) break; } while (1); scsi_ulto2b(ranges * 16 + 6, &buf[0]); scsi_ulto2b(ranges * 16, &buf[2]); scsi_unmap(&ccb->csio, /*retries*/da_retry_count, /*cbfcnp*/dadone, /*tag_action*/MSG_SIMPLE_Q_TAG, /*byte2*/0, /*data_ptr*/ buf, /*dxfer_len*/ ranges * 16 + 8, /*sense_len*/SSD_FULL_SIZE, da_default_timeout * 1000); ccb->ccb_h.ccb_state = DA_CCB_DELETE; ccb->ccb_h.flags |= CAM_UNLOCKED; } static void da_delete_trim(struct cam_periph *periph, union ccb *ccb, struct bio *bp) { struct da_softc *softc = (struct da_softc *)periph->softc; struct bio *bp1; uint8_t *buf = softc->unmap_buf; uint64_t lastlba = (uint64_t)-1; uint64_t count; uint64_t lba; uint32_t lastcount = 0, c, requestcount; int ranges = 0, off, block_count; softc->delete_running = 1; bzero(softc->unmap_buf, sizeof(softc->unmap_buf)); bp1 = bp; do { bioq_remove(&softc->delete_queue, bp1); if (bp1 != bp) bioq_insert_tail(&softc->delete_run_queue, bp1); lba = bp1->bio_pblkno; count = bp1->bio_bcount / softc->params.secsize; requestcount = count; /* Try to extend the previous range. */ if (lba == lastlba) { c = min(count, ATA_DSM_RANGE_MAX - lastcount); lastcount += c; off = (ranges - 1) * 8; buf[off + 6] = lastcount & 0xff; buf[off + 7] = (lastcount >> 8) & 0xff; count -= c; lba += c; } while (count > 0) { c = min(count, ATA_DSM_RANGE_MAX); off = ranges * 8; buf[off + 0] = lba & 0xff; buf[off + 1] = (lba >> 8) & 0xff; buf[off + 2] = (lba >> 16) & 0xff; buf[off + 3] = (lba >> 24) & 0xff; buf[off + 4] = (lba >> 32) & 0xff; buf[off + 5] = (lba >> 40) & 0xff; buf[off + 6] = c & 0xff; buf[off + 7] = (c >> 8) & 0xff; lba += c; ranges++; count -= c; lastcount = c; if (count != 0 && ranges == softc->trim_max_ranges) { xpt_print(periph->path, "%s issuing short delete %ld > %ld\n", da_delete_method_desc[softc->delete_method], requestcount, (softc->trim_max_ranges - ranges) * ATA_DSM_RANGE_MAX); break; } } lastlba = lba; bp1 = bioq_first(&softc->delete_queue); if (bp1 == NULL || bp1->bio_bcount / softc->params.secsize > (softc->trim_max_ranges - ranges) * ATA_DSM_RANGE_MAX) break; } while (1); block_count = (ranges + ATA_DSM_BLK_RANGES - 1) / ATA_DSM_BLK_RANGES; scsi_ata_trim(&ccb->csio, /*retries*/da_retry_count, /*cbfcnp*/dadone, /*tag_action*/MSG_SIMPLE_Q_TAG, block_count, /*data_ptr*/buf, /*dxfer_len*/block_count * ATA_DSM_BLK_SIZE, /*sense_len*/SSD_FULL_SIZE, da_default_timeout * 1000); ccb->ccb_h.ccb_state = DA_CCB_DELETE; ccb->ccb_h.flags |= CAM_UNLOCKED; } /* * We calculate ws_max_blks here based off d_delmaxsize instead * of using softc->ws_max_blks as it is absolute max for the * device not the protocol max which may well be lower. */ static void da_delete_ws(struct cam_periph *periph, union ccb *ccb, struct bio *bp) { struct da_softc *softc; struct bio *bp1; uint64_t ws_max_blks; uint64_t lba; uint64_t count; /* forward compat with WS32 */ softc = (struct da_softc *)periph->softc; ws_max_blks = softc->disk->d_delmaxsize / softc->params.secsize; softc->delete_running = 1; lba = bp->bio_pblkno; count = 0; bp1 = bp; do { bioq_remove(&softc->delete_queue, bp1); if (bp1 != bp) bioq_insert_tail(&softc->delete_run_queue, bp1); count += bp1->bio_bcount / softc->params.secsize; if (count > ws_max_blks) { xpt_print(periph->path, "%s issuing short delete %ld > %ld\n", da_delete_method_desc[softc->delete_method], count, ws_max_blks); count = min(count, ws_max_blks); break; } bp1 = bioq_first(&softc->delete_queue); if (bp1 == NULL || lba + count != bp1->bio_pblkno || count + bp1->bio_bcount / softc->params.secsize > ws_max_blks) break; } while (1); scsi_write_same(&ccb->csio, /*retries*/da_retry_count, /*cbfcnp*/dadone, /*tag_action*/MSG_SIMPLE_Q_TAG, /*byte2*/softc->delete_method == DA_DELETE_ZERO ? 0 : SWS_UNMAP, softc->delete_method == DA_DELETE_WS16 ? 16 : 10, /*lba*/lba, /*block_count*/count, /*data_ptr*/ __DECONST(void *, zero_region), /*dxfer_len*/ softc->params.secsize, /*sense_len*/SSD_FULL_SIZE, da_default_timeout * 1000); ccb->ccb_h.ccb_state = DA_CCB_DELETE; ccb->ccb_h.flags |= CAM_UNLOCKED; } static int cmd6workaround(union ccb *ccb) { struct scsi_rw_6 cmd6; struct scsi_rw_10 *cmd10; struct da_softc *softc; u_int8_t *cdb; struct bio *bp; int frozen; cdb = ccb->csio.cdb_io.cdb_bytes; softc = (struct da_softc *)xpt_path_periph(ccb->ccb_h.path)->softc; if (ccb->ccb_h.ccb_state == DA_CCB_DELETE) { da_delete_methods old_method = softc->delete_method; /* * Typically there are two reasons for failure here * 1. Delete method was detected as supported but isn't * 2. Delete failed due to invalid params e.g. too big * * While we will attempt to choose an alternative delete method * this may result in short deletes if the existing delete * requests from geom are big for the new method choosen. * * This method assumes that the error which triggered this * will not retry the io otherwise a panic will occur */ dadeleteflag(softc, old_method, 0); dadeletemethodchoose(softc, DA_DELETE_DISABLE); if (softc->delete_method == DA_DELETE_DISABLE) xpt_print(ccb->ccb_h.path, "%s failed, disabling BIO_DELETE\n", da_delete_method_desc[old_method]); else xpt_print(ccb->ccb_h.path, "%s failed, switching to %s BIO_DELETE\n", da_delete_method_desc[old_method], da_delete_method_desc[softc->delete_method]); while ((bp = bioq_takefirst(&softc->delete_run_queue)) != NULL) bioq_disksort(&softc->delete_queue, bp); bioq_disksort(&softc->delete_queue, (struct bio *)ccb->ccb_h.ccb_bp); ccb->ccb_h.ccb_bp = NULL; return (0); } /* Detect unsupported PREVENT ALLOW MEDIUM REMOVAL. */ if ((ccb->ccb_h.flags & CAM_CDB_POINTER) == 0 && (*cdb == PREVENT_ALLOW) && (softc->quirks & DA_Q_NO_PREVENT) == 0) { if (bootverbose) xpt_print(ccb->ccb_h.path, "PREVENT ALLOW MEDIUM REMOVAL not supported.\n"); softc->quirks |= DA_Q_NO_PREVENT; return (0); } /* Detect unsupported SYNCHRONIZE CACHE(10). */ if ((ccb->ccb_h.flags & CAM_CDB_POINTER) == 0 && (*cdb == SYNCHRONIZE_CACHE) && (softc->quirks & DA_Q_NO_SYNC_CACHE) == 0) { if (bootverbose) xpt_print(ccb->ccb_h.path, "SYNCHRONIZE CACHE(10) not supported.\n"); softc->quirks |= DA_Q_NO_SYNC_CACHE; softc->disk->d_flags &= ~DISKFLAG_CANFLUSHCACHE; return (0); } /* Translation only possible if CDB is an array and cmd is R/W6 */ if ((ccb->ccb_h.flags & CAM_CDB_POINTER) != 0 || (*cdb != READ_6 && *cdb != WRITE_6)) return 0; xpt_print(ccb->ccb_h.path, "READ(6)/WRITE(6) not supported, " "increasing minimum_cmd_size to 10.\n"); softc->minimum_cmd_size = 10; bcopy(cdb, &cmd6, sizeof(struct scsi_rw_6)); cmd10 = (struct scsi_rw_10 *)cdb; cmd10->opcode = (cmd6.opcode == READ_6) ? READ_10 : WRITE_10; cmd10->byte2 = 0; scsi_ulto4b(scsi_3btoul(cmd6.addr), cmd10->addr); cmd10->reserved = 0; scsi_ulto2b(cmd6.length, cmd10->length); cmd10->control = cmd6.control; ccb->csio.cdb_len = sizeof(*cmd10); /* Requeue request, unfreezing queue if necessary */ frozen = (ccb->ccb_h.status & CAM_DEV_QFRZN) != 0; ccb->ccb_h.status = CAM_REQUEUE_REQ; xpt_action(ccb); if (frozen) { cam_release_devq(ccb->ccb_h.path, /*relsim_flags*/0, /*reduction*/0, /*timeout*/0, /*getcount_only*/0); } return (ERESTART); } static void dadone(struct cam_periph *periph, union ccb *done_ccb) { struct da_softc *softc; struct ccb_scsiio *csio; u_int32_t priority; da_ccb_state state; softc = (struct da_softc *)periph->softc; priority = done_ccb->ccb_h.pinfo.priority; CAM_DEBUG(periph->path, CAM_DEBUG_TRACE, ("dadone\n")); csio = &done_ccb->csio; state = csio->ccb_h.ccb_state & DA_CCB_TYPE_MASK; switch (state) { case DA_CCB_BUFFER_IO: case DA_CCB_DELETE: { struct bio *bp, *bp1; cam_periph_lock(periph); bp = (struct bio *)done_ccb->ccb_h.ccb_bp; if ((done_ccb->ccb_h.status & CAM_STATUS_MASK) != CAM_REQ_CMP) { int error; int sf; if ((csio->ccb_h.ccb_state & DA_CCB_RETRY_UA) != 0) sf = SF_RETRY_UA; else sf = 0; error = daerror(done_ccb, CAM_RETRY_SELTO, sf); if (error == ERESTART) { /* * A retry was scheduled, so * just return. */ cam_periph_unlock(periph); return; } bp = (struct bio *)done_ccb->ccb_h.ccb_bp; if (error != 0) { int queued_error; /* * return all queued I/O with EIO, so that * the client can retry these I/Os in the * proper order should it attempt to recover. */ queued_error = EIO; if (error == ENXIO && (softc->flags & DA_FLAG_PACK_INVALID)== 0) { /* * Catastrophic error. Mark our pack as * invalid. */ /* * XXX See if this is really a media * XXX change first? */ xpt_print(periph->path, "Invalidating pack\n"); softc->flags |= DA_FLAG_PACK_INVALID; queued_error = ENXIO; } bioq_flush(&softc->bio_queue, NULL, queued_error); if (bp != NULL) { bp->bio_error = error; bp->bio_resid = bp->bio_bcount; bp->bio_flags |= BIO_ERROR; } } else if (bp != NULL) { if (state == DA_CCB_DELETE) bp->bio_resid = 0; else bp->bio_resid = csio->resid; bp->bio_error = 0; if (bp->bio_resid != 0) bp->bio_flags |= BIO_ERROR; } if ((done_ccb->ccb_h.status & CAM_DEV_QFRZN) != 0) cam_release_devq(done_ccb->ccb_h.path, /*relsim_flags*/0, /*reduction*/0, /*timeout*/0, /*getcount_only*/0); } else if (bp != NULL) { if ((done_ccb->ccb_h.status & CAM_DEV_QFRZN) != 0) panic("REQ_CMP with QFRZN"); if (state == DA_CCB_DELETE) bp->bio_resid = 0; else bp->bio_resid = csio->resid; if (csio->resid > 0) bp->bio_flags |= BIO_ERROR; if (softc->error_inject != 0) { bp->bio_error = softc->error_inject; bp->bio_resid = bp->bio_bcount; bp->bio_flags |= BIO_ERROR; softc->error_inject = 0; } } LIST_REMOVE(&done_ccb->ccb_h, periph_links.le); if (LIST_EMPTY(&softc->pending_ccbs)) softc->flags |= DA_FLAG_WAS_OTAG; xpt_release_ccb(done_ccb); if (state == DA_CCB_DELETE) { TAILQ_HEAD(, bio) queue; TAILQ_INIT(&queue); TAILQ_CONCAT(&queue, &softc->delete_run_queue.queue, bio_queue); softc->delete_run_queue.insert_point = NULL; softc->delete_running = 0; daschedule(periph); cam_periph_unlock(periph); while ((bp1 = TAILQ_FIRST(&queue)) != NULL) { TAILQ_REMOVE(&queue, bp1, bio_queue); bp1->bio_error = bp->bio_error; if (bp->bio_flags & BIO_ERROR) { bp1->bio_flags |= BIO_ERROR; bp1->bio_resid = bp1->bio_bcount; } else bp1->bio_resid = 0; biodone(bp1); } } else cam_periph_unlock(periph); if (bp != NULL) biodone(bp); return; } case DA_CCB_PROBE_RC: case DA_CCB_PROBE_RC16: { struct scsi_read_capacity_data *rdcap; struct scsi_read_capacity_data_long *rcaplong; char announce_buf[80]; int lbp; lbp = 0; rdcap = NULL; rcaplong = NULL; if (state == DA_CCB_PROBE_RC) rdcap =(struct scsi_read_capacity_data *)csio->data_ptr; else rcaplong = (struct scsi_read_capacity_data_long *) csio->data_ptr; if ((csio->ccb_h.status & CAM_STATUS_MASK) == CAM_REQ_CMP) { struct disk_params *dp; uint32_t block_size; uint64_t maxsector; u_int lbppbe; /* LB per physical block exponent. */ u_int lalba; /* Lowest aligned LBA. */ if (state == DA_CCB_PROBE_RC) { block_size = scsi_4btoul(rdcap->length); maxsector = scsi_4btoul(rdcap->addr); lbppbe = 0; lalba = 0; /* * According to SBC-2, if the standard 10 * byte READ CAPACITY command returns 2^32, * we should issue the 16 byte version of * the command, since the device in question * has more sectors than can be represented * with the short version of the command. */ if (maxsector == 0xffffffff) { free(rdcap, M_SCSIDA); xpt_release_ccb(done_ccb); softc->state = DA_STATE_PROBE_RC16; xpt_schedule(periph, priority); return; } } else { block_size = scsi_4btoul(rcaplong->length); maxsector = scsi_8btou64(rcaplong->addr); lbppbe = rcaplong->prot_lbppbe & SRC16_LBPPBE; lalba = scsi_2btoul(rcaplong->lalba_lbp); } /* * Because GEOM code just will panic us if we * give them an 'illegal' value we'll avoid that * here. */ if (block_size == 0 && maxsector == 0) { block_size = 512; maxsector = -1; } if (block_size >= MAXPHYS || block_size == 0) { xpt_print(periph->path, "unsupportable block size %ju\n", (uintmax_t) block_size); announce_buf[0] = '\0'; cam_periph_invalidate(periph); } else { /* * We pass rcaplong into dasetgeom(), * because it will only use it if it is * non-NULL. */ dasetgeom(periph, block_size, maxsector, rcaplong, sizeof(*rcaplong)); lbp = (lalba & SRC16_LBPME_A); dp = &softc->params; snprintf(announce_buf, sizeof(announce_buf), "%juMB (%ju %u byte sectors: %dH %dS/T " "%dC)", (uintmax_t) (((uintmax_t)dp->secsize * dp->sectors) / (1024*1024)), (uintmax_t)dp->sectors, dp->secsize, dp->heads, dp->secs_per_track, dp->cylinders); } } else { int error; announce_buf[0] = '\0'; /* * Retry any UNIT ATTENTION type errors. They * are expected at boot. */ error = daerror(done_ccb, CAM_RETRY_SELTO, SF_RETRY_UA|SF_NO_PRINT); if (error == ERESTART) { /* * A retry was scheuled, so * just return. */ return; } else if (error != 0) { int asc, ascq; int sense_key, error_code; int have_sense; cam_status status; struct ccb_getdev cgd; /* Don't wedge this device's queue */ status = done_ccb->ccb_h.status; if ((status & CAM_DEV_QFRZN) != 0) cam_release_devq(done_ccb->ccb_h.path, /*relsim_flags*/0, /*reduction*/0, /*timeout*/0, /*getcount_only*/0); xpt_setup_ccb(&cgd.ccb_h, done_ccb->ccb_h.path, CAM_PRIORITY_NORMAL); cgd.ccb_h.func_code = XPT_GDEV_TYPE; xpt_action((union ccb *)&cgd); if (scsi_extract_sense_ccb(done_ccb, &error_code, &sense_key, &asc, &ascq)) have_sense = TRUE; else have_sense = FALSE; /* * If we tried READ CAPACITY(16) and failed, * fallback to READ CAPACITY(10). */ if ((state == DA_CCB_PROBE_RC16) && (softc->flags & DA_FLAG_CAN_RC16) && (((csio->ccb_h.status & CAM_STATUS_MASK) == CAM_REQ_INVALID) || ((have_sense) && (error_code == SSD_CURRENT_ERROR) && (sense_key == SSD_KEY_ILLEGAL_REQUEST)))) { softc->flags &= ~DA_FLAG_CAN_RC16; free(rdcap, M_SCSIDA); xpt_release_ccb(done_ccb); softc->state = DA_STATE_PROBE_RC; xpt_schedule(periph, priority); return; } else /* * Attach to anything that claims to be a * direct access or optical disk device, * as long as it doesn't return a "Logical * unit not supported" (0x25) error. */ if ((have_sense) && (asc != 0x25) && (error_code == SSD_CURRENT_ERROR)) { const char *sense_key_desc; const char *asc_desc; dasetgeom(periph, 512, -1, NULL, 0); scsi_sense_desc(sense_key, asc, ascq, &cgd.inq_data, &sense_key_desc, &asc_desc); snprintf(announce_buf, sizeof(announce_buf), "Attempt to query device " "size failed: %s, %s", sense_key_desc, asc_desc); } else { if (have_sense) scsi_sense_print( &done_ccb->csio); else { xpt_print(periph->path, "got CAM status %#x\n", done_ccb->ccb_h.status); } xpt_print(periph->path, "fatal error, " "failed to attach to device\n"); /* * Free up resources. */ cam_periph_invalidate(periph); } } } free(csio->data_ptr, M_SCSIDA); if (announce_buf[0] != '\0' && ((softc->flags & DA_FLAG_ANNOUNCED) == 0)) { /* * Create our sysctl variables, now that we know * we have successfully attached. */ /* increase the refcount */ if (cam_periph_acquire(periph) == CAM_REQ_CMP) { taskqueue_enqueue(taskqueue_thread, &softc->sysctl_task); xpt_announce_periph(periph, announce_buf); xpt_announce_quirks(periph, softc->quirks, DA_Q_BIT_STRING); } else { xpt_print(periph->path, "fatal error, " "could not acquire reference count\n"); } } /* We already probed the device. */ if (softc->flags & DA_FLAG_PROBED) { daprobedone(periph, done_ccb); return; } /* Ensure re-probe doesn't see old delete. */ softc->delete_available = 0; if (lbp && (softc->quirks & DA_Q_NO_UNMAP) == 0) { /* * Based on older SBC-3 spec revisions * any of the UNMAP methods "may" be * available via LBP given this flag so * we flag all of them as availble and * then remove those which further * probes confirm aren't available * later. * * We could also check readcap(16) p_type * flag to exclude one or more invalid * write same (X) types here */ dadeleteflag(softc, DA_DELETE_WS16, 1); dadeleteflag(softc, DA_DELETE_WS10, 1); dadeleteflag(softc, DA_DELETE_ZERO, 1); dadeleteflag(softc, DA_DELETE_UNMAP, 1); xpt_release_ccb(done_ccb); softc->state = DA_STATE_PROBE_LBP; xpt_schedule(periph, priority); return; } xpt_release_ccb(done_ccb); softc->state = DA_STATE_PROBE_BDC; xpt_schedule(periph, priority); return; } case DA_CCB_PROBE_LBP: { struct scsi_vpd_logical_block_prov *lbp; lbp = (struct scsi_vpd_logical_block_prov *)csio->data_ptr; if ((csio->ccb_h.status & CAM_STATUS_MASK) == CAM_REQ_CMP) { /* * T10/1799-D Revision 31 states at least one of these * must be supported but we don't currently enforce this. */ dadeleteflag(softc, DA_DELETE_WS16, (lbp->flags & SVPD_LBP_WS16)); dadeleteflag(softc, DA_DELETE_WS10, (lbp->flags & SVPD_LBP_WS10)); dadeleteflag(softc, DA_DELETE_ZERO, (lbp->flags & SVPD_LBP_WS10)); dadeleteflag(softc, DA_DELETE_UNMAP, (lbp->flags & SVPD_LBP_UNMAP)); } else { int error; error = daerror(done_ccb, CAM_RETRY_SELTO, SF_RETRY_UA|SF_NO_PRINT); if (error == ERESTART) return; else if (error != 0) { if ((done_ccb->ccb_h.status & CAM_DEV_QFRZN) != 0) { /* Don't wedge this device's queue */ cam_release_devq(done_ccb->ccb_h.path, /*relsim_flags*/0, /*reduction*/0, /*timeout*/0, /*getcount_only*/0); } /* * Failure indicates we don't support any SBC-3 * delete methods with UNMAP */ } } free(lbp, M_SCSIDA); xpt_release_ccb(done_ccb); softc->state = DA_STATE_PROBE_BLK_LIMITS; xpt_schedule(periph, priority); return; } case DA_CCB_PROBE_BLK_LIMITS: { struct scsi_vpd_block_limits *block_limits; block_limits = (struct scsi_vpd_block_limits *)csio->data_ptr; if ((csio->ccb_h.status & CAM_STATUS_MASK) == CAM_REQ_CMP) { uint32_t max_txfer_len = scsi_4btoul( block_limits->max_txfer_len); uint32_t max_unmap_lba_cnt = scsi_4btoul( block_limits->max_unmap_lba_cnt); uint32_t max_unmap_blk_cnt = scsi_4btoul( block_limits->max_unmap_blk_cnt); uint64_t ws_max_blks = scsi_8btou64( block_limits->max_write_same_length); if (max_txfer_len != 0) { softc->disk->d_maxsize = MIN(softc->maxio, (off_t)max_txfer_len * softc->params.secsize); } /* * We should already support UNMAP but we check lba * and block count to be sure */ if (max_unmap_lba_cnt != 0x00L && max_unmap_blk_cnt != 0x00L) { softc->unmap_max_lba = max_unmap_lba_cnt; softc->unmap_max_ranges = min(max_unmap_blk_cnt, UNMAP_MAX_RANGES); } else { /* * Unexpected UNMAP limits which means the * device doesn't actually support UNMAP */ dadeleteflag(softc, DA_DELETE_UNMAP, 0); } if (ws_max_blks != 0x00L) softc->ws_max_blks = ws_max_blks; } else { int error; error = daerror(done_ccb, CAM_RETRY_SELTO, SF_RETRY_UA|SF_NO_PRINT); if (error == ERESTART) return; else if (error != 0) { if ((done_ccb->ccb_h.status & CAM_DEV_QFRZN) != 0) { /* Don't wedge this device's queue */ cam_release_devq(done_ccb->ccb_h.path, /*relsim_flags*/0, /*reduction*/0, /*timeout*/0, /*getcount_only*/0); } /* * Failure here doesn't mean UNMAP is not * supported as this is an optional page. */ softc->unmap_max_lba = 1; softc->unmap_max_ranges = 1; } } free(block_limits, M_SCSIDA); xpt_release_ccb(done_ccb); softc->state = DA_STATE_PROBE_BDC; xpt_schedule(periph, priority); return; } case DA_CCB_PROBE_BDC: { struct scsi_vpd_block_characteristics *bdc; bdc = (struct scsi_vpd_block_characteristics *)csio->data_ptr; if ((csio->ccb_h.status & CAM_STATUS_MASK) == CAM_REQ_CMP) { /* * Disable queue sorting for non-rotational media * by default. */ - if (scsi_2btoul(bdc->medium_rotation_rate) == - SVPD_BDC_RATE_NONE_ROTATING) + u_int old_rate = softc->disk->d_rotation_rate; + + softc->disk->d_rotation_rate = + scsi_2btoul(bdc->medium_rotation_rate); + if (softc->disk->d_rotation_rate == + SVPD_BDC_RATE_NON_ROTATING) { softc->sort_io_queue = 0; + } + if (softc->disk->d_rotation_rate != old_rate) { + disk_attr_changed(softc->disk, + "GEOM::rotation_rate", M_NOWAIT); + } } else { int error; error = daerror(done_ccb, CAM_RETRY_SELTO, SF_RETRY_UA|SF_NO_PRINT); if (error == ERESTART) return; else if (error != 0) { if ((done_ccb->ccb_h.status & CAM_DEV_QFRZN) != 0) { /* Don't wedge this device's queue */ cam_release_devq(done_ccb->ccb_h.path, /*relsim_flags*/0, /*reduction*/0, /*timeout*/0, /*getcount_only*/0); } } } free(bdc, M_SCSIDA); xpt_release_ccb(done_ccb); softc->state = DA_STATE_PROBE_ATA; xpt_schedule(periph, priority); return; } case DA_CCB_PROBE_ATA: { int i; struct ata_params *ata_params; int16_t *ptr; ata_params = (struct ata_params *)csio->data_ptr; ptr = (uint16_t *)ata_params; if ((csio->ccb_h.status & CAM_STATUS_MASK) == CAM_REQ_CMP) { + uint16_t old_rate; + for (i = 0; i < sizeof(*ata_params) / 2; i++) ptr[i] = le16toh(ptr[i]); if (ata_params->support_dsm & ATA_SUPPORT_DSM_TRIM && (softc->quirks & DA_Q_NO_UNMAP) == 0) { dadeleteflag(softc, DA_DELETE_ATA_TRIM, 1); if (ata_params->max_dsm_blocks != 0) softc->trim_max_ranges = min( softc->trim_max_ranges, ata_params->max_dsm_blocks * ATA_DSM_BLK_RANGES); } /* * Disable queue sorting for non-rotational media * by default. */ - if (ata_params->media_rotation_rate == 1) + old_rate = softc->disk->d_rotation_rate; + softc->disk->d_rotation_rate = + ata_params->media_rotation_rate; + if (softc->disk->d_rotation_rate == + ATA_RATE_NON_ROTATING) { softc->sort_io_queue = 0; + } + + if (softc->disk->d_rotation_rate != old_rate) { + disk_attr_changed(softc->disk, + "GEOM::rotation_rate", M_NOWAIT); + } } else { int error; error = daerror(done_ccb, CAM_RETRY_SELTO, SF_RETRY_UA|SF_NO_PRINT); if (error == ERESTART) return; else if (error != 0) { if ((done_ccb->ccb_h.status & CAM_DEV_QFRZN) != 0) { /* Don't wedge this device's queue */ cam_release_devq(done_ccb->ccb_h.path, /*relsim_flags*/0, /*reduction*/0, /*timeout*/0, /*getcount_only*/0); } } } free(ata_params, M_SCSIDA); daprobedone(periph, done_ccb); return; } case DA_CCB_DUMP: /* No-op. We're polling */ return; case DA_CCB_TUR: { if ((done_ccb->ccb_h.status & CAM_STATUS_MASK) != CAM_REQ_CMP) { if (daerror(done_ccb, CAM_RETRY_SELTO, SF_RETRY_UA | SF_NO_RECOVERY | SF_NO_PRINT) == ERESTART) return; if ((done_ccb->ccb_h.status & CAM_DEV_QFRZN) != 0) cam_release_devq(done_ccb->ccb_h.path, /*relsim_flags*/0, /*reduction*/0, /*timeout*/0, /*getcount_only*/0); } xpt_release_ccb(done_ccb); cam_periph_release_locked(periph); return; } default: break; } xpt_release_ccb(done_ccb); } static void dareprobe(struct cam_periph *periph) { struct da_softc *softc; cam_status status; softc = (struct da_softc *)periph->softc; /* Probe in progress; don't interfere. */ if (softc->state != DA_STATE_NORMAL) return; status = cam_periph_acquire(periph); KASSERT(status == CAM_REQ_CMP, ("dareprobe: cam_periph_acquire failed")); if (softc->flags & DA_FLAG_CAN_RC16) softc->state = DA_STATE_PROBE_RC16; else softc->state = DA_STATE_PROBE_RC; xpt_schedule(periph, CAM_PRIORITY_DEV); } static int daerror(union ccb *ccb, u_int32_t cam_flags, u_int32_t sense_flags) { struct da_softc *softc; struct cam_periph *periph; int error, error_code, sense_key, asc, ascq; periph = xpt_path_periph(ccb->ccb_h.path); softc = (struct da_softc *)periph->softc; /* * Automatically detect devices that do not support * READ(6)/WRITE(6) and upgrade to using 10 byte cdbs. */ error = 0; if ((ccb->ccb_h.status & CAM_STATUS_MASK) == CAM_REQ_INVALID) { error = cmd6workaround(ccb); } else if (scsi_extract_sense_ccb(ccb, &error_code, &sense_key, &asc, &ascq)) { if (sense_key == SSD_KEY_ILLEGAL_REQUEST) error = cmd6workaround(ccb); /* * If the target replied with CAPACITY DATA HAS CHANGED UA, * query the capacity and notify upper layers. */ else if (sense_key == SSD_KEY_UNIT_ATTENTION && asc == 0x2A && ascq == 0x09) { xpt_print(periph->path, "Capacity data has changed\n"); softc->flags &= ~DA_FLAG_PROBED; dareprobe(periph); sense_flags |= SF_NO_PRINT; } else if (sense_key == SSD_KEY_UNIT_ATTENTION && asc == 0x28 && ascq == 0x00) { softc->flags &= ~DA_FLAG_PROBED; disk_media_changed(softc->disk, M_NOWAIT); } else if (sense_key == SSD_KEY_UNIT_ATTENTION && asc == 0x3F && ascq == 0x03) { xpt_print(periph->path, "INQUIRY data has changed\n"); softc->flags &= ~DA_FLAG_PROBED; dareprobe(periph); sense_flags |= SF_NO_PRINT; } else if (sense_key == SSD_KEY_NOT_READY && asc == 0x3a && (softc->flags & DA_FLAG_PACK_INVALID) == 0) { softc->flags |= DA_FLAG_PACK_INVALID; disk_media_gone(softc->disk, M_NOWAIT); } } if (error == ERESTART) return (ERESTART); /* * XXX * Until we have a better way of doing pack validation, * don't treat UAs as errors. */ sense_flags |= SF_RETRY_UA; return(cam_periph_error(ccb, cam_flags, sense_flags, &softc->saved_ccb)); } static void damediapoll(void *arg) { struct cam_periph *periph = arg; struct da_softc *softc = periph->softc; if (!softc->tur && LIST_EMPTY(&softc->pending_ccbs)) { if (cam_periph_acquire(periph) == CAM_REQ_CMP) { softc->tur = 1; daschedule(periph); } } /* Queue us up again */ if (da_poll_period != 0) callout_schedule(&softc->mediapoll_c, da_poll_period * hz); } static void daprevent(struct cam_periph *periph, int action) { struct da_softc *softc; union ccb *ccb; int error; softc = (struct da_softc *)periph->softc; if (((action == PR_ALLOW) && (softc->flags & DA_FLAG_PACK_LOCKED) == 0) || ((action == PR_PREVENT) && (softc->flags & DA_FLAG_PACK_LOCKED) != 0)) { return; } ccb = cam_periph_getccb(periph, CAM_PRIORITY_NORMAL); scsi_prevent(&ccb->csio, /*retries*/1, /*cbcfp*/dadone, MSG_SIMPLE_Q_TAG, action, SSD_FULL_SIZE, 5000); error = cam_periph_runccb(ccb, daerror, CAM_RETRY_SELTO, SF_RETRY_UA | SF_NO_PRINT, softc->disk->d_devstat); if (error == 0) { if (action == PR_ALLOW) softc->flags &= ~DA_FLAG_PACK_LOCKED; else softc->flags |= DA_FLAG_PACK_LOCKED; } xpt_release_ccb(ccb); } static void dasetgeom(struct cam_periph *periph, uint32_t block_len, uint64_t maxsector, struct scsi_read_capacity_data_long *rcaplong, size_t rcap_len) { struct ccb_calc_geometry ccg; struct da_softc *softc; struct disk_params *dp; u_int lbppbe, lalba; int error; softc = (struct da_softc *)periph->softc; dp = &softc->params; dp->secsize = block_len; dp->sectors = maxsector + 1; if (rcaplong != NULL) { lbppbe = rcaplong->prot_lbppbe & SRC16_LBPPBE; lalba = scsi_2btoul(rcaplong->lalba_lbp); lalba &= SRC16_LALBA_A; } else { lbppbe = 0; lalba = 0; } if (lbppbe > 0) { dp->stripesize = block_len << lbppbe; dp->stripeoffset = (dp->stripesize - block_len * lalba) % dp->stripesize; } else if (softc->quirks & DA_Q_4K) { dp->stripesize = 4096; dp->stripeoffset = 0; } else { dp->stripesize = 0; dp->stripeoffset = 0; } /* * Have the controller provide us with a geometry * for this disk. The only time the geometry * matters is when we boot and the controller * is the only one knowledgeable enough to come * up with something that will make this a bootable * device. */ xpt_setup_ccb(&ccg.ccb_h, periph->path, CAM_PRIORITY_NORMAL); ccg.ccb_h.func_code = XPT_CALC_GEOMETRY; ccg.block_size = dp->secsize; ccg.volume_size = dp->sectors; ccg.heads = 0; ccg.secs_per_track = 0; ccg.cylinders = 0; xpt_action((union ccb*)&ccg); if ((ccg.ccb_h.status & CAM_STATUS_MASK) != CAM_REQ_CMP) { /* * We don't know what went wrong here- but just pick * a geometry so we don't have nasty things like divide * by zero. */ dp->heads = 255; dp->secs_per_track = 255; dp->cylinders = dp->sectors / (255 * 255); if (dp->cylinders == 0) { dp->cylinders = 1; } } else { dp->heads = ccg.heads; dp->secs_per_track = ccg.secs_per_track; dp->cylinders = ccg.cylinders; } /* * If the user supplied a read capacity buffer, and if it is * different than the previous buffer, update the data in the EDT. * If it's the same, we don't bother. This avoids sending an * update every time someone opens this device. */ if ((rcaplong != NULL) && (bcmp(rcaplong, &softc->rcaplong, min(sizeof(softc->rcaplong), rcap_len)) != 0)) { struct ccb_dev_advinfo cdai; xpt_setup_ccb(&cdai.ccb_h, periph->path, CAM_PRIORITY_NORMAL); cdai.ccb_h.func_code = XPT_DEV_ADVINFO; cdai.buftype = CDAI_TYPE_RCAPLONG; cdai.flags |= CDAI_FLAG_STORE; cdai.bufsiz = rcap_len; cdai.buf = (uint8_t *)rcaplong; xpt_action((union ccb *)&cdai); if ((cdai.ccb_h.status & CAM_DEV_QFRZN) != 0) cam_release_devq(cdai.ccb_h.path, 0, 0, 0, FALSE); if (cdai.ccb_h.status != CAM_REQ_CMP) { xpt_print(periph->path, "%s: failed to set read " "capacity advinfo\n", __func__); /* Use cam_error_print() to decode the status */ cam_error_print((union ccb *)&cdai, CAM_ESF_CAM_STATUS, CAM_EPF_ALL); } else { bcopy(rcaplong, &softc->rcaplong, min(sizeof(softc->rcaplong), rcap_len)); } } softc->disk->d_sectorsize = softc->params.secsize; softc->disk->d_mediasize = softc->params.secsize * (off_t)softc->params.sectors; softc->disk->d_stripesize = softc->params.stripesize; softc->disk->d_stripeoffset = softc->params.stripeoffset; /* XXX: these are not actually "firmware" values, so they may be wrong */ softc->disk->d_fwsectors = softc->params.secs_per_track; softc->disk->d_fwheads = softc->params.heads; softc->disk->d_devstat->block_size = softc->params.secsize; softc->disk->d_devstat->flags &= ~DEVSTAT_BS_UNAVAILABLE; error = disk_resize(softc->disk, M_NOWAIT); if (error != 0) xpt_print(periph->path, "disk_resize(9) failed, error = %d\n", error); } static void dasendorderedtag(void *arg) { struct da_softc *softc = arg; if (da_send_ordered) { if (!LIST_EMPTY(&softc->pending_ccbs)) { if ((softc->flags & DA_FLAG_WAS_OTAG) == 0) softc->flags |= DA_FLAG_NEED_OTAG; softc->flags &= ~DA_FLAG_WAS_OTAG; } } /* Queue us up again */ callout_reset(&softc->sendordered_c, (da_default_timeout * hz) / DA_ORDEREDTAG_INTERVAL, dasendorderedtag, softc); } /* * Step through all DA peripheral drivers, and if the device is still open, * sync the disk cache to physical media. */ static void dashutdown(void * arg, int howto) { struct cam_periph *periph; struct da_softc *softc; union ccb *ccb; int error; CAM_PERIPH_FOREACH(periph, &dadriver) { softc = (struct da_softc *)periph->softc; if (SCHEDULER_STOPPED()) { /* If we paniced with the lock held, do not recurse. */ if (!cam_periph_owned(periph) && (softc->flags & DA_FLAG_OPEN)) { dadump(softc->disk, NULL, 0, 0, 0); } continue; } cam_periph_lock(periph); /* * We only sync the cache if the drive is still open, and * if the drive is capable of it.. */ if (((softc->flags & DA_FLAG_OPEN) == 0) || (softc->quirks & DA_Q_NO_SYNC_CACHE)) { cam_periph_unlock(periph); continue; } ccb = cam_periph_getccb(periph, CAM_PRIORITY_NORMAL); scsi_synchronize_cache(&ccb->csio, /*retries*/0, /*cbfcnp*/dadone, MSG_SIMPLE_Q_TAG, /*begin_lba*/0, /* whole disk */ /*lb_count*/0, SSD_FULL_SIZE, 60 * 60 * 1000); error = cam_periph_runccb(ccb, daerror, /*cam_flags*/0, /*sense_flags*/ SF_NO_RECOVERY | SF_NO_RETRY | SF_QUIET_IR, softc->disk->d_devstat); if (error != 0) xpt_print(periph->path, "Synchronize cache failed\n"); xpt_release_ccb(ccb); cam_periph_unlock(periph); } } #else /* !_KERNEL */ /* * XXX This is only left out of the kernel build to silence warnings. If, * for some reason this function is used in the kernel, the ifdefs should * be moved so it is included both in the kernel and userland. */ void scsi_format_unit(struct ccb_scsiio *csio, u_int32_t retries, void (*cbfcnp)(struct cam_periph *, union ccb *), u_int8_t tag_action, u_int8_t byte2, u_int16_t ileave, u_int8_t *data_ptr, u_int32_t dxfer_len, u_int8_t sense_len, u_int32_t timeout) { struct scsi_format_unit *scsi_cmd; scsi_cmd = (struct scsi_format_unit *)&csio->cdb_io.cdb_bytes; scsi_cmd->opcode = FORMAT_UNIT; scsi_cmd->byte2 = byte2; scsi_ulto2b(ileave, scsi_cmd->interleave); cam_fill_csio(csio, retries, cbfcnp, /*flags*/ (dxfer_len > 0) ? CAM_DIR_OUT : CAM_DIR_NONE, tag_action, data_ptr, dxfer_len, sense_len, sizeof(*scsi_cmd), timeout); } void scsi_sanitize(struct ccb_scsiio *csio, u_int32_t retries, void (*cbfcnp)(struct cam_periph *, union ccb *), u_int8_t tag_action, u_int8_t byte2, u_int16_t control, u_int8_t *data_ptr, u_int32_t dxfer_len, u_int8_t sense_len, u_int32_t timeout) { struct scsi_sanitize *scsi_cmd; scsi_cmd = (struct scsi_sanitize *)&csio->cdb_io.cdb_bytes; scsi_cmd->opcode = SANITIZE; scsi_cmd->byte2 = byte2; scsi_cmd->control = control; scsi_ulto2b(dxfer_len, scsi_cmd->length); cam_fill_csio(csio, retries, cbfcnp, /*flags*/ (dxfer_len > 0) ? CAM_DIR_OUT : CAM_DIR_NONE, tag_action, data_ptr, dxfer_len, sense_len, sizeof(*scsi_cmd), timeout); } #endif /* _KERNEL */ Index: stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/vdev.h =================================================================== --- stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/vdev.h (revision 271237) +++ stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/vdev.h (revision 271238) @@ -1,166 +1,169 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2013 by Delphix. All rights reserved. */ #ifndef _SYS_VDEV_H #define _SYS_VDEV_H #include #include #include #include #include #ifdef __cplusplus extern "C" { #endif typedef enum vdev_dtl_type { DTL_MISSING, /* 0% replication: no copies of the data */ DTL_PARTIAL, /* less than 100% replication: some copies missing */ DTL_SCRUB, /* unable to fully repair during scrub/resilver */ DTL_OUTAGE, /* temporarily missing (used to attempt detach) */ DTL_TYPES } vdev_dtl_type_t; extern boolean_t zfs_nocacheflush; extern boolean_t zfs_trim_enabled; extern int vdev_open(vdev_t *); extern void vdev_open_children(vdev_t *); extern boolean_t vdev_uses_zvols(vdev_t *); extern int vdev_validate(vdev_t *, boolean_t); extern void vdev_close(vdev_t *); extern int vdev_create(vdev_t *, uint64_t txg, boolean_t isreplace); extern void vdev_reopen(vdev_t *); extern int vdev_validate_aux(vdev_t *vd); extern zio_t *vdev_probe(vdev_t *vd, zio_t *pio); extern boolean_t vdev_is_bootable(vdev_t *vd); extern vdev_t *vdev_lookup_top(spa_t *spa, uint64_t vdev); extern vdev_t *vdev_lookup_by_guid(vdev_t *vd, uint64_t guid); extern void vdev_dtl_dirty(vdev_t *vd, vdev_dtl_type_t d, uint64_t txg, uint64_t size); extern boolean_t vdev_dtl_contains(vdev_t *vd, vdev_dtl_type_t d, uint64_t txg, uint64_t size); extern boolean_t vdev_dtl_empty(vdev_t *vd, vdev_dtl_type_t d); extern void vdev_dtl_reassess(vdev_t *vd, uint64_t txg, uint64_t scrub_txg, int scrub_done); extern boolean_t vdev_dtl_required(vdev_t *vd); extern boolean_t vdev_resilver_needed(vdev_t *vd, uint64_t *minp, uint64_t *maxp); extern void vdev_hold(vdev_t *); extern void vdev_rele(vdev_t *); extern int vdev_metaslab_init(vdev_t *vd, uint64_t txg); extern void vdev_metaslab_fini(vdev_t *vd); extern void vdev_metaslab_set_size(vdev_t *); extern void vdev_ashift_optimize(vdev_t *); extern void vdev_expand(vdev_t *vd, uint64_t txg); extern void vdev_split(vdev_t *vd); extern void vdev_deadman(vdev_t *vd); extern void vdev_get_stats(vdev_t *vd, vdev_stat_t *vs); extern void vdev_clear_stats(vdev_t *vd); extern void vdev_stat_update(zio_t *zio, uint64_t psize); extern void vdev_scan_stat_init(vdev_t *vd); extern void vdev_propagate_state(vdev_t *vd); extern void vdev_set_state(vdev_t *vd, boolean_t isopen, vdev_state_t state, vdev_aux_t aux); extern void vdev_space_update(vdev_t *vd, int64_t alloc_delta, int64_t defer_delta, int64_t space_delta); extern uint64_t vdev_psize_to_asize(vdev_t *vd, uint64_t psize); extern int vdev_fault(spa_t *spa, uint64_t guid, vdev_aux_t aux); extern int vdev_degrade(spa_t *spa, uint64_t guid, vdev_aux_t aux); extern int vdev_online(spa_t *spa, uint64_t guid, uint64_t flags, vdev_state_t *); extern int vdev_offline(spa_t *spa, uint64_t guid, uint64_t flags); extern void vdev_clear(spa_t *spa, vdev_t *vd); extern boolean_t vdev_is_dead(vdev_t *vd); extern boolean_t vdev_readable(vdev_t *vd); extern boolean_t vdev_writeable(vdev_t *vd); extern boolean_t vdev_allocatable(vdev_t *vd); extern boolean_t vdev_accessible(vdev_t *vd, zio_t *zio); extern void vdev_cache_init(vdev_t *vd); extern void vdev_cache_fini(vdev_t *vd); extern boolean_t vdev_cache_read(zio_t *zio); extern void vdev_cache_write(zio_t *zio); extern void vdev_cache_purge(vdev_t *vd); extern void vdev_queue_init(vdev_t *vd); extern void vdev_queue_fini(vdev_t *vd); extern zio_t *vdev_queue_io(zio_t *zio); extern void vdev_queue_io_done(zio_t *zio); +extern int vdev_queue_length(vdev_t *vd); +extern uint64_t vdev_queue_lastoffset(vdev_t *vd); +extern void vdev_queue_register_lastoffset(vdev_t *vd, zio_t *zio); extern void vdev_config_dirty(vdev_t *vd); extern void vdev_config_clean(vdev_t *vd); extern int vdev_config_sync(vdev_t **svd, int svdcount, uint64_t txg, boolean_t); extern void vdev_state_dirty(vdev_t *vd); extern void vdev_state_clean(vdev_t *vd); typedef enum vdev_config_flag { VDEV_CONFIG_SPARE = 1 << 0, VDEV_CONFIG_L2CACHE = 1 << 1, VDEV_CONFIG_REMOVING = 1 << 2 } vdev_config_flag_t; extern void vdev_top_config_generate(spa_t *spa, nvlist_t *config); extern nvlist_t *vdev_config_generate(spa_t *spa, vdev_t *vd, boolean_t getstats, vdev_config_flag_t flags); /* * Label routines */ struct uberblock; extern uint64_t vdev_label_offset(uint64_t psize, int l, uint64_t offset); extern int vdev_label_number(uint64_t psise, uint64_t offset); extern nvlist_t *vdev_label_read_config(vdev_t *vd, uint64_t txg); extern void vdev_uberblock_load(vdev_t *, struct uberblock *, nvlist_t **); typedef enum { VDEV_LABEL_CREATE, /* create/add a new device */ VDEV_LABEL_REPLACE, /* replace an existing device */ VDEV_LABEL_SPARE, /* add a new hot spare */ VDEV_LABEL_REMOVE, /* remove an existing device */ VDEV_LABEL_L2CACHE, /* add an L2ARC cache device */ VDEV_LABEL_SPLIT /* generating new label for split-off dev */ } vdev_labeltype_t; extern int vdev_label_init(vdev_t *vd, uint64_t txg, vdev_labeltype_t reason); #ifdef __cplusplus } #endif #endif /* _SYS_VDEV_H */ Index: stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/vdev_impl.h =================================================================== --- stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/vdev_impl.h (revision 271237) +++ stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/vdev_impl.h (revision 271238) @@ -1,379 +1,383 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2012, 2014 by Delphix. All rights reserved. */ #ifndef _SYS_VDEV_IMPL_H #define _SYS_VDEV_IMPL_H #include #include #include #include #include #include #include #include #ifdef __cplusplus extern "C" { #endif /* * Virtual device descriptors. * * All storage pool operations go through the virtual device framework, * which provides data replication and I/O scheduling. */ /* * Forward declarations that lots of things need. */ typedef struct vdev_queue vdev_queue_t; typedef struct vdev_cache vdev_cache_t; typedef struct vdev_cache_entry vdev_cache_entry_t; /* * Virtual device operations */ typedef int vdev_open_func_t(vdev_t *vd, uint64_t *size, uint64_t *max_size, uint64_t *logical_ashift, uint64_t *physical_ashift); typedef void vdev_close_func_t(vdev_t *vd); typedef uint64_t vdev_asize_func_t(vdev_t *vd, uint64_t psize); typedef int vdev_io_start_func_t(zio_t *zio); typedef void vdev_io_done_func_t(zio_t *zio); typedef void vdev_state_change_func_t(vdev_t *vd, int, int); typedef void vdev_hold_func_t(vdev_t *vd); typedef void vdev_rele_func_t(vdev_t *vd); typedef struct vdev_ops { vdev_open_func_t *vdev_op_open; vdev_close_func_t *vdev_op_close; vdev_asize_func_t *vdev_op_asize; vdev_io_start_func_t *vdev_op_io_start; vdev_io_done_func_t *vdev_op_io_done; vdev_state_change_func_t *vdev_op_state_change; vdev_hold_func_t *vdev_op_hold; vdev_rele_func_t *vdev_op_rele; char vdev_op_type[16]; boolean_t vdev_op_leaf; } vdev_ops_t; /* * Virtual device properties */ struct vdev_cache_entry { char *ve_data; uint64_t ve_offset; uint64_t ve_lastused; avl_node_t ve_offset_node; avl_node_t ve_lastused_node; uint32_t ve_hits; uint16_t ve_missed_update; zio_t *ve_fill_io; }; struct vdev_cache { avl_tree_t vc_offset_tree; avl_tree_t vc_lastused_tree; kmutex_t vc_lock; }; typedef struct vdev_queue_class { uint32_t vqc_active; /* * Sorted by offset or timestamp, depending on if the queue is * LBA-ordered vs FIFO. */ avl_tree_t vqc_queued_tree; } vdev_queue_class_t; struct vdev_queue { vdev_t *vq_vdev; vdev_queue_class_t vq_class[ZIO_PRIORITY_NUM_QUEUEABLE]; avl_tree_t vq_active_tree; uint64_t vq_last_offset; hrtime_t vq_io_complete_ts; /* time last i/o completed */ kmutex_t vq_lock; + uint64_t vq_lastoffset; }; /* * Virtual device descriptor */ struct vdev { /* * Common to all vdev types. */ uint64_t vdev_id; /* child number in vdev parent */ uint64_t vdev_guid; /* unique ID for this vdev */ uint64_t vdev_guid_sum; /* self guid + all child guids */ uint64_t vdev_orig_guid; /* orig. guid prior to remove */ uint64_t vdev_asize; /* allocatable device capacity */ uint64_t vdev_min_asize; /* min acceptable asize */ uint64_t vdev_max_asize; /* max acceptable asize */ uint64_t vdev_ashift; /* block alignment shift */ /* * Logical block alignment shift * * The smallest sized/aligned I/O supported by the device. */ uint64_t vdev_logical_ashift; /* * Physical block alignment shift * * The device supports logical I/Os with vdev_logical_ashift * size/alignment, but optimum performance will be achieved by * aligning/sizing requests to vdev_physical_ashift. Smaller * requests may be inflated or incur device level read-modify-write * operations. * * May be 0 to indicate no preference (i.e. use vdev_logical_ashift). */ uint64_t vdev_physical_ashift; uint64_t vdev_state; /* see VDEV_STATE_* #defines */ uint64_t vdev_prevstate; /* used when reopening a vdev */ vdev_ops_t *vdev_ops; /* vdev operations */ spa_t *vdev_spa; /* spa for this vdev */ void *vdev_tsd; /* type-specific data */ vnode_t *vdev_name_vp; /* vnode for pathname */ vnode_t *vdev_devid_vp; /* vnode for devid */ vdev_t *vdev_top; /* top-level vdev */ vdev_t *vdev_parent; /* parent vdev */ vdev_t **vdev_child; /* array of children */ uint64_t vdev_children; /* number of children */ vdev_stat_t vdev_stat; /* virtual device statistics */ boolean_t vdev_expanding; /* expand the vdev? */ boolean_t vdev_reopening; /* reopen in progress? */ int vdev_open_error; /* error on last open */ kthread_t *vdev_open_thread; /* thread opening children */ uint64_t vdev_crtxg; /* txg when top-level was added */ /* * Top-level vdev state. */ uint64_t vdev_ms_array; /* metaslab array object */ uint64_t vdev_ms_shift; /* metaslab size shift */ uint64_t vdev_ms_count; /* number of metaslabs */ metaslab_group_t *vdev_mg; /* metaslab group */ metaslab_t **vdev_ms; /* metaslab array */ txg_list_t vdev_ms_list; /* per-txg dirty metaslab lists */ txg_list_t vdev_dtl_list; /* per-txg dirty DTL lists */ txg_node_t vdev_txg_node; /* per-txg dirty vdev linkage */ boolean_t vdev_remove_wanted; /* async remove wanted? */ boolean_t vdev_probe_wanted; /* async probe wanted? */ list_node_t vdev_config_dirty_node; /* config dirty list */ list_node_t vdev_state_dirty_node; /* state dirty list */ uint64_t vdev_deflate_ratio; /* deflation ratio (x512) */ uint64_t vdev_islog; /* is an intent log device */ uint64_t vdev_removing; /* device is being removed? */ boolean_t vdev_ishole; /* is a hole in the namespace */ /* * Leaf vdev state. */ range_tree_t *vdev_dtl[DTL_TYPES]; /* dirty time logs */ space_map_t *vdev_dtl_sm; /* dirty time log space map */ txg_node_t vdev_dtl_node; /* per-txg dirty DTL linkage */ uint64_t vdev_dtl_object; /* DTL object */ uint64_t vdev_psize; /* physical device capacity */ uint64_t vdev_wholedisk; /* true if this is a whole disk */ uint64_t vdev_offline; /* persistent offline state */ uint64_t vdev_faulted; /* persistent faulted state */ uint64_t vdev_degraded; /* persistent degraded state */ uint64_t vdev_removed; /* persistent removed state */ uint64_t vdev_resilver_txg; /* persistent resilvering state */ uint64_t vdev_nparity; /* number of parity devices for raidz */ char *vdev_path; /* vdev path (if any) */ char *vdev_devid; /* vdev devid (if any) */ char *vdev_physpath; /* vdev device path (if any) */ char *vdev_fru; /* physical FRU location */ uint64_t vdev_not_present; /* not present during import */ uint64_t vdev_unspare; /* unspare when resilvering done */ boolean_t vdev_nowritecache; /* true if flushwritecache failed */ boolean_t vdev_notrim; /* true if trim failed */ boolean_t vdev_checkremove; /* temporary online test */ boolean_t vdev_forcefault; /* force online fault */ boolean_t vdev_splitting; /* split or repair in progress */ boolean_t vdev_delayed_close; /* delayed device close? */ boolean_t vdev_tmpoffline; /* device taken offline temporarily? */ boolean_t vdev_detached; /* device detached? */ boolean_t vdev_cant_read; /* vdev is failing all reads */ boolean_t vdev_cant_write; /* vdev is failing all writes */ boolean_t vdev_isspare; /* was a hot spare */ boolean_t vdev_isl2cache; /* was a l2cache device */ vdev_queue_t vdev_queue; /* I/O deadline schedule queue */ vdev_cache_t vdev_cache; /* physical block cache */ spa_aux_vdev_t *vdev_aux; /* for l2cache vdevs */ zio_t *vdev_probe_zio; /* root of current probe */ vdev_aux_t vdev_label_aux; /* on-disk aux state */ - struct trim_map *vdev_trimmap; + struct trim_map *vdev_trimmap; /* map on outstanding trims */ + uint16_t vdev_rotation_rate; /* rotational rate of the media */ +#define VDEV_RATE_UNKNOWN 0 +#define VDEV_RATE_NON_ROTATING 1 /* * For DTrace to work in userland (libzpool) context, these fields must * remain at the end of the structure. DTrace will use the kernel's * CTF definition for 'struct vdev', and since the size of a kmutex_t is * larger in userland, the offsets for the rest of the fields would be * incorrect. */ kmutex_t vdev_dtl_lock; /* vdev_dtl_{map,resilver} */ kmutex_t vdev_stat_lock; /* vdev_stat */ kmutex_t vdev_probe_lock; /* protects vdev_probe_zio */ }; #define VDEV_RAIDZ_MAXPARITY 3 #define VDEV_PAD_SIZE (8 << 10) /* 2 padding areas (vl_pad1 and vl_pad2) to skip */ #define VDEV_SKIP_SIZE VDEV_PAD_SIZE * 2 #define VDEV_PHYS_SIZE (112 << 10) #define VDEV_UBERBLOCK_RING (128 << 10) /* The largest uberblock we support is 8k. */ #define MAX_UBERBLOCK_SHIFT (13) #define VDEV_UBERBLOCK_SHIFT(vd) \ MIN(MAX((vd)->vdev_top->vdev_ashift, UBERBLOCK_SHIFT), \ MAX_UBERBLOCK_SHIFT) #define VDEV_UBERBLOCK_COUNT(vd) \ (VDEV_UBERBLOCK_RING >> VDEV_UBERBLOCK_SHIFT(vd)) #define VDEV_UBERBLOCK_OFFSET(vd, n) \ offsetof(vdev_label_t, vl_uberblock[(n) << VDEV_UBERBLOCK_SHIFT(vd)]) #define VDEV_UBERBLOCK_SIZE(vd) (1ULL << VDEV_UBERBLOCK_SHIFT(vd)) typedef struct vdev_phys { char vp_nvlist[VDEV_PHYS_SIZE - sizeof (zio_eck_t)]; zio_eck_t vp_zbt; } vdev_phys_t; typedef struct vdev_label { char vl_pad1[VDEV_PAD_SIZE]; /* 8K */ char vl_pad2[VDEV_PAD_SIZE]; /* 8K */ vdev_phys_t vl_vdev_phys; /* 112K */ char vl_uberblock[VDEV_UBERBLOCK_RING]; /* 128K */ } vdev_label_t; /* 256K total */ /* * vdev_dirty() flags */ #define VDD_METASLAB 0x01 #define VDD_DTL 0x02 /* Offset of embedded boot loader region on each label */ #define VDEV_BOOT_OFFSET (2 * sizeof (vdev_label_t)) /* * Size of embedded boot loader region on each label. * The total size of the first two labels plus the boot area is 4MB. */ #define VDEV_BOOT_SIZE (7ULL << 19) /* 3.5M */ /* * Size of label regions at the start and end of each leaf device. */ #define VDEV_LABEL_START_SIZE (2 * sizeof (vdev_label_t) + VDEV_BOOT_SIZE) #define VDEV_LABEL_END_SIZE (2 * sizeof (vdev_label_t)) #define VDEV_LABELS 4 #define VDEV_BEST_LABEL VDEV_LABELS #define VDEV_ALLOC_LOAD 0 #define VDEV_ALLOC_ADD 1 #define VDEV_ALLOC_SPARE 2 #define VDEV_ALLOC_L2CACHE 3 #define VDEV_ALLOC_ROOTPOOL 4 #define VDEV_ALLOC_SPLIT 5 #define VDEV_ALLOC_ATTACH 6 /* * Allocate or free a vdev */ extern vdev_t *vdev_alloc_common(spa_t *spa, uint_t id, uint64_t guid, vdev_ops_t *ops); extern int vdev_alloc(spa_t *spa, vdev_t **vdp, nvlist_t *config, vdev_t *parent, uint_t id, int alloctype); extern void vdev_free(vdev_t *vd); /* * Add or remove children and parents */ extern void vdev_add_child(vdev_t *pvd, vdev_t *cvd); extern void vdev_remove_child(vdev_t *pvd, vdev_t *cvd); extern void vdev_compact_children(vdev_t *pvd); extern vdev_t *vdev_add_parent(vdev_t *cvd, vdev_ops_t *ops); extern void vdev_remove_parent(vdev_t *cvd); /* * vdev sync load and sync */ extern void vdev_load_log_state(vdev_t *nvd, vdev_t *ovd); extern boolean_t vdev_log_state_valid(vdev_t *vd); extern void vdev_load(vdev_t *vd); extern int vdev_dtl_load(vdev_t *vd); extern void vdev_sync(vdev_t *vd, uint64_t txg); extern void vdev_sync_done(vdev_t *vd, uint64_t txg); extern void vdev_dirty(vdev_t *vd, int flags, void *arg, uint64_t txg); extern void vdev_dirty_leaves(vdev_t *vd, int flags, uint64_t txg); /* * Available vdev types. */ extern vdev_ops_t vdev_root_ops; extern vdev_ops_t vdev_mirror_ops; extern vdev_ops_t vdev_replacing_ops; extern vdev_ops_t vdev_raidz_ops; #ifdef _KERNEL extern vdev_ops_t vdev_geom_ops; #else extern vdev_ops_t vdev_disk_ops; #endif extern vdev_ops_t vdev_file_ops; extern vdev_ops_t vdev_missing_ops; extern vdev_ops_t vdev_hole_ops; extern vdev_ops_t vdev_spare_ops; /* * Common size functions */ extern uint64_t vdev_default_asize(vdev_t *vd, uint64_t psize); extern uint64_t vdev_get_min_asize(vdev_t *vd); extern void vdev_set_min_asize(vdev_t *vd); /* * Global variables */ /* zdb uses this tunable, so it must be declared here to make lint happy. */ extern int zfs_vdev_cache_size; #ifdef illumos /* * The vdev_buf_t is used to translate between zio_t and buf_t, and back again. */ typedef struct vdev_buf { buf_t vb_buf; /* buffer that describes the io */ zio_t *vb_io; /* pointer back to the original zio_t */ } vdev_buf_t; #endif #ifdef __cplusplus } #endif #endif /* _SYS_VDEV_IMPL_H */ Index: stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev_geom.c =================================================================== --- stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev_geom.c (revision 271237) +++ stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev_geom.c (revision 271238) @@ -1,863 +1,898 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2006 Pawel Jakub Dawidek * All rights reserved. * * Portions Copyright (c) 2012 Martin Matuska */ #include #include #include #include #include #include #include #include #include #include #include #include /* * Virtual device vector for GEOM. */ +static g_attrchanged_t vdev_geom_attrchanged; struct g_class zfs_vdev_class = { .name = "ZFS::VDEV", .version = G_VERSION, + .attrchanged = vdev_geom_attrchanged, }; DECLARE_GEOM_CLASS(zfs_vdev_class, zfs_vdev); SYSCTL_DECL(_vfs_zfs_vdev); /* Don't send BIO_FLUSH. */ static int vdev_geom_bio_flush_disable = 0; TUNABLE_INT("vfs.zfs.vdev.bio_flush_disable", &vdev_geom_bio_flush_disable); SYSCTL_INT(_vfs_zfs_vdev, OID_AUTO, bio_flush_disable, CTLFLAG_RW, &vdev_geom_bio_flush_disable, 0, "Disable BIO_FLUSH"); /* Don't send BIO_DELETE. */ static int vdev_geom_bio_delete_disable = 0; TUNABLE_INT("vfs.zfs.vdev.bio_delete_disable", &vdev_geom_bio_delete_disable); SYSCTL_INT(_vfs_zfs_vdev, OID_AUTO, bio_delete_disable, CTLFLAG_RW, &vdev_geom_bio_delete_disable, 0, "Disable BIO_DELETE"); static void +vdev_geom_set_rotation_rate(vdev_t *vd, struct g_consumer *cp) +{ + int error; + uint16_t rate; + + error = g_getattr("GEOM::rotation_rate", cp, &rate); + if (error == 0) + vd->vdev_rotation_rate = rate; + else + vd->vdev_rotation_rate = VDEV_RATE_UNKNOWN; +} + +static void +vdev_geom_attrchanged(struct g_consumer *cp, const char *attr) +{ + vdev_t *vd; + + vd = cp->private; + if (vd == NULL) + return; + + if (strcmp(attr, "GEOM::rotation_rate") == 0) { + vdev_geom_set_rotation_rate(vd, cp); + return; + } +} + +static void vdev_geom_orphan(struct g_consumer *cp) { vdev_t *vd; g_topology_assert(); vd = cp->private; if (vd == NULL) return; /* * Orphan callbacks occur from the GEOM event thread. * Concurrent with this call, new I/O requests may be * working their way through GEOM about to find out * (only once executed by the g_down thread) that we've * been orphaned from our disk provider. These I/Os * must be retired before we can detach our consumer. * This is most easily achieved by acquiring the * SPA ZIO configuration lock as a writer, but doing * so with the GEOM topology lock held would cause * a lock order reversal. Instead, rely on the SPA's * async removal support to invoke a close on this * vdev once it is safe to do so. */ zfs_post_remove(vd->vdev_spa, vd); vd->vdev_remove_wanted = B_TRUE; spa_async_request(vd->vdev_spa, SPA_ASYNC_REMOVE); } static struct g_consumer * vdev_geom_attach(struct g_provider *pp) { struct g_geom *gp; struct g_consumer *cp; g_topology_assert(); ZFS_LOG(1, "Attaching to %s.", pp->name); /* Do we have geom already? No? Create one. */ LIST_FOREACH(gp, &zfs_vdev_class.geom, geom) { if (gp->flags & G_GEOM_WITHER) continue; if (strcmp(gp->name, "zfs::vdev") != 0) continue; break; } if (gp == NULL) { gp = g_new_geomf(&zfs_vdev_class, "zfs::vdev"); gp->orphan = vdev_geom_orphan; cp = g_new_consumer(gp); if (g_attach(cp, pp) != 0) { g_wither_geom(gp, ENXIO); return (NULL); } if (g_access(cp, 1, 0, 1) != 0) { g_wither_geom(gp, ENXIO); return (NULL); } ZFS_LOG(1, "Created geom and consumer for %s.", pp->name); } else { /* Check if we are already connected to this provider. */ LIST_FOREACH(cp, &gp->consumer, consumer) { if (cp->provider == pp) { ZFS_LOG(1, "Found consumer for %s.", pp->name); break; } } if (cp == NULL) { cp = g_new_consumer(gp); if (g_attach(cp, pp) != 0) { g_destroy_consumer(cp); return (NULL); } if (g_access(cp, 1, 0, 1) != 0) { g_detach(cp); g_destroy_consumer(cp); return (NULL); } ZFS_LOG(1, "Created consumer for %s.", pp->name); } else { if (g_access(cp, 1, 0, 1) != 0) return (NULL); ZFS_LOG(1, "Used existing consumer for %s.", pp->name); } } cp->flags |= G_CF_DIRECT_SEND | G_CF_DIRECT_RECEIVE; return (cp); } static void vdev_geom_detach(void *arg, int flag __unused) { struct g_geom *gp; struct g_consumer *cp; g_topology_assert(); cp = arg; gp = cp->geom; ZFS_LOG(1, "Closing access to %s.", cp->provider->name); g_access(cp, -1, 0, -1); /* Destroy consumer on last close. */ if (cp->acr == 0 && cp->ace == 0) { ZFS_LOG(1, "Destroyed consumer to %s.", cp->provider->name); if (cp->acw > 0) g_access(cp, 0, -cp->acw, 0); g_detach(cp); g_destroy_consumer(cp); } /* Destroy geom if there are no consumers left. */ if (LIST_EMPTY(&gp->consumer)) { ZFS_LOG(1, "Destroyed geom %s.", gp->name); g_wither_geom(gp, ENXIO); } } static uint64_t nvlist_get_guid(nvlist_t *list) { uint64_t value; value = 0; nvlist_lookup_uint64(list, ZPOOL_CONFIG_GUID, &value); return (value); } static int vdev_geom_io(struct g_consumer *cp, int cmd, void *data, off_t offset, off_t size) { struct bio *bp; u_char *p; off_t off, maxio; int error; ASSERT((offset % cp->provider->sectorsize) == 0); ASSERT((size % cp->provider->sectorsize) == 0); bp = g_alloc_bio(); off = offset; offset += size; p = data; maxio = MAXPHYS - (MAXPHYS % cp->provider->sectorsize); error = 0; for (; off < offset; off += maxio, p += maxio, size -= maxio) { bzero(bp, sizeof(*bp)); bp->bio_cmd = cmd; bp->bio_done = NULL; bp->bio_offset = off; bp->bio_length = MIN(size, maxio); bp->bio_data = p; g_io_request(bp, cp); error = biowait(bp, "vdev_geom_io"); if (error != 0) break; } g_destroy_bio(bp); return (error); } static void vdev_geom_taste_orphan(struct g_consumer *cp) { KASSERT(1 == 0, ("%s called while tasting %s.", __func__, cp->provider->name)); } static int vdev_geom_read_config(struct g_consumer *cp, nvlist_t **config) { struct g_provider *pp; vdev_label_t *label; char *p, *buf; size_t buflen; uint64_t psize; off_t offset, size; uint64_t guid, state, txg; int error, l, len; g_topology_assert_not(); pp = cp->provider; ZFS_LOG(1, "Reading config from %s...", pp->name); psize = pp->mediasize; psize = P2ALIGN(psize, (uint64_t)sizeof(vdev_label_t)); size = sizeof(*label) + pp->sectorsize - ((sizeof(*label) - 1) % pp->sectorsize) - 1; guid = 0; label = kmem_alloc(size, KM_SLEEP); buflen = sizeof(label->vl_vdev_phys.vp_nvlist); *config = NULL; for (l = 0; l < VDEV_LABELS; l++) { offset = vdev_label_offset(psize, l, 0); if ((offset % pp->sectorsize) != 0) continue; if (vdev_geom_io(cp, BIO_READ, label, offset, size) != 0) continue; buf = label->vl_vdev_phys.vp_nvlist; if (nvlist_unpack(buf, buflen, config, 0) != 0) continue; if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_STATE, &state) != 0 || state > POOL_STATE_L2CACHE) { nvlist_free(*config); *config = NULL; continue; } if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE && (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_TXG, &txg) != 0 || txg == 0)) { nvlist_free(*config); *config = NULL; continue; } break; } kmem_free(label, size); return (*config == NULL ? ENOENT : 0); } static void resize_configs(nvlist_t ***configs, uint64_t *count, uint64_t id) { nvlist_t **new_configs; uint64_t i; if (id < *count) return; new_configs = kmem_zalloc((id + 1) * sizeof(nvlist_t *), KM_SLEEP); for (i = 0; i < *count; i++) new_configs[i] = (*configs)[i]; if (*configs != NULL) kmem_free(*configs, *count * sizeof(void *)); *configs = new_configs; *count = id + 1; } static void process_vdev_config(nvlist_t ***configs, uint64_t *count, nvlist_t *cfg, const char *name, uint64_t* known_pool_guid) { nvlist_t *vdev_tree; uint64_t pool_guid; uint64_t vdev_guid, known_guid; uint64_t id, txg, known_txg; char *pname; int i; if (nvlist_lookup_string(cfg, ZPOOL_CONFIG_POOL_NAME, &pname) != 0 || strcmp(pname, name) != 0) goto ignore; if (nvlist_lookup_uint64(cfg, ZPOOL_CONFIG_POOL_GUID, &pool_guid) != 0) goto ignore; if (nvlist_lookup_uint64(cfg, ZPOOL_CONFIG_TOP_GUID, &vdev_guid) != 0) goto ignore; if (nvlist_lookup_nvlist(cfg, ZPOOL_CONFIG_VDEV_TREE, &vdev_tree) != 0) goto ignore; if (nvlist_lookup_uint64(vdev_tree, ZPOOL_CONFIG_ID, &id) != 0) goto ignore; VERIFY(nvlist_lookup_uint64(cfg, ZPOOL_CONFIG_POOL_TXG, &txg) == 0); if (*known_pool_guid != 0) { if (pool_guid != *known_pool_guid) goto ignore; } else *known_pool_guid = pool_guid; resize_configs(configs, count, id); if ((*configs)[id] != NULL) { VERIFY(nvlist_lookup_uint64((*configs)[id], ZPOOL_CONFIG_POOL_TXG, &known_txg) == 0); if (txg <= known_txg) goto ignore; nvlist_free((*configs)[id]); } (*configs)[id] = cfg; return; ignore: nvlist_free(cfg); } static int vdev_geom_attach_taster(struct g_consumer *cp, struct g_provider *pp) { int error; if (pp->flags & G_PF_WITHER) return (EINVAL); g_attach(cp, pp); error = g_access(cp, 1, 0, 0); if (error == 0) { if (pp->sectorsize > VDEV_PAD_SIZE || !ISP2(pp->sectorsize)) error = EINVAL; else if (pp->mediasize < SPA_MINDEVSIZE) error = EINVAL; if (error != 0) g_access(cp, -1, 0, 0); } if (error != 0) g_detach(cp); return (error); } static void vdev_geom_detach_taster(struct g_consumer *cp) { g_access(cp, -1, 0, 0); g_detach(cp); } int vdev_geom_read_pool_label(const char *name, nvlist_t ***configs, uint64_t *count) { struct g_class *mp; struct g_geom *gp, *zgp; struct g_provider *pp; struct g_consumer *zcp; nvlist_t *vdev_cfg; uint64_t pool_guid; int error; DROP_GIANT(); g_topology_lock(); zgp = g_new_geomf(&zfs_vdev_class, "zfs::vdev::taste"); /* This orphan function should be never called. */ zgp->orphan = vdev_geom_taste_orphan; zcp = g_new_consumer(zgp); *configs = NULL; *count = 0; pool_guid = 0; LIST_FOREACH(mp, &g_classes, class) { if (mp == &zfs_vdev_class) continue; LIST_FOREACH(gp, &mp->geom, geom) { if (gp->flags & G_GEOM_WITHER) continue; LIST_FOREACH(pp, &gp->provider, provider) { if (pp->flags & G_PF_WITHER) continue; if (vdev_geom_attach_taster(zcp, pp) != 0) continue; g_topology_unlock(); error = vdev_geom_read_config(zcp, &vdev_cfg); g_topology_lock(); vdev_geom_detach_taster(zcp); if (error) continue; ZFS_LOG(1, "successfully read vdev config"); process_vdev_config(configs, count, vdev_cfg, name, &pool_guid); } } } g_destroy_consumer(zcp); g_destroy_geom(zgp); g_topology_unlock(); PICKUP_GIANT(); return (*count > 0 ? 0 : ENOENT); } static uint64_t vdev_geom_read_guid(struct g_consumer *cp) { nvlist_t *config; uint64_t guid; g_topology_assert_not(); guid = 0; if (vdev_geom_read_config(cp, &config) == 0) { guid = nvlist_get_guid(config); nvlist_free(config); } return (guid); } static struct g_consumer * vdev_geom_attach_by_guid(uint64_t guid) { struct g_class *mp; struct g_geom *gp, *zgp; struct g_provider *pp; struct g_consumer *cp, *zcp; uint64_t pguid; g_topology_assert(); zgp = g_new_geomf(&zfs_vdev_class, "zfs::vdev::taste"); /* This orphan function should be never called. */ zgp->orphan = vdev_geom_taste_orphan; zcp = g_new_consumer(zgp); cp = NULL; LIST_FOREACH(mp, &g_classes, class) { if (mp == &zfs_vdev_class) continue; LIST_FOREACH(gp, &mp->geom, geom) { if (gp->flags & G_GEOM_WITHER) continue; LIST_FOREACH(pp, &gp->provider, provider) { if (vdev_geom_attach_taster(zcp, pp) != 0) continue; g_topology_unlock(); pguid = vdev_geom_read_guid(zcp); g_topology_lock(); vdev_geom_detach_taster(zcp); if (pguid != guid) continue; cp = vdev_geom_attach(pp); if (cp == NULL) { printf("ZFS WARNING: Unable to attach to %s.\n", pp->name); continue; } break; } if (cp != NULL) break; } if (cp != NULL) break; } end: g_destroy_consumer(zcp); g_destroy_geom(zgp); return (cp); } static struct g_consumer * vdev_geom_open_by_guid(vdev_t *vd) { struct g_consumer *cp; char *buf; size_t len; g_topology_assert(); ZFS_LOG(1, "Searching by guid [%ju].", (uintmax_t)vd->vdev_guid); cp = vdev_geom_attach_by_guid(vd->vdev_guid); if (cp != NULL) { len = strlen(cp->provider->name) + strlen("/dev/") + 1; buf = kmem_alloc(len, KM_SLEEP); snprintf(buf, len, "/dev/%s", cp->provider->name); spa_strfree(vd->vdev_path); vd->vdev_path = buf; ZFS_LOG(1, "Attach by guid [%ju] succeeded, provider %s.", (uintmax_t)vd->vdev_guid, vd->vdev_path); } else { ZFS_LOG(1, "Search by guid [%ju] failed.", (uintmax_t)vd->vdev_guid); } return (cp); } static struct g_consumer * vdev_geom_open_by_path(vdev_t *vd, int check_guid) { struct g_provider *pp; struct g_consumer *cp; uint64_t guid; g_topology_assert(); cp = NULL; pp = g_provider_by_name(vd->vdev_path + sizeof("/dev/") - 1); if (pp != NULL) { ZFS_LOG(1, "Found provider by name %s.", vd->vdev_path); cp = vdev_geom_attach(pp); if (cp != NULL && check_guid && ISP2(pp->sectorsize) && pp->sectorsize <= VDEV_PAD_SIZE) { g_topology_unlock(); guid = vdev_geom_read_guid(cp); g_topology_lock(); if (guid != vd->vdev_guid) { vdev_geom_detach(cp, 0); cp = NULL; ZFS_LOG(1, "guid mismatch for provider %s: " "%ju != %ju.", vd->vdev_path, (uintmax_t)vd->vdev_guid, (uintmax_t)guid); } else { ZFS_LOG(1, "guid match for provider %s.", vd->vdev_path); } } } return (cp); } static int vdev_geom_open(vdev_t *vd, uint64_t *psize, uint64_t *max_psize, uint64_t *logical_ashift, uint64_t *physical_ashift) { struct g_provider *pp; struct g_consumer *cp; size_t bufsize; int error; /* * We must have a pathname, and it must be absolute. */ if (vd->vdev_path == NULL || vd->vdev_path[0] != '/') { vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL; return (EINVAL); } vd->vdev_tsd = NULL; DROP_GIANT(); g_topology_lock(); error = 0; /* * If we're creating or splitting a pool, just find the GEOM provider * by its name and ignore GUID mismatches. */ if (vd->vdev_spa->spa_load_state == SPA_LOAD_NONE || vd->vdev_spa->spa_splitting_newspa == B_TRUE) cp = vdev_geom_open_by_path(vd, 0); else { cp = vdev_geom_open_by_path(vd, 1); if (cp == NULL) { /* * The device at vd->vdev_path doesn't have the * expected guid. The disks might have merely * moved around so try all other GEOM providers * to find one with the right guid. */ cp = vdev_geom_open_by_guid(vd); } } if (cp == NULL) { ZFS_LOG(1, "Provider %s not found.", vd->vdev_path); error = ENOENT; } else if (cp->provider->sectorsize > VDEV_PAD_SIZE || !ISP2(cp->provider->sectorsize)) { ZFS_LOG(1, "Provider %s has unsupported sectorsize.", vd->vdev_path); vdev_geom_detach(cp, 0); error = EINVAL; cp = NULL; } else if (cp->acw == 0 && (spa_mode(vd->vdev_spa) & FWRITE) != 0) { int i; for (i = 0; i < 5; i++) { error = g_access(cp, 0, 1, 0); if (error == 0) break; g_topology_unlock(); tsleep(vd, 0, "vdev", hz / 2); g_topology_lock(); } if (error != 0) { printf("ZFS WARNING: Unable to open %s for writing (error=%d).\n", vd->vdev_path, error); vdev_geom_detach(cp, 0); cp = NULL; } } g_topology_unlock(); PICKUP_GIANT(); if (cp == NULL) { vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED; return (error); } cp->private = vd; vd->vdev_tsd = cp; pp = cp->provider; /* * Determine the actual size of the device. */ *max_psize = *psize = pp->mediasize; /* * Determine the device's minimum transfer size and preferred * transfer size. */ *logical_ashift = highbit(MAX(pp->sectorsize, SPA_MINBLOCKSIZE)) - 1; *physical_ashift = 0; if (pp->stripesize) *physical_ashift = highbit(pp->stripesize) - 1; /* * Clear the nowritecache settings, so that on a vdev_reopen() * we will try again. */ vd->vdev_nowritecache = B_FALSE; if (vd->vdev_physpath != NULL) spa_strfree(vd->vdev_physpath); bufsize = sizeof("/dev/") + strlen(pp->name); vd->vdev_physpath = kmem_alloc(bufsize, KM_SLEEP); snprintf(vd->vdev_physpath, bufsize, "/dev/%s", pp->name); + + /* + * Determine the device's rotation rate. + */ + vdev_geom_set_rotation_rate(vd, cp); return (0); } static void vdev_geom_close(vdev_t *vd) { struct g_consumer *cp; cp = vd->vdev_tsd; if (cp == NULL) return; vd->vdev_tsd = NULL; vd->vdev_delayed_close = B_FALSE; cp->private = NULL; /* XXX locking */ g_post_event(vdev_geom_detach, cp, M_WAITOK, NULL); } static void vdev_geom_io_intr(struct bio *bp) { vdev_t *vd; zio_t *zio; zio = bp->bio_caller1; vd = zio->io_vd; zio->io_error = bp->bio_error; if (zio->io_error == 0 && bp->bio_resid != 0) zio->io_error = SET_ERROR(EIO); switch(zio->io_error) { case ENOTSUP: /* * If we get ENOTSUP for BIO_FLUSH or BIO_DELETE we know * that future attempts will never succeed. In this case * we set a persistent flag so that we don't bother with * requests in the future. */ switch(bp->bio_cmd) { case BIO_FLUSH: vd->vdev_nowritecache = B_TRUE; break; case BIO_DELETE: vd->vdev_notrim = B_TRUE; break; } break; case ENXIO: if (!vd->vdev_remove_wanted) { /* * If provider's error is set we assume it is being * removed. */ if (bp->bio_to->error != 0) { vd->vdev_remove_wanted = B_TRUE; spa_async_request(zio->io_spa, SPA_ASYNC_REMOVE); } else if (!vd->vdev_delayed_close) { vd->vdev_delayed_close = B_TRUE; } } break; } g_destroy_bio(bp); zio_interrupt(zio); } static int vdev_geom_io_start(zio_t *zio) { vdev_t *vd; struct g_consumer *cp; struct bio *bp; int error; vd = zio->io_vd; switch (zio->io_type) { case ZIO_TYPE_IOCTL: /* XXPOLICY */ if (!vdev_readable(vd)) { zio->io_error = SET_ERROR(ENXIO); } else { switch (zio->io_cmd) { case DKIOCFLUSHWRITECACHE: if (zfs_nocacheflush || vdev_geom_bio_flush_disable) break; if (vd->vdev_nowritecache) { zio->io_error = SET_ERROR(ENOTSUP); break; } goto sendreq; default: zio->io_error = SET_ERROR(ENOTSUP); } } zio_interrupt(zio); return (ZIO_PIPELINE_STOP); case ZIO_TYPE_FREE: if (vd->vdev_notrim) { zio->io_error = SET_ERROR(ENOTSUP); } else if (!vdev_geom_bio_delete_disable) { goto sendreq; } zio_interrupt(zio); return (ZIO_PIPELINE_STOP); } sendreq: cp = vd->vdev_tsd; if (cp == NULL) { zio->io_error = SET_ERROR(ENXIO); zio_interrupt(zio); return (ZIO_PIPELINE_STOP); } bp = g_alloc_bio(); bp->bio_caller1 = zio; switch (zio->io_type) { case ZIO_TYPE_READ: case ZIO_TYPE_WRITE: bp->bio_cmd = zio->io_type == ZIO_TYPE_READ ? BIO_READ : BIO_WRITE; bp->bio_data = zio->io_data; bp->bio_offset = zio->io_offset; bp->bio_length = zio->io_size; break; case ZIO_TYPE_FREE: bp->bio_cmd = BIO_DELETE; bp->bio_data = NULL; bp->bio_offset = zio->io_offset; bp->bio_length = zio->io_size; break; case ZIO_TYPE_IOCTL: bp->bio_cmd = BIO_FLUSH; bp->bio_flags |= BIO_ORDERED; bp->bio_data = NULL; bp->bio_offset = cp->provider->mediasize; bp->bio_length = 0; break; } bp->bio_done = vdev_geom_io_intr; g_io_request(bp, cp); return (ZIO_PIPELINE_STOP); } static void vdev_geom_io_done(zio_t *zio) { } static void vdev_geom_hold(vdev_t *vd) { } static void vdev_geom_rele(vdev_t *vd) { } vdev_ops_t vdev_geom_ops = { vdev_geom_open, vdev_geom_close, vdev_default_asize, vdev_geom_io_start, vdev_geom_io_done, NULL, vdev_geom_hold, vdev_geom_rele, VDEV_TYPE_DISK, /* name of this vdev type */ B_TRUE /* leaf vdev */ }; Index: stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev_mirror.c =================================================================== --- stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev_mirror.c (revision 271237) +++ stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev_mirror.c (revision 271238) @@ -1,498 +1,667 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright 2010 Sun Microsystems, Inc. All rights reserved. * Use is subject to license terms. */ /* * Copyright (c) 2013 by Delphix. All rights reserved. */ #include #include #include #include #include /* * Virtual device vector for mirroring. */ typedef struct mirror_child { vdev_t *mc_vd; uint64_t mc_offset; int mc_error; + int mc_load; uint8_t mc_tried; uint8_t mc_skipped; uint8_t mc_speculative; } mirror_child_t; typedef struct mirror_map { + int *mm_preferred; + int mm_preferred_cnt; int mm_children; - int mm_replacing; - int mm_preferred; - int mm_root; - mirror_child_t mm_child[1]; + boolean_t mm_replacing; + boolean_t mm_root; + mirror_child_t mm_child[]; } mirror_map_t; -int vdev_mirror_shift = 21; +static int vdev_mirror_shift = 21; +SYSCTL_DECL(_vfs_zfs_vdev); +static SYSCTL_NODE(_vfs_zfs_vdev, OID_AUTO, mirror, CTLFLAG_RD, 0, + "ZFS VDEV Mirror"); + +/* + * The load configuration settings below are tuned by default for + * the case where all devices are of the same rotational type. + * + * If there is a mixture of rotating and non-rotating media, setting + * non_rotating_seek_inc to 0 may well provide better results as it + * will direct more reads to the non-rotating vdevs which are more + * likely to have a higher performance. + */ + +/* Rotating media load calculation configuration. */ +static int rotating_inc = 0; +TUNABLE_INT("vfs.zfs.vdev.mirror.rotating_inc", &rotating_inc); +SYSCTL_INT(_vfs_zfs_vdev_mirror, OID_AUTO, rotating_inc, CTLFLAG_RW, + &rotating_inc, 0, "Rotating media load increment for non-seeking I/O's"); + +static int rotating_seek_inc = 5; +TUNABLE_INT("vfs.zfs.vdev.mirror.rotating_seek_inc", &rotating_seek_inc); +SYSCTL_INT(_vfs_zfs_vdev_mirror, OID_AUTO, rotating_seek_inc, CTLFLAG_RW, + &rotating_seek_inc, 0, "Rotating media load increment for seeking I/O's"); + +static int rotating_seek_offset = 1 * 1024 * 1024; +TUNABLE_INT("vfs.zfs.vdev.mirror.rotating_seek_offset", &rotating_seek_offset); +SYSCTL_INT(_vfs_zfs_vdev_mirror, OID_AUTO, rotating_seek_offset, CTLFLAG_RW, + &rotating_seek_offset, 0, "Offset in bytes from the last I/O which " + "triggers a reduced rotating media seek increment"); + +/* Non-rotating media load calculation configuration. */ +static int non_rotating_inc = 0; +TUNABLE_INT("vfs.zfs.vdev.mirror.non_rotating_inc", &non_rotating_inc); +SYSCTL_INT(_vfs_zfs_vdev_mirror, OID_AUTO, non_rotating_inc, CTLFLAG_RW, + &non_rotating_inc, 0, + "Non-rotating media load increment for non-seeking I/O's"); + +static int non_rotating_seek_inc = 1; +TUNABLE_INT("vfs.zfs.vdev.mirror.non_rotating_seek_inc", + &non_rotating_seek_inc); +SYSCTL_INT(_vfs_zfs_vdev_mirror, OID_AUTO, non_rotating_seek_inc, CTLFLAG_RW, + &non_rotating_seek_inc, 0, + "Non-rotating media load increment for seeking I/O's"); + + +static inline size_t +vdev_mirror_map_size(int children) +{ + return (offsetof(mirror_map_t, mm_child[children]) + + sizeof(int) * children); +} + +static inline mirror_map_t * +vdev_mirror_map_alloc(int children, boolean_t replacing, boolean_t root) +{ + mirror_map_t *mm; + + mm = kmem_zalloc(vdev_mirror_map_size(children), KM_SLEEP); + mm->mm_children = children; + mm->mm_replacing = replacing; + mm->mm_root = root; + mm->mm_preferred = (int *)((uintptr_t)mm + + offsetof(mirror_map_t, mm_child[children])); + + return mm; +} + static void vdev_mirror_map_free(zio_t *zio) { mirror_map_t *mm = zio->io_vsd; - kmem_free(mm, offsetof(mirror_map_t, mm_child[mm->mm_children])); + kmem_free(mm, vdev_mirror_map_size(mm->mm_children)); } static const zio_vsd_ops_t vdev_mirror_vsd_ops = { vdev_mirror_map_free, zio_vsd_default_cksum_report }; +static int +vdev_mirror_load(mirror_map_t *mm, vdev_t *vd, uint64_t zio_offset) +{ + uint64_t lastoffset; + int load; + + /* All DVAs have equal weight at the root. */ + if (mm->mm_root) + return (INT_MAX); + + /* + * We don't return INT_MAX if the device is resilvering i.e. + * vdev_resilver_txg != 0 as when tested performance was slightly + * worse overall when resilvering with compared to without. + */ + + /* Standard load based on pending queue length. */ + load = vdev_queue_length(vd); + lastoffset = vdev_queue_lastoffset(vd); + + if (vd->vdev_rotation_rate == VDEV_RATE_NON_ROTATING) { + /* Non-rotating media. */ + if (lastoffset == zio_offset) + return (load + non_rotating_inc); + + /* + * Apply a seek penalty even for non-rotating devices as + * sequential I/O'a can be aggregated into fewer operations + * on the device, thus avoiding unnecessary per-command + * overhead and boosting performance. + */ + return (load + non_rotating_seek_inc); + } + + /* Rotating media I/O's which directly follow the last I/O. */ + if (lastoffset == zio_offset) + return (load + rotating_inc); + + /* + * Apply half the seek increment to I/O's within seek offset + * of the last I/O queued to this vdev as they should incure less + * of a seek increment. + */ + if (ABS(lastoffset - zio_offset) < rotating_seek_offset) + return (load + (rotating_seek_inc / 2)); + + /* Apply the full seek increment to all other I/O's. */ + return (load + rotating_seek_inc); +} + + static mirror_map_t * -vdev_mirror_map_alloc(zio_t *zio) +vdev_mirror_map_init(zio_t *zio) { mirror_map_t *mm = NULL; mirror_child_t *mc; vdev_t *vd = zio->io_vd; - int c, d; + int c; if (vd == NULL) { dva_t *dva = zio->io_bp->blk_dva; spa_t *spa = zio->io_spa; - c = BP_GET_NDVAS(zio->io_bp); - - mm = kmem_zalloc(offsetof(mirror_map_t, mm_child[c]), KM_SLEEP); - mm->mm_children = c; - mm->mm_replacing = B_FALSE; - mm->mm_preferred = spa_get_random(c); - mm->mm_root = B_TRUE; - - /* - * Check the other, lower-index DVAs to see if they're on - * the same vdev as the child we picked. If they are, use - * them since they are likely to have been allocated from - * the primary metaslab in use at the time, and hence are - * more likely to have locality with single-copy data. - */ - for (c = mm->mm_preferred, d = c - 1; d >= 0; d--) { - if (DVA_GET_VDEV(&dva[d]) == DVA_GET_VDEV(&dva[c])) - mm->mm_preferred = d; - } - + mm = vdev_mirror_map_alloc(BP_GET_NDVAS(zio->io_bp), B_FALSE, + B_TRUE); for (c = 0; c < mm->mm_children; c++) { mc = &mm->mm_child[c]; - mc->mc_vd = vdev_lookup_top(spa, DVA_GET_VDEV(&dva[c])); mc->mc_offset = DVA_GET_OFFSET(&dva[c]); } } else { - c = vd->vdev_children; - - mm = kmem_zalloc(offsetof(mirror_map_t, mm_child[c]), KM_SLEEP); - mm->mm_children = c; - mm->mm_replacing = (vd->vdev_ops == &vdev_replacing_ops || - vd->vdev_ops == &vdev_spare_ops); - mm->mm_preferred = mm->mm_replacing ? 0 : - (zio->io_offset >> vdev_mirror_shift) % c; - mm->mm_root = B_FALSE; - + mm = vdev_mirror_map_alloc(vd->vdev_children, + (vd->vdev_ops == &vdev_replacing_ops || + vd->vdev_ops == &vdev_spare_ops), B_FALSE); for (c = 0; c < mm->mm_children; c++) { mc = &mm->mm_child[c]; mc->mc_vd = vd->vdev_child[c]; mc->mc_offset = zio->io_offset; } } zio->io_vsd = mm; zio->io_vsd_ops = &vdev_mirror_vsd_ops; return (mm); } static int vdev_mirror_open(vdev_t *vd, uint64_t *asize, uint64_t *max_asize, uint64_t *logical_ashift, uint64_t *physical_ashift) { int numerrors = 0; int lasterror = 0; if (vd->vdev_children == 0) { vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL; return (SET_ERROR(EINVAL)); } vdev_open_children(vd); for (int c = 0; c < vd->vdev_children; c++) { vdev_t *cvd = vd->vdev_child[c]; if (cvd->vdev_open_error) { lasterror = cvd->vdev_open_error; numerrors++; continue; } *asize = MIN(*asize - 1, cvd->vdev_asize - 1) + 1; *max_asize = MIN(*max_asize - 1, cvd->vdev_max_asize - 1) + 1; *logical_ashift = MAX(*logical_ashift, cvd->vdev_ashift); *physical_ashift = MAX(*physical_ashift, cvd->vdev_physical_ashift); } if (numerrors == vd->vdev_children) { vd->vdev_stat.vs_aux = VDEV_AUX_NO_REPLICAS; return (lasterror); } return (0); } static void vdev_mirror_close(vdev_t *vd) { for (int c = 0; c < vd->vdev_children; c++) vdev_close(vd->vdev_child[c]); } static void vdev_mirror_child_done(zio_t *zio) { mirror_child_t *mc = zio->io_private; mc->mc_error = zio->io_error; mc->mc_tried = 1; mc->mc_skipped = 0; } static void vdev_mirror_scrub_done(zio_t *zio) { mirror_child_t *mc = zio->io_private; if (zio->io_error == 0) { zio_t *pio; mutex_enter(&zio->io_lock); while ((pio = zio_walk_parents(zio)) != NULL) { mutex_enter(&pio->io_lock); ASSERT3U(zio->io_size, >=, pio->io_size); bcopy(zio->io_data, pio->io_data, pio->io_size); mutex_exit(&pio->io_lock); } mutex_exit(&zio->io_lock); } zio_buf_free(zio->io_data, zio->io_size); mc->mc_error = zio->io_error; mc->mc_tried = 1; mc->mc_skipped = 0; } /* - * Try to find a child whose DTL doesn't contain the block we want to read. + * Check the other, lower-index DVAs to see if they're on the same + * vdev as the child we picked. If they are, use them since they + * are likely to have been allocated from the primary metaslab in + * use at the time, and hence are more likely to have locality with + * single-copy data. + */ +static int +vdev_mirror_dva_select(zio_t *zio, int p) +{ + dva_t *dva = zio->io_bp->blk_dva; + mirror_map_t *mm = zio->io_vsd; + int preferred; + int c; + + preferred = mm->mm_preferred[p]; + for (p-- ; p >= 0; p--) { + c = mm->mm_preferred[p]; + if (DVA_GET_VDEV(&dva[c]) == DVA_GET_VDEV(&dva[preferred])) + preferred = c; + } + return (preferred); +} + +static int +vdev_mirror_preferred_child_randomize(zio_t *zio) +{ + mirror_map_t *mm = zio->io_vsd; + int p; + + if (mm->mm_root) { + p = spa_get_random(mm->mm_preferred_cnt); + return (vdev_mirror_dva_select(zio, p)); + } + + /* + * To ensure we don't always favour the first matching vdev, + * which could lead to wear leveling issues on SSD's, we + * use the I/O offset as a pseudo random seed into the vdevs + * which have the lowest load. + */ + p = (zio->io_offset >> vdev_mirror_shift) % mm->mm_preferred_cnt; + return (mm->mm_preferred[p]); +} + +/* + * Try to find a vdev whose DTL doesn't contain the block we want to read + * prefering vdevs based on determined load. + * * If we can't, try the read on any vdev we haven't already tried. */ static int vdev_mirror_child_select(zio_t *zio) { mirror_map_t *mm = zio->io_vsd; - mirror_child_t *mc; uint64_t txg = zio->io_txg; - int i, c; + int c, lowest_load; ASSERT(zio->io_bp == NULL || BP_PHYSICAL_BIRTH(zio->io_bp) == txg); - /* - * Try to find a child whose DTL doesn't contain the block to read. - * If a child is known to be completely inaccessible (indicated by - * vdev_readable() returning B_FALSE), don't even try. - */ - for (i = 0, c = mm->mm_preferred; i < mm->mm_children; i++, c++) { - if (c >= mm->mm_children) - c = 0; + lowest_load = INT_MAX; + mm->mm_preferred_cnt = 0; + for (c = 0; c < mm->mm_children; c++) { + mirror_child_t *mc; + mc = &mm->mm_child[c]; if (mc->mc_tried || mc->mc_skipped) continue; + if (!vdev_readable(mc->mc_vd)) { mc->mc_error = SET_ERROR(ENXIO); mc->mc_tried = 1; /* don't even try */ mc->mc_skipped = 1; continue; } - if (!vdev_dtl_contains(mc->mc_vd, DTL_MISSING, txg, 1)) - return (c); - mc->mc_error = SET_ERROR(ESTALE); - mc->mc_skipped = 1; - mc->mc_speculative = 1; + + if (vdev_dtl_contains(mc->mc_vd, DTL_MISSING, txg, 1)) { + mc->mc_error = SET_ERROR(ESTALE); + mc->mc_skipped = 1; + mc->mc_speculative = 1; + continue; + } + + mc->mc_load = vdev_mirror_load(mm, mc->mc_vd, mc->mc_offset); + if (mc->mc_load > lowest_load) + continue; + + if (mc->mc_load < lowest_load) { + lowest_load = mc->mc_load; + mm->mm_preferred_cnt = 0; + } + mm->mm_preferred[mm->mm_preferred_cnt] = c; + mm->mm_preferred_cnt++; } + if (mm->mm_preferred_cnt == 1) { + vdev_queue_register_lastoffset( + mm->mm_child[mm->mm_preferred[0]].mc_vd, zio); + return (mm->mm_preferred[0]); + } + + if (mm->mm_preferred_cnt > 1) { + int c = vdev_mirror_preferred_child_randomize(zio); + + vdev_queue_register_lastoffset(mm->mm_child[c].mc_vd, zio); + return (c); + } + /* * Every device is either missing or has this txg in its DTL. * Look for any child we haven't already tried before giving up. */ - for (c = 0; c < mm->mm_children; c++) - if (!mm->mm_child[c].mc_tried) + for (c = 0; c < mm->mm_children; c++) { + if (!mm->mm_child[c].mc_tried) { + vdev_queue_register_lastoffset(mm->mm_child[c].mc_vd, + zio); return (c); + } + } /* * Every child failed. There's no place left to look. */ return (-1); } static int vdev_mirror_io_start(zio_t *zio) { mirror_map_t *mm; mirror_child_t *mc; int c, children; - mm = vdev_mirror_map_alloc(zio); + mm = vdev_mirror_map_init(zio); if (zio->io_type == ZIO_TYPE_READ) { if ((zio->io_flags & ZIO_FLAG_SCRUB) && !mm->mm_replacing) { /* * For scrubbing reads we need to allocate a read * buffer for each child and issue reads to all * children. If any child succeeds, it will copy its * data into zio->io_data in vdev_mirror_scrub_done. */ for (c = 0; c < mm->mm_children; c++) { mc = &mm->mm_child[c]; zio_nowait(zio_vdev_child_io(zio, zio->io_bp, mc->mc_vd, mc->mc_offset, zio_buf_alloc(zio->io_size), zio->io_size, zio->io_type, zio->io_priority, 0, vdev_mirror_scrub_done, mc)); } zio_interrupt(zio); return (ZIO_PIPELINE_STOP); } /* * For normal reads just pick one child. */ c = vdev_mirror_child_select(zio); children = (c >= 0); } else { ASSERT(zio->io_type == ZIO_TYPE_WRITE || zio->io_type == ZIO_TYPE_FREE); /* * Writes and frees go to all children. */ c = 0; children = mm->mm_children; } while (children--) { mc = &mm->mm_child[c]; zio_nowait(zio_vdev_child_io(zio, zio->io_bp, mc->mc_vd, mc->mc_offset, zio->io_data, zio->io_size, zio->io_type, zio->io_priority, 0, vdev_mirror_child_done, mc)); c++; } zio_interrupt(zio); return (ZIO_PIPELINE_STOP); } static int vdev_mirror_worst_error(mirror_map_t *mm) { int error[2] = { 0, 0 }; for (int c = 0; c < mm->mm_children; c++) { mirror_child_t *mc = &mm->mm_child[c]; int s = mc->mc_speculative; error[s] = zio_worst_error(error[s], mc->mc_error); } return (error[0] ? error[0] : error[1]); } static void vdev_mirror_io_done(zio_t *zio) { mirror_map_t *mm = zio->io_vsd; mirror_child_t *mc; int c; int good_copies = 0; int unexpected_errors = 0; for (c = 0; c < mm->mm_children; c++) { mc = &mm->mm_child[c]; if (mc->mc_error) { if (!mc->mc_skipped) unexpected_errors++; } else if (mc->mc_tried) { good_copies++; } } if (zio->io_type == ZIO_TYPE_WRITE) { /* * XXX -- for now, treat partial writes as success. * * Now that we support write reallocation, it would be better * to treat partial failure as real failure unless there are * no non-degraded top-level vdevs left, and not update DTLs * if we intend to reallocate. */ /* XXPOLICY */ if (good_copies != mm->mm_children) { /* * Always require at least one good copy. * * For ditto blocks (io_vd == NULL), require * all copies to be good. * * XXX -- for replacing vdevs, there's no great answer. * If the old device is really dead, we may not even * be able to access it -- so we only want to * require good writes to the new device. But if * the new device turns out to be flaky, we want * to be able to detach it -- which requires all * writes to the old device to have succeeded. */ if (good_copies == 0 || zio->io_vd == NULL) zio->io_error = vdev_mirror_worst_error(mm); } return; } else if (zio->io_type == ZIO_TYPE_FREE) { return; } ASSERT(zio->io_type == ZIO_TYPE_READ); /* * If we don't have a good copy yet, keep trying other children. */ /* XXPOLICY */ if (good_copies == 0 && (c = vdev_mirror_child_select(zio)) != -1) { ASSERT(c >= 0 && c < mm->mm_children); mc = &mm->mm_child[c]; zio_vdev_io_redone(zio); zio_nowait(zio_vdev_child_io(zio, zio->io_bp, mc->mc_vd, mc->mc_offset, zio->io_data, zio->io_size, ZIO_TYPE_READ, zio->io_priority, 0, vdev_mirror_child_done, mc)); return; } /* XXPOLICY */ if (good_copies == 0) { zio->io_error = vdev_mirror_worst_error(mm); ASSERT(zio->io_error != 0); } if (good_copies && spa_writeable(zio->io_spa) && (unexpected_errors || (zio->io_flags & ZIO_FLAG_RESILVER) || ((zio->io_flags & ZIO_FLAG_SCRUB) && mm->mm_replacing))) { /* * Use the good data we have in hand to repair damaged children. */ for (c = 0; c < mm->mm_children; c++) { /* * Don't rewrite known good children. * Not only is it unnecessary, it could * actually be harmful: if the system lost * power while rewriting the only good copy, * there would be no good copies left! */ mc = &mm->mm_child[c]; if (mc->mc_error == 0) { if (mc->mc_tried) continue; if (!(zio->io_flags & ZIO_FLAG_SCRUB) && !vdev_dtl_contains(mc->mc_vd, DTL_PARTIAL, zio->io_txg, 1)) continue; mc->mc_error = SET_ERROR(ESTALE); } zio_nowait(zio_vdev_child_io(zio, zio->io_bp, mc->mc_vd, mc->mc_offset, zio->io_data, zio->io_size, ZIO_TYPE_WRITE, ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_IO_REPAIR | (unexpected_errors ? ZIO_FLAG_SELF_HEAL : 0), NULL, NULL)); } } } static void vdev_mirror_state_change(vdev_t *vd, int faulted, int degraded) { if (faulted == vd->vdev_children) vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN, VDEV_AUX_NO_REPLICAS); else if (degraded + faulted != 0) vdev_set_state(vd, B_FALSE, VDEV_STATE_DEGRADED, VDEV_AUX_NONE); else vdev_set_state(vd, B_FALSE, VDEV_STATE_HEALTHY, VDEV_AUX_NONE); } vdev_ops_t vdev_mirror_ops = { vdev_mirror_open, vdev_mirror_close, vdev_default_asize, vdev_mirror_io_start, vdev_mirror_io_done, vdev_mirror_state_change, NULL, NULL, VDEV_TYPE_MIRROR, /* name of this vdev type */ B_FALSE /* not a leaf vdev */ }; vdev_ops_t vdev_replacing_ops = { vdev_mirror_open, vdev_mirror_close, vdev_default_asize, vdev_mirror_io_start, vdev_mirror_io_done, vdev_mirror_state_change, NULL, NULL, VDEV_TYPE_REPLACING, /* name of this vdev type */ B_FALSE /* not a leaf vdev */ }; vdev_ops_t vdev_spare_ops = { vdev_mirror_open, vdev_mirror_close, vdev_default_asize, vdev_mirror_io_start, vdev_mirror_io_done, vdev_mirror_state_change, NULL, NULL, VDEV_TYPE_SPARE, /* name of this vdev type */ B_FALSE /* not a leaf vdev */ }; Index: stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev_queue.c =================================================================== --- stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev_queue.c (revision 271237) +++ stable/10/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev_queue.c (revision 271238) @@ -1,817 +1,842 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright 2009 Sun Microsystems, Inc. All rights reserved. * Use is subject to license terms. */ /* * Copyright (c) 2012, 2014 by Delphix. All rights reserved. */ #include #include #include #include #include #include /* * ZFS I/O Scheduler * --------------- * * ZFS issues I/O operations to leaf vdevs to satisfy and complete zios. The * I/O scheduler determines when and in what order those operations are * issued. The I/O scheduler divides operations into six I/O classes * prioritized in the following order: sync read, sync write, async read, * async write, scrub/resilver and trim. Each queue defines the minimum and * maximum number of concurrent operations that may be issued to the device. * In addition, the device has an aggregate maximum. Note that the sum of the * per-queue minimums must not exceed the aggregate maximum, and if the * aggregate maximum is equal to or greater than the sum of the per-queue * maximums, the per-queue minimum has no effect. * * For many physical devices, throughput increases with the number of * concurrent operations, but latency typically suffers. Further, physical * devices typically have a limit at which more concurrent operations have no * effect on throughput or can actually cause it to decrease. * * The scheduler selects the next operation to issue by first looking for an * I/O class whose minimum has not been satisfied. Once all are satisfied and * the aggregate maximum has not been hit, the scheduler looks for classes * whose maximum has not been satisfied. Iteration through the I/O classes is * done in the order specified above. No further operations are issued if the * aggregate maximum number of concurrent operations has been hit or if there * are no operations queued for an I/O class that has not hit its maximum. * Every time an I/O is queued or an operation completes, the I/O scheduler * looks for new operations to issue. * * All I/O classes have a fixed maximum number of outstanding operations * except for the async write class. Asynchronous writes represent the data * that is committed to stable storage during the syncing stage for * transaction groups (see txg.c). Transaction groups enter the syncing state * periodically so the number of queued async writes will quickly burst up and * then bleed down to zero. Rather than servicing them as quickly as possible, * the I/O scheduler changes the maximum number of active async write I/Os * according to the amount of dirty data in the pool (see dsl_pool.c). Since * both throughput and latency typically increase with the number of * concurrent operations issued to physical devices, reducing the burstiness * in the number of concurrent operations also stabilizes the response time of * operations from other -- and in particular synchronous -- queues. In broad * strokes, the I/O scheduler will issue more concurrent operations from the * async write queue as there's more dirty data in the pool. * * Async Writes * * The number of concurrent operations issued for the async write I/O class * follows a piece-wise linear function defined by a few adjustable points. * * | o---------| <-- zfs_vdev_async_write_max_active * ^ | /^ | * | | / | | * active | / | | * I/O | / | | * count | / | | * | / | | * |------------o | | <-- zfs_vdev_async_write_min_active * 0|____________^______|_________| * 0% | | 100% of zfs_dirty_data_max * | | * | `-- zfs_vdev_async_write_active_max_dirty_percent * `--------- zfs_vdev_async_write_active_min_dirty_percent * * Until the amount of dirty data exceeds a minimum percentage of the dirty * data allowed in the pool, the I/O scheduler will limit the number of * concurrent operations to the minimum. As that threshold is crossed, the * number of concurrent operations issued increases linearly to the maximum at * the specified maximum percentage of the dirty data allowed in the pool. * * Ideally, the amount of dirty data on a busy pool will stay in the sloped * part of the function between zfs_vdev_async_write_active_min_dirty_percent * and zfs_vdev_async_write_active_max_dirty_percent. If it exceeds the * maximum percentage, this indicates that the rate of incoming data is * greater than the rate that the backend storage can handle. In this case, we * must further throttle incoming writes (see dmu_tx_delay() for details). */ /* * The maximum number of I/Os active to each device. Ideally, this will be >= * the sum of each queue's max_active. It must be at least the sum of each * queue's min_active. */ uint32_t zfs_vdev_max_active = 1000; /* * Per-queue limits on the number of I/Os active to each device. If the * sum of the queue's max_active is < zfs_vdev_max_active, then the * min_active comes into play. We will send min_active from each queue, * and then select from queues in the order defined by zio_priority_t. * * In general, smaller max_active's will lead to lower latency of synchronous * operations. Larger max_active's may lead to higher overall throughput, * depending on underlying storage. * * The ratio of the queues' max_actives determines the balance of performance * between reads, writes, and scrubs. E.g., increasing * zfs_vdev_scrub_max_active will cause the scrub or resilver to complete * more quickly, but reads and writes to have higher latency and lower * throughput. */ uint32_t zfs_vdev_sync_read_min_active = 10; uint32_t zfs_vdev_sync_read_max_active = 10; uint32_t zfs_vdev_sync_write_min_active = 10; uint32_t zfs_vdev_sync_write_max_active = 10; uint32_t zfs_vdev_async_read_min_active = 1; uint32_t zfs_vdev_async_read_max_active = 3; uint32_t zfs_vdev_async_write_min_active = 1; uint32_t zfs_vdev_async_write_max_active = 10; uint32_t zfs_vdev_scrub_min_active = 1; uint32_t zfs_vdev_scrub_max_active = 2; uint32_t zfs_vdev_trim_min_active = 1; /* * TRIM max active is large in comparison to the other values due to the fact * that TRIM IOs are coalesced at the device layer. This value is set such * that a typical SSD can process the queued IOs in a single request. */ uint32_t zfs_vdev_trim_max_active = 64; /* * When the pool has less than zfs_vdev_async_write_active_min_dirty_percent * dirty data, use zfs_vdev_async_write_min_active. When it has more than * zfs_vdev_async_write_active_max_dirty_percent, use * zfs_vdev_async_write_max_active. The value is linearly interpolated * between min and max. */ int zfs_vdev_async_write_active_min_dirty_percent = 30; int zfs_vdev_async_write_active_max_dirty_percent = 60; /* * To reduce IOPs, we aggregate small adjacent I/Os into one large I/O. * For read I/Os, we also aggregate across small adjacency gaps; for writes * we include spans of optional I/Os to aid aggregation at the disk even when * they aren't able to help us aggregate at this level. */ int zfs_vdev_aggregation_limit = SPA_MAXBLOCKSIZE; int zfs_vdev_read_gap_limit = 32 << 10; int zfs_vdev_write_gap_limit = 4 << 10; #ifdef __FreeBSD__ SYSCTL_DECL(_vfs_zfs_vdev); TUNABLE_INT("vfs.zfs.vdev.max_active", &zfs_vdev_max_active); SYSCTL_UINT(_vfs_zfs_vdev, OID_AUTO, max_active, CTLFLAG_RW, &zfs_vdev_max_active, 0, "The maximum number of I/Os of all types active for each device."); #define ZFS_VDEV_QUEUE_KNOB_MIN(name) \ TUNABLE_INT("vfs.zfs.vdev." #name "_min_active", \ &zfs_vdev_ ## name ## _min_active); \ SYSCTL_UINT(_vfs_zfs_vdev, OID_AUTO, name ## _min_active, CTLFLAG_RW, \ &zfs_vdev_ ## name ## _min_active, 0, \ "Initial number of I/O requests of type " #name \ " active for each device"); #define ZFS_VDEV_QUEUE_KNOB_MAX(name) \ TUNABLE_INT("vfs.zfs.vdev." #name "_max_active", \ &zfs_vdev_ ## name ## _max_active); \ SYSCTL_UINT(_vfs_zfs_vdev, OID_AUTO, name ## _max_active, CTLFLAG_RW, \ &zfs_vdev_ ## name ## _max_active, 0, \ "Maximum number of I/O requests of type " #name \ " active for each device"); ZFS_VDEV_QUEUE_KNOB_MIN(sync_read); ZFS_VDEV_QUEUE_KNOB_MAX(sync_read); ZFS_VDEV_QUEUE_KNOB_MIN(sync_write); ZFS_VDEV_QUEUE_KNOB_MAX(sync_write); ZFS_VDEV_QUEUE_KNOB_MIN(async_read); ZFS_VDEV_QUEUE_KNOB_MAX(async_read); ZFS_VDEV_QUEUE_KNOB_MIN(async_write); ZFS_VDEV_QUEUE_KNOB_MAX(async_write); ZFS_VDEV_QUEUE_KNOB_MIN(scrub); ZFS_VDEV_QUEUE_KNOB_MAX(scrub); ZFS_VDEV_QUEUE_KNOB_MIN(trim); ZFS_VDEV_QUEUE_KNOB_MAX(trim); #undef ZFS_VDEV_QUEUE_KNOB TUNABLE_INT("vfs.zfs.vdev.aggregation_limit", &zfs_vdev_aggregation_limit); SYSCTL_INT(_vfs_zfs_vdev, OID_AUTO, aggregation_limit, CTLFLAG_RW, &zfs_vdev_aggregation_limit, 0, "I/O requests are aggregated up to this size"); TUNABLE_INT("vfs.zfs.vdev.read_gap_limit", &zfs_vdev_read_gap_limit); SYSCTL_INT(_vfs_zfs_vdev, OID_AUTO, read_gap_limit, CTLFLAG_RW, &zfs_vdev_read_gap_limit, 0, "Acceptable gap between two reads being aggregated"); TUNABLE_INT("vfs.zfs.vdev.write_gap_limit", &zfs_vdev_write_gap_limit); SYSCTL_INT(_vfs_zfs_vdev, OID_AUTO, write_gap_limit, CTLFLAG_RW, &zfs_vdev_write_gap_limit, 0, "Acceptable gap between two writes being aggregated"); #endif int vdev_queue_offset_compare(const void *x1, const void *x2) { const zio_t *z1 = x1; const zio_t *z2 = x2; if (z1->io_offset < z2->io_offset) return (-1); if (z1->io_offset > z2->io_offset) return (1); if (z1 < z2) return (-1); if (z1 > z2) return (1); return (0); } int vdev_queue_timestamp_compare(const void *x1, const void *x2) { const zio_t *z1 = x1; const zio_t *z2 = x2; if (z1->io_timestamp < z2->io_timestamp) return (-1); if (z1->io_timestamp > z2->io_timestamp) return (1); if (z1 < z2) return (-1); if (z1 > z2) return (1); return (0); } void vdev_queue_init(vdev_t *vd) { vdev_queue_t *vq = &vd->vdev_queue; mutex_init(&vq->vq_lock, NULL, MUTEX_DEFAULT, NULL); vq->vq_vdev = vd; avl_create(&vq->vq_active_tree, vdev_queue_offset_compare, sizeof (zio_t), offsetof(struct zio, io_queue_node)); for (zio_priority_t p = 0; p < ZIO_PRIORITY_NUM_QUEUEABLE; p++) { /* * The synchronous i/o queues are FIFO rather than LBA ordered. * This provides more consistent latency for these i/os, and * they tend to not be tightly clustered anyway so there is * little to no throughput loss. */ boolean_t fifo = (p == ZIO_PRIORITY_SYNC_READ || p == ZIO_PRIORITY_SYNC_WRITE); avl_create(&vq->vq_class[p].vqc_queued_tree, fifo ? vdev_queue_timestamp_compare : vdev_queue_offset_compare, sizeof (zio_t), offsetof(struct zio, io_queue_node)); } + + vq->vq_lastoffset = 0; } void vdev_queue_fini(vdev_t *vd) { vdev_queue_t *vq = &vd->vdev_queue; for (zio_priority_t p = 0; p < ZIO_PRIORITY_NUM_QUEUEABLE; p++) avl_destroy(&vq->vq_class[p].vqc_queued_tree); avl_destroy(&vq->vq_active_tree); mutex_destroy(&vq->vq_lock); } static void vdev_queue_io_add(vdev_queue_t *vq, zio_t *zio) { spa_t *spa = zio->io_spa; ASSERT(MUTEX_HELD(&vq->vq_lock)); ASSERT3U(zio->io_priority, <, ZIO_PRIORITY_NUM_QUEUEABLE); avl_add(&vq->vq_class[zio->io_priority].vqc_queued_tree, zio); #ifdef illumos mutex_enter(&spa->spa_iokstat_lock); spa->spa_queue_stats[zio->io_priority].spa_queued++; if (spa->spa_iokstat != NULL) kstat_waitq_enter(spa->spa_iokstat->ks_data); mutex_exit(&spa->spa_iokstat_lock); #endif } static void vdev_queue_io_remove(vdev_queue_t *vq, zio_t *zio) { spa_t *spa = zio->io_spa; ASSERT(MUTEX_HELD(&vq->vq_lock)); ASSERT3U(zio->io_priority, <, ZIO_PRIORITY_NUM_QUEUEABLE); avl_remove(&vq->vq_class[zio->io_priority].vqc_queued_tree, zio); #ifdef illumos mutex_enter(&spa->spa_iokstat_lock); ASSERT3U(spa->spa_queue_stats[zio->io_priority].spa_queued, >, 0); spa->spa_queue_stats[zio->io_priority].spa_queued--; if (spa->spa_iokstat != NULL) kstat_waitq_exit(spa->spa_iokstat->ks_data); mutex_exit(&spa->spa_iokstat_lock); #endif } static void vdev_queue_pending_add(vdev_queue_t *vq, zio_t *zio) { spa_t *spa = zio->io_spa; ASSERT(MUTEX_HELD(&vq->vq_lock)); ASSERT3U(zio->io_priority, <, ZIO_PRIORITY_NUM_QUEUEABLE); vq->vq_class[zio->io_priority].vqc_active++; avl_add(&vq->vq_active_tree, zio); #ifdef illumos mutex_enter(&spa->spa_iokstat_lock); spa->spa_queue_stats[zio->io_priority].spa_active++; if (spa->spa_iokstat != NULL) kstat_runq_enter(spa->spa_iokstat->ks_data); mutex_exit(&spa->spa_iokstat_lock); #endif } static void vdev_queue_pending_remove(vdev_queue_t *vq, zio_t *zio) { spa_t *spa = zio->io_spa; ASSERT(MUTEX_HELD(&vq->vq_lock)); ASSERT3U(zio->io_priority, <, ZIO_PRIORITY_NUM_QUEUEABLE); vq->vq_class[zio->io_priority].vqc_active--; avl_remove(&vq->vq_active_tree, zio); #ifdef illumos mutex_enter(&spa->spa_iokstat_lock); ASSERT3U(spa->spa_queue_stats[zio->io_priority].spa_active, >, 0); spa->spa_queue_stats[zio->io_priority].spa_active--; if (spa->spa_iokstat != NULL) { kstat_io_t *ksio = spa->spa_iokstat->ks_data; kstat_runq_exit(spa->spa_iokstat->ks_data); if (zio->io_type == ZIO_TYPE_READ) { ksio->reads++; ksio->nread += zio->io_size; } else if (zio->io_type == ZIO_TYPE_WRITE) { ksio->writes++; ksio->nwritten += zio->io_size; } } mutex_exit(&spa->spa_iokstat_lock); #endif } static void vdev_queue_agg_io_done(zio_t *aio) { if (aio->io_type == ZIO_TYPE_READ) { zio_t *pio; while ((pio = zio_walk_parents(aio)) != NULL) { bcopy((char *)aio->io_data + (pio->io_offset - aio->io_offset), pio->io_data, pio->io_size); } } zio_buf_free(aio->io_data, aio->io_size); } static int vdev_queue_class_min_active(zio_priority_t p) { switch (p) { case ZIO_PRIORITY_SYNC_READ: return (zfs_vdev_sync_read_min_active); case ZIO_PRIORITY_SYNC_WRITE: return (zfs_vdev_sync_write_min_active); case ZIO_PRIORITY_ASYNC_READ: return (zfs_vdev_async_read_min_active); case ZIO_PRIORITY_ASYNC_WRITE: return (zfs_vdev_async_write_min_active); case ZIO_PRIORITY_SCRUB: return (zfs_vdev_scrub_min_active); case ZIO_PRIORITY_TRIM: return (zfs_vdev_trim_min_active); default: panic("invalid priority %u", p); return (0); } } static int vdev_queue_max_async_writes(spa_t *spa) { int writes; uint64_t dirty = spa->spa_dsl_pool->dp_dirty_total; uint64_t min_bytes = zfs_dirty_data_max * zfs_vdev_async_write_active_min_dirty_percent / 100; uint64_t max_bytes = zfs_dirty_data_max * zfs_vdev_async_write_active_max_dirty_percent / 100; /* * Sync tasks correspond to interactive user actions. To reduce the * execution time of those actions we push data out as fast as possible. */ if (spa_has_pending_synctask(spa)) { return (zfs_vdev_async_write_max_active); } if (dirty < min_bytes) return (zfs_vdev_async_write_min_active); if (dirty > max_bytes) return (zfs_vdev_async_write_max_active); /* * linear interpolation: * slope = (max_writes - min_writes) / (max_bytes - min_bytes) * move right by min_bytes * move up by min_writes */ writes = (dirty - min_bytes) * (zfs_vdev_async_write_max_active - zfs_vdev_async_write_min_active) / (max_bytes - min_bytes) + zfs_vdev_async_write_min_active; ASSERT3U(writes, >=, zfs_vdev_async_write_min_active); ASSERT3U(writes, <=, zfs_vdev_async_write_max_active); return (writes); } static int vdev_queue_class_max_active(spa_t *spa, zio_priority_t p) { switch (p) { case ZIO_PRIORITY_SYNC_READ: return (zfs_vdev_sync_read_max_active); case ZIO_PRIORITY_SYNC_WRITE: return (zfs_vdev_sync_write_max_active); case ZIO_PRIORITY_ASYNC_READ: return (zfs_vdev_async_read_max_active); case ZIO_PRIORITY_ASYNC_WRITE: return (vdev_queue_max_async_writes(spa)); case ZIO_PRIORITY_SCRUB: return (zfs_vdev_scrub_max_active); case ZIO_PRIORITY_TRIM: return (zfs_vdev_trim_max_active); default: panic("invalid priority %u", p); return (0); } } /* * Return the i/o class to issue from, or ZIO_PRIORITY_MAX_QUEUEABLE if * there is no eligible class. */ static zio_priority_t vdev_queue_class_to_issue(vdev_queue_t *vq) { spa_t *spa = vq->vq_vdev->vdev_spa; zio_priority_t p; ASSERT(MUTEX_HELD(&vq->vq_lock)); if (avl_numnodes(&vq->vq_active_tree) >= zfs_vdev_max_active) return (ZIO_PRIORITY_NUM_QUEUEABLE); /* find a queue that has not reached its minimum # outstanding i/os */ for (p = 0; p < ZIO_PRIORITY_NUM_QUEUEABLE; p++) { if (avl_numnodes(&vq->vq_class[p].vqc_queued_tree) > 0 && vq->vq_class[p].vqc_active < vdev_queue_class_min_active(p)) return (p); } /* * If we haven't found a queue, look for one that hasn't reached its * maximum # outstanding i/os. */ for (p = 0; p < ZIO_PRIORITY_NUM_QUEUEABLE; p++) { if (avl_numnodes(&vq->vq_class[p].vqc_queued_tree) > 0 && vq->vq_class[p].vqc_active < vdev_queue_class_max_active(spa, p)) return (p); } /* No eligible queued i/os */ return (ZIO_PRIORITY_NUM_QUEUEABLE); } /* * Compute the range spanned by two i/os, which is the endpoint of the last * (lio->io_offset + lio->io_size) minus start of the first (fio->io_offset). * Conveniently, the gap between fio and lio is given by -IO_SPAN(lio, fio); * thus fio and lio are adjacent if and only if IO_SPAN(lio, fio) == 0. */ #define IO_SPAN(fio, lio) ((lio)->io_offset + (lio)->io_size - (fio)->io_offset) #define IO_GAP(fio, lio) (-IO_SPAN(lio, fio)) static zio_t * vdev_queue_aggregate(vdev_queue_t *vq, zio_t *zio) { zio_t *first, *last, *aio, *dio, *mandatory, *nio; uint64_t maxgap = 0; uint64_t size; boolean_t stretch; avl_tree_t *t; enum zio_flag flags; ASSERT(MUTEX_HELD(&vq->vq_lock)); if (zio->io_flags & ZIO_FLAG_DONT_AGGREGATE) return (NULL); /* * The synchronous i/o queues are not sorted by LBA, so we can't * find adjacent i/os. These i/os tend to not be tightly clustered, * or too large to aggregate, so this has little impact on performance. */ if (zio->io_priority == ZIO_PRIORITY_SYNC_READ || zio->io_priority == ZIO_PRIORITY_SYNC_WRITE) return (NULL); first = last = zio; if (zio->io_type == ZIO_TYPE_READ) maxgap = zfs_vdev_read_gap_limit; /* * We can aggregate I/Os that are sufficiently adjacent and of * the same flavor, as expressed by the AGG_INHERIT flags. * The latter requirement is necessary so that certain * attributes of the I/O, such as whether it's a normal I/O * or a scrub/resilver, can be preserved in the aggregate. * We can include optional I/Os, but don't allow them * to begin a range as they add no benefit in that situation. */ /* * We keep track of the last non-optional I/O. */ mandatory = (first->io_flags & ZIO_FLAG_OPTIONAL) ? NULL : first; /* * Walk backwards through sufficiently contiguous I/Os * recording the last non-option I/O. */ flags = zio->io_flags & ZIO_FLAG_AGG_INHERIT; t = &vq->vq_class[zio->io_priority].vqc_queued_tree; while ((dio = AVL_PREV(t, first)) != NULL && (dio->io_flags & ZIO_FLAG_AGG_INHERIT) == flags && IO_SPAN(dio, last) <= zfs_vdev_aggregation_limit && IO_GAP(dio, first) <= maxgap) { first = dio; if (mandatory == NULL && !(first->io_flags & ZIO_FLAG_OPTIONAL)) mandatory = first; } /* * Skip any initial optional I/Os. */ while ((first->io_flags & ZIO_FLAG_OPTIONAL) && first != last) { first = AVL_NEXT(t, first); ASSERT(first != NULL); } /* * Walk forward through sufficiently contiguous I/Os. */ while ((dio = AVL_NEXT(t, last)) != NULL && (dio->io_flags & ZIO_FLAG_AGG_INHERIT) == flags && IO_SPAN(first, dio) <= zfs_vdev_aggregation_limit && IO_GAP(last, dio) <= maxgap) { last = dio; if (!(last->io_flags & ZIO_FLAG_OPTIONAL)) mandatory = last; } /* * Now that we've established the range of the I/O aggregation * we must decide what to do with trailing optional I/Os. * For reads, there's nothing to do. While we are unable to * aggregate further, it's possible that a trailing optional * I/O would allow the underlying device to aggregate with * subsequent I/Os. We must therefore determine if the next * non-optional I/O is close enough to make aggregation * worthwhile. */ stretch = B_FALSE; if (zio->io_type == ZIO_TYPE_WRITE && mandatory != NULL) { zio_t *nio = last; while ((dio = AVL_NEXT(t, nio)) != NULL && IO_GAP(nio, dio) == 0 && IO_GAP(mandatory, dio) <= zfs_vdev_write_gap_limit) { nio = dio; if (!(nio->io_flags & ZIO_FLAG_OPTIONAL)) { stretch = B_TRUE; break; } } } if (stretch) { /* This may be a no-op. */ dio = AVL_NEXT(t, last); dio->io_flags &= ~ZIO_FLAG_OPTIONAL; } else { while (last != mandatory && last != first) { ASSERT(last->io_flags & ZIO_FLAG_OPTIONAL); last = AVL_PREV(t, last); ASSERT(last != NULL); } } if (first == last) return (NULL); size = IO_SPAN(first, last); ASSERT3U(size, <=, zfs_vdev_aggregation_limit); aio = zio_vdev_delegated_io(first->io_vd, first->io_offset, zio_buf_alloc(size), size, first->io_type, zio->io_priority, flags | ZIO_FLAG_DONT_CACHE | ZIO_FLAG_DONT_QUEUE, vdev_queue_agg_io_done, NULL); aio->io_timestamp = first->io_timestamp; nio = first; do { dio = nio; nio = AVL_NEXT(t, dio); ASSERT3U(dio->io_type, ==, aio->io_type); if (dio->io_flags & ZIO_FLAG_NODATA) { ASSERT3U(dio->io_type, ==, ZIO_TYPE_WRITE); bzero((char *)aio->io_data + (dio->io_offset - aio->io_offset), dio->io_size); } else if (dio->io_type == ZIO_TYPE_WRITE) { bcopy(dio->io_data, (char *)aio->io_data + (dio->io_offset - aio->io_offset), dio->io_size); } zio_add_child(dio, aio); vdev_queue_io_remove(vq, dio); zio_vdev_io_bypass(dio); zio_execute(dio); } while (dio != last); return (aio); } static zio_t * vdev_queue_io_to_issue(vdev_queue_t *vq) { zio_t *zio, *aio; zio_priority_t p; avl_index_t idx; vdev_queue_class_t *vqc; zio_t search; again: ASSERT(MUTEX_HELD(&vq->vq_lock)); p = vdev_queue_class_to_issue(vq); if (p == ZIO_PRIORITY_NUM_QUEUEABLE) { /* No eligible queued i/os */ return (NULL); } /* * For LBA-ordered queues (async / scrub), issue the i/o which follows * the most recently issued i/o in LBA (offset) order. * * For FIFO queues (sync), issue the i/o with the lowest timestamp. */ vqc = &vq->vq_class[p]; search.io_timestamp = 0; search.io_offset = vq->vq_last_offset + 1; VERIFY3P(avl_find(&vqc->vqc_queued_tree, &search, &idx), ==, NULL); zio = avl_nearest(&vqc->vqc_queued_tree, idx, AVL_AFTER); if (zio == NULL) zio = avl_first(&vqc->vqc_queued_tree); ASSERT3U(zio->io_priority, ==, p); aio = vdev_queue_aggregate(vq, zio); if (aio != NULL) zio = aio; else vdev_queue_io_remove(vq, zio); /* * If the I/O is or was optional and therefore has no data, we need to * simply discard it. We need to drop the vdev queue's lock to avoid a * deadlock that we could encounter since this I/O will complete * immediately. */ if (zio->io_flags & ZIO_FLAG_NODATA) { mutex_exit(&vq->vq_lock); zio_vdev_io_bypass(zio); zio_execute(zio); mutex_enter(&vq->vq_lock); goto again; } vdev_queue_pending_add(vq, zio); vq->vq_last_offset = zio->io_offset; return (zio); } zio_t * vdev_queue_io(zio_t *zio) { vdev_queue_t *vq = &zio->io_vd->vdev_queue; zio_t *nio; if (zio->io_flags & ZIO_FLAG_DONT_QUEUE) return (zio); /* * Children i/os inherent their parent's priority, which might * not match the child's i/o type. Fix it up here. */ if (zio->io_type == ZIO_TYPE_READ) { if (zio->io_priority != ZIO_PRIORITY_SYNC_READ && zio->io_priority != ZIO_PRIORITY_ASYNC_READ && zio->io_priority != ZIO_PRIORITY_SCRUB) zio->io_priority = ZIO_PRIORITY_ASYNC_READ; } else if (zio->io_type == ZIO_TYPE_WRITE) { if (zio->io_priority != ZIO_PRIORITY_SYNC_WRITE && zio->io_priority != ZIO_PRIORITY_ASYNC_WRITE) zio->io_priority = ZIO_PRIORITY_ASYNC_WRITE; } else { ASSERT(zio->io_type == ZIO_TYPE_FREE); zio->io_priority = ZIO_PRIORITY_TRIM; } zio->io_flags |= ZIO_FLAG_DONT_CACHE | ZIO_FLAG_DONT_QUEUE; mutex_enter(&vq->vq_lock); zio->io_timestamp = gethrtime(); vdev_queue_io_add(vq, zio); nio = vdev_queue_io_to_issue(vq); mutex_exit(&vq->vq_lock); if (nio == NULL) return (NULL); if (nio->io_done == vdev_queue_agg_io_done) { zio_nowait(nio); return (NULL); } return (nio); } void vdev_queue_io_done(zio_t *zio) { vdev_queue_t *vq = &zio->io_vd->vdev_queue; zio_t *nio; if (zio_injection_enabled) delay(SEC_TO_TICK(zio_handle_io_delay(zio))); mutex_enter(&vq->vq_lock); vdev_queue_pending_remove(vq, zio); vq->vq_io_complete_ts = gethrtime(); while ((nio = vdev_queue_io_to_issue(vq)) != NULL) { mutex_exit(&vq->vq_lock); if (nio->io_done == vdev_queue_agg_io_done) { zio_nowait(nio); } else { zio_vdev_io_reissue(nio); zio_execute(nio); } mutex_enter(&vq->vq_lock); } mutex_exit(&vq->vq_lock); +} + +/* + * As these three methods are only used for load calculations we're not concerned + * if we get an incorrect value on 32bit platforms due to lack of vq_lock mutex + * use here, instead we prefer to keep it lock free for performance. + */ +int +vdev_queue_length(vdev_t *vd) +{ + return (avl_numnodes(&vd->vdev_queue.vq_active_tree)); +} + +uint64_t +vdev_queue_lastoffset(vdev_t *vd) +{ + return (vd->vdev_queue.vq_lastoffset); +} + +void +vdev_queue_register_lastoffset(vdev_t *vd, zio_t *zio) +{ + vd->vdev_queue.vq_lastoffset = zio->io_offset + zio->io_size; } Index: stable/10/sys/geom/geom.h =================================================================== --- stable/10/sys/geom/geom.h (revision 271237) +++ stable/10/sys/geom/geom.h (revision 271238) @@ -1,415 +1,416 @@ /*- * Copyright (c) 2002 Poul-Henning Kamp * Copyright (c) 2002 Networks Associates Technology, Inc. * All rights reserved. * * This software was developed for the FreeBSD Project by Poul-Henning Kamp * and NAI Labs, the Security Research Division of Network Associates, Inc. * under DARPA/SPAWAR contract N66001-01-C-8035 ("CBOSS"), as part of the * DARPA CHATS research program. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. The names of the authors may not be used to endorse or promote * products derived from this software without specific prior written * permission. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * $FreeBSD$ */ #ifndef _GEOM_GEOM_H_ #define _GEOM_GEOM_H_ #include #include #include #include #include #include #include struct g_class; struct g_geom; struct g_consumer; struct g_provider; struct g_stat; struct thread; struct bio; struct sbuf; struct gctl_req; struct g_configargs; typedef int g_config_t (struct g_configargs *ca); typedef void g_ctl_req_t (struct gctl_req *, struct g_class *cp, char const *verb); typedef int g_ctl_create_geom_t (struct gctl_req *, struct g_class *cp, struct g_provider *pp); typedef int g_ctl_destroy_geom_t (struct gctl_req *, struct g_class *cp, struct g_geom *gp); typedef int g_ctl_config_geom_t (struct gctl_req *, struct g_geom *gp, const char *verb); typedef void g_init_t (struct g_class *mp); typedef void g_fini_t (struct g_class *mp); typedef struct g_geom * g_taste_t (struct g_class *, struct g_provider *, int flags); typedef int g_ioctl_t(struct g_provider *pp, u_long cmd, void *data, int fflag, struct thread *td); #define G_TF_NORMAL 0 #define G_TF_INSIST 1 #define G_TF_TRANSPARENT 2 typedef int g_access_t (struct g_provider *, int, int, int); /* XXX: not sure about the thread arg */ typedef void g_orphan_t (struct g_consumer *); typedef void g_start_t (struct bio *); typedef void g_spoiled_t (struct g_consumer *); typedef void g_attrchanged_t (struct g_consumer *, const char *attr); typedef void g_provgone_t (struct g_provider *); typedef void g_dumpconf_t (struct sbuf *, const char *indent, struct g_geom *, struct g_consumer *, struct g_provider *); typedef void g_resize_t(struct g_consumer *cp); /* * The g_class structure describes a transformation class. In other words * all BSD disklabel handlers share one g_class, all MBR handlers share * one common g_class and so on. * Certain operations are instantiated on the class, most notably the * taste and config_geom functions. */ struct g_class { const char *name; u_int version; u_int spare0; g_taste_t *taste; g_config_t *config; g_ctl_req_t *ctlreq; g_init_t *init; g_fini_t *fini; g_ctl_destroy_geom_t *destroy_geom; /* * Default values for geom methods */ g_start_t *start; g_spoiled_t *spoiled; g_attrchanged_t *attrchanged; g_dumpconf_t *dumpconf; g_access_t *access; g_orphan_t *orphan; g_ioctl_t *ioctl; g_provgone_t *providergone; g_resize_t *resize; void *spare1; void *spare2; /* * The remaining elements are private */ LIST_ENTRY(g_class) class; LIST_HEAD(,g_geom) geom; }; #define G_VERSION_00 0x19950323 #define G_VERSION_01 0x20041207 /* add fflag to g_ioctl_t */ #define G_VERSION G_VERSION_01 /* * The g_geom is an instance of a g_class. */ struct g_geom { char *name; struct g_class *class; LIST_ENTRY(g_geom) geom; LIST_HEAD(,g_consumer) consumer; LIST_HEAD(,g_provider) provider; TAILQ_ENTRY(g_geom) geoms; /* XXX: better name */ int rank; g_start_t *start; g_spoiled_t *spoiled; g_attrchanged_t *attrchanged; g_dumpconf_t *dumpconf; g_access_t *access; g_orphan_t *orphan; g_ioctl_t *ioctl; g_provgone_t *providergone; g_resize_t *resize; void *spare0; void *spare1; void *softc; unsigned flags; #define G_GEOM_WITHER 1 #define G_GEOM_VOLATILE_BIO 2 }; /* * The g_bioq is a queue of struct bio's. * XXX: possibly collection point for statistics. * XXX: should (possibly) be collapsed with sys/bio.h::bio_queue_head. */ struct g_bioq { TAILQ_HEAD(, bio) bio_queue; struct mtx bio_queue_lock; int bio_queue_length; }; /* * A g_consumer is an attachment point for a g_provider. One g_consumer * can only be attached to one g_provider, but multiple g_consumers * can be attached to one g_provider. */ struct g_consumer { struct g_geom *geom; LIST_ENTRY(g_consumer) consumer; struct g_provider *provider; LIST_ENTRY(g_consumer) consumers; /* XXX: better name */ int acr, acw, ace; int flags; #define G_CF_SPOILED 0x1 #define G_CF_ORPHAN 0x4 #define G_CF_DIRECT_SEND 0x10 #define G_CF_DIRECT_RECEIVE 0x20 struct devstat *stat; u_int nstart, nend; /* Two fields for the implementing class to use */ void *private; u_int index; }; /* * A g_provider is a "logical disk". */ struct g_provider { char *name; LIST_ENTRY(g_provider) provider; struct g_geom *geom; LIST_HEAD(,g_consumer) consumers; int acr, acw, ace; int error; TAILQ_ENTRY(g_provider) orphan; off_t mediasize; u_int sectorsize; u_int stripesize; u_int stripeoffset; struct devstat *stat; u_int nstart, nend; u_int flags; #define G_PF_WITHER 0x2 #define G_PF_ORPHAN 0x4 #define G_PF_ACCEPT_UNMAPPED 0x8 #define G_PF_DIRECT_SEND 0x10 #define G_PF_DIRECT_RECEIVE 0x20 /* Two fields for the implementing class to use */ void *private; u_int index; }; /* * Descriptor of a classifier. We can register a function and * an argument, which is called by g_io_request() on bio's * that are not previously classified. */ struct g_classifier_hook { TAILQ_ENTRY(g_classifier_hook) link; int (*func)(void *arg, struct bio *bp); void *arg; }; /* BIO_GETATTR("GEOM::setstate") argument values. */ #define G_STATE_FAILED 0 #define G_STATE_REBUILD 1 #define G_STATE_RESYNC 2 #define G_STATE_ACTIVE 3 /* geom_dev.c */ struct cdev; void g_dev_print(void); void g_dev_physpath_changed(void); struct g_provider *g_dev_getprovider(struct cdev *dev); /* geom_dump.c */ void g_trace(int level, const char *, ...); # define G_T_TOPOLOGY 1 # define G_T_BIO 2 # define G_T_ACCESS 4 /* geom_event.c */ typedef void g_event_t(void *, int flag); #define EV_CANCEL 1 int g_post_event(g_event_t *func, void *arg, int flag, ...); int g_waitfor_event(g_event_t *func, void *arg, int flag, ...); void g_cancel_event(void *ref); int g_attr_changed(struct g_provider *pp, const char *attr, int flag); int g_media_changed(struct g_provider *pp, int flag); int g_media_gone(struct g_provider *pp, int flag); void g_orphan_provider(struct g_provider *pp, int error); void g_waitidlelock(void); /* geom_subr.c */ int g_access(struct g_consumer *cp, int nread, int nwrite, int nexcl); int g_attach(struct g_consumer *cp, struct g_provider *pp); int g_compare_names(const char *namea, const char *nameb); void g_destroy_consumer(struct g_consumer *cp); void g_destroy_geom(struct g_geom *pp); void g_destroy_provider(struct g_provider *pp); void g_detach(struct g_consumer *cp); void g_error_provider(struct g_provider *pp, int error); struct g_provider *g_provider_by_name(char const *arg); int g_getattr__(const char *attr, struct g_consumer *cp, void *var, int len); #define g_getattr(a, c, v) g_getattr__((a), (c), (v), sizeof *(v)) int g_handleattr(struct bio *bp, const char *attribute, const void *val, int len); int g_handleattr_int(struct bio *bp, const char *attribute, int val); int g_handleattr_off_t(struct bio *bp, const char *attribute, off_t val); +int g_handleattr_uint16_t(struct bio *bp, const char *attribute, uint16_t val); int g_handleattr_str(struct bio *bp, const char *attribute, const char *str); struct g_consumer * g_new_consumer(struct g_geom *gp); struct g_geom * g_new_geomf(struct g_class *mp, const char *fmt, ...) __printflike(2, 3); struct g_provider * g_new_providerf(struct g_geom *gp, const char *fmt, ...) __printflike(2, 3); void g_resize_provider(struct g_provider *pp, off_t size); int g_retaste(struct g_class *mp); void g_spoil(struct g_provider *pp, struct g_consumer *cp); int g_std_access(struct g_provider *pp, int dr, int dw, int de); void g_std_done(struct bio *bp); void g_std_spoiled(struct g_consumer *cp); void g_wither_geom(struct g_geom *gp, int error); void g_wither_geom_close(struct g_geom *gp, int error); void g_wither_provider(struct g_provider *pp, int error); #if defined(DIAGNOSTIC) || defined(DDB) int g_valid_obj(void const *ptr); #endif #ifdef DIAGNOSTIC #define G_VALID_CLASS(foo) \ KASSERT(g_valid_obj(foo) == 1, ("%p is not a g_class", foo)) #define G_VALID_GEOM(foo) \ KASSERT(g_valid_obj(foo) == 2, ("%p is not a g_geom", foo)) #define G_VALID_CONSUMER(foo) \ KASSERT(g_valid_obj(foo) == 3, ("%p is not a g_consumer", foo)) #define G_VALID_PROVIDER(foo) \ KASSERT(g_valid_obj(foo) == 4, ("%p is not a g_provider", foo)) #else #define G_VALID_CLASS(foo) do { } while (0) #define G_VALID_GEOM(foo) do { } while (0) #define G_VALID_CONSUMER(foo) do { } while (0) #define G_VALID_PROVIDER(foo) do { } while (0) #endif int g_modevent(module_t, int, void *); /* geom_io.c */ struct bio * g_clone_bio(struct bio *); struct bio * g_duplicate_bio(struct bio *); void g_destroy_bio(struct bio *); void g_io_deliver(struct bio *bp, int error); int g_io_getattr(const char *attr, struct g_consumer *cp, int *len, void *ptr); int g_io_flush(struct g_consumer *cp); int g_register_classifier(struct g_classifier_hook *hook); void g_unregister_classifier(struct g_classifier_hook *hook); void g_io_request(struct bio *bp, struct g_consumer *cp); struct bio *g_new_bio(void); struct bio *g_alloc_bio(void); void * g_read_data(struct g_consumer *cp, off_t offset, off_t length, int *error); int g_write_data(struct g_consumer *cp, off_t offset, void *ptr, off_t length); int g_delete_data(struct g_consumer *cp, off_t offset, off_t length); void g_print_bio(struct bio *bp); /* geom_kern.c / geom_kernsim.c */ #ifdef _KERNEL extern struct sx topology_lock; struct g_kerneldump { off_t offset; off_t length; struct dumperinfo di; }; MALLOC_DECLARE(M_GEOM); static __inline void * g_malloc(int size, int flags) { void *p; p = malloc(size, M_GEOM, flags); return (p); } static __inline void g_free(void *ptr) { #ifdef DIAGNOSTIC if (sx_xlocked(&topology_lock)) { KASSERT(g_valid_obj(ptr) == 0, ("g_free(%p) of live object, type %d", ptr, g_valid_obj(ptr))); } #endif free(ptr, M_GEOM); } #define g_topology_lock() \ do { \ mtx_assert(&Giant, MA_NOTOWNED); \ sx_xlock(&topology_lock); \ } while (0) #define g_topology_try_lock() sx_try_xlock(&topology_lock) #define g_topology_unlock() \ do { \ sx_xunlock(&topology_lock); \ } while (0) #define g_topology_assert() \ do { \ sx_assert(&topology_lock, SX_XLOCKED); \ } while (0) #define g_topology_assert_not() \ do { \ sx_assert(&topology_lock, SX_UNLOCKED); \ } while (0) #define g_topology_sleep(chan, timo) \ sx_sleep(chan, &topology_lock, 0, "gtopol", timo) #define DECLARE_GEOM_CLASS(class, name) \ static moduledata_t name##_mod = { \ #name, g_modevent, &class \ }; \ DECLARE_MODULE(name, name##_mod, SI_SUB_DRIVERS, SI_ORDER_FIRST); int g_is_geom_thread(struct thread *td); #endif /* _KERNEL */ /* geom_ctl.c */ int gctl_set_param(struct gctl_req *req, const char *param, void const *ptr, int len); void gctl_set_param_err(struct gctl_req *req, const char *param, void const *ptr, int len); void *gctl_get_param(struct gctl_req *req, const char *param, int *len); char const *gctl_get_asciiparam(struct gctl_req *req, const char *param); void *gctl_get_paraml(struct gctl_req *req, const char *param, int len); int gctl_error(struct gctl_req *req, const char *fmt, ...) __printflike(2, 3); struct g_class *gctl_get_class(struct gctl_req *req, char const *arg); struct g_geom *gctl_get_geom(struct gctl_req *req, struct g_class *mpr, char const *arg); struct g_provider *gctl_get_provider(struct gctl_req *req, char const *arg); #endif /* _GEOM_GEOM_H_ */ Index: stable/10/sys/geom/geom_disk.c =================================================================== --- stable/10/sys/geom/geom_disk.c (revision 271237) +++ stable/10/sys/geom/geom_disk.c (revision 271238) @@ -1,826 +1,837 @@ /*- * Copyright (c) 2002 Poul-Henning Kamp * Copyright (c) 2002 Networks Associates Technology, Inc. * All rights reserved. * * This software was developed for the FreeBSD Project by Poul-Henning Kamp * and NAI Labs, the Security Research Division of Network Associates, Inc. * under DARPA/SPAWAR contract N66001-01-C-8035 ("CBOSS"), as part of the * DARPA CHATS research program. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. The names of the authors may not be used to endorse or promote * products derived from this software without specific prior written * permission. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include "opt_geom.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include struct g_disk_softc { struct mtx done_mtx; struct disk *dp; struct sysctl_ctx_list sysctl_ctx; struct sysctl_oid *sysctl_tree; char led[64]; uint32_t state; struct mtx start_mtx; }; static g_access_t g_disk_access; static g_start_t g_disk_start; static g_ioctl_t g_disk_ioctl; static g_dumpconf_t g_disk_dumpconf; static g_provgone_t g_disk_providergone; static struct g_class g_disk_class = { .name = G_DISK_CLASS_NAME, .version = G_VERSION, .start = g_disk_start, .access = g_disk_access, .ioctl = g_disk_ioctl, .providergone = g_disk_providergone, .dumpconf = g_disk_dumpconf, }; SYSCTL_DECL(_kern_geom); static SYSCTL_NODE(_kern_geom, OID_AUTO, disk, CTLFLAG_RW, 0, "GEOM_DISK stuff"); DECLARE_GEOM_CLASS(g_disk_class, g_disk); static void __inline g_disk_lock_giant(struct disk *dp) { if (dp->d_flags & DISKFLAG_NEEDSGIANT) mtx_lock(&Giant); } static void __inline g_disk_unlock_giant(struct disk *dp) { if (dp->d_flags & DISKFLAG_NEEDSGIANT) mtx_unlock(&Giant); } static int g_disk_access(struct g_provider *pp, int r, int w, int e) { struct disk *dp; struct g_disk_softc *sc; int error; g_trace(G_T_ACCESS, "g_disk_access(%s, %d, %d, %d)", pp->name, r, w, e); g_topology_assert(); sc = pp->private; if (sc == NULL || (dp = sc->dp) == NULL || dp->d_destroyed) { /* * Allow decreasing access count even if disk is not * avaliable anymore. */ if (r <= 0 && w <= 0 && e <= 0) return (0); return (ENXIO); } r += pp->acr; w += pp->acw; e += pp->ace; error = 0; if ((pp->acr + pp->acw + pp->ace) == 0 && (r + w + e) > 0) { if (dp->d_open != NULL) { g_disk_lock_giant(dp); error = dp->d_open(dp); if (bootverbose && error != 0) printf("Opened disk %s -> %d\n", pp->name, error); g_disk_unlock_giant(dp); if (error != 0) return (error); } pp->mediasize = dp->d_mediasize; pp->sectorsize = dp->d_sectorsize; if (dp->d_maxsize == 0) { printf("WARNING: Disk drive %s%d has no d_maxsize\n", dp->d_name, dp->d_unit); dp->d_maxsize = DFLTPHYS; } if (dp->d_delmaxsize == 0) { if (bootverbose && dp->d_flags & DISKFLAG_CANDELETE) { printf("WARNING: Disk drive %s%d has no " "d_delmaxsize\n", dp->d_name, dp->d_unit); } dp->d_delmaxsize = dp->d_maxsize; } pp->stripeoffset = dp->d_stripeoffset; pp->stripesize = dp->d_stripesize; dp->d_flags |= DISKFLAG_OPEN; } else if ((pp->acr + pp->acw + pp->ace) > 0 && (r + w + e) == 0) { if (dp->d_close != NULL) { g_disk_lock_giant(dp); error = dp->d_close(dp); if (error != 0) printf("Closed disk %s -> %d\n", pp->name, error); g_disk_unlock_giant(dp); } sc->state = G_STATE_ACTIVE; if (sc->led[0] != 0) led_set(sc->led, "0"); dp->d_flags &= ~DISKFLAG_OPEN; } return (error); } static void g_disk_kerneldump(struct bio *bp, struct disk *dp) { struct g_kerneldump *gkd; struct g_geom *gp; gkd = (struct g_kerneldump*)bp->bio_data; gp = bp->bio_to->geom; g_trace(G_T_TOPOLOGY, "g_disk_kerneldump(%s, %jd, %jd)", gp->name, (intmax_t)gkd->offset, (intmax_t)gkd->length); if (dp->d_dump == NULL) { g_io_deliver(bp, ENODEV); return; } gkd->di.dumper = dp->d_dump; gkd->di.priv = dp; gkd->di.blocksize = dp->d_sectorsize; gkd->di.maxiosize = dp->d_maxsize; gkd->di.mediaoffset = gkd->offset; if ((gkd->offset + gkd->length) > dp->d_mediasize) gkd->length = dp->d_mediasize - gkd->offset; gkd->di.mediasize = gkd->length; g_io_deliver(bp, 0); } static void g_disk_setstate(struct bio *bp, struct g_disk_softc *sc) { const char *cmd; memcpy(&sc->state, bp->bio_data, sizeof(sc->state)); if (sc->led[0] != 0) { switch (sc->state) { case G_STATE_FAILED: cmd = "1"; break; case G_STATE_REBUILD: cmd = "f5"; break; case G_STATE_RESYNC: cmd = "f1"; break; default: cmd = "0"; break; } led_set(sc->led, cmd); } g_io_deliver(bp, 0); } static void g_disk_done(struct bio *bp) { struct bintime now; struct bio *bp2; struct g_disk_softc *sc; /* See "notes" for why we need a mutex here */ /* XXX: will witness accept a mix of Giant/unGiant drivers here ? */ bp2 = bp->bio_parent; sc = bp2->bio_to->private; bp->bio_completed = bp->bio_length - bp->bio_resid; binuptime(&now); mtx_lock(&sc->done_mtx); if (bp2->bio_error == 0) bp2->bio_error = bp->bio_error; bp2->bio_completed += bp->bio_completed; if ((bp->bio_cmd & (BIO_READ|BIO_WRITE|BIO_DELETE|BIO_FLUSH)) != 0) devstat_end_transaction_bio_bt(sc->dp->d_devstat, bp, &now); bp2->bio_inbed++; if (bp2->bio_children == bp2->bio_inbed) { mtx_unlock(&sc->done_mtx); bp2->bio_resid = bp2->bio_bcount - bp2->bio_completed; g_io_deliver(bp2, bp2->bio_error); } else mtx_unlock(&sc->done_mtx); g_destroy_bio(bp); } static void g_disk_done_single(struct bio *bp) { struct bintime now; struct g_disk_softc *sc; bp->bio_completed = bp->bio_length - bp->bio_resid; bp->bio_done = (void *)bp->bio_to; bp->bio_to = LIST_FIRST(&bp->bio_disk->d_geom->provider); if ((bp->bio_cmd & (BIO_READ|BIO_WRITE|BIO_DELETE|BIO_FLUSH)) != 0) { binuptime(&now); sc = bp->bio_to->private; mtx_lock(&sc->done_mtx); devstat_end_transaction_bio_bt(sc->dp->d_devstat, bp, &now); mtx_unlock(&sc->done_mtx); } g_io_deliver(bp, bp->bio_error); } static int g_disk_ioctl(struct g_provider *pp, u_long cmd, void * data, int fflag, struct thread *td) { struct disk *dp; struct g_disk_softc *sc; int error; sc = pp->private; dp = sc->dp; if (dp->d_ioctl == NULL) return (ENOIOCTL); g_disk_lock_giant(dp); error = dp->d_ioctl(dp, cmd, data, fflag, td); g_disk_unlock_giant(dp); return (error); } static void g_disk_start(struct bio *bp) { struct bio *bp2, *bp3; struct disk *dp; struct g_disk_softc *sc; int error; off_t d_maxsize, off; sc = bp->bio_to->private; if (sc == NULL || (dp = sc->dp) == NULL || dp->d_destroyed) { g_io_deliver(bp, ENXIO); return; } error = EJUSTRETURN; switch(bp->bio_cmd) { case BIO_DELETE: if (!(dp->d_flags & DISKFLAG_CANDELETE)) { error = EOPNOTSUPP; break; } /* fall-through */ case BIO_READ: case BIO_WRITE: d_maxsize = (bp->bio_cmd == BIO_DELETE) ? dp->d_delmaxsize : dp->d_maxsize; if (bp->bio_length <= d_maxsize) { bp->bio_disk = dp; bp->bio_to = (void *)bp->bio_done; bp->bio_done = g_disk_done_single; bp->bio_pblkno = bp->bio_offset / dp->d_sectorsize; bp->bio_bcount = bp->bio_length; mtx_lock(&sc->start_mtx); devstat_start_transaction_bio(dp->d_devstat, bp); mtx_unlock(&sc->start_mtx); g_disk_lock_giant(dp); dp->d_strategy(bp); g_disk_unlock_giant(dp); break; } off = 0; bp3 = NULL; bp2 = g_clone_bio(bp); if (bp2 == NULL) { error = ENOMEM; break; } do { bp2->bio_offset += off; bp2->bio_length -= off; if ((bp->bio_flags & BIO_UNMAPPED) == 0) { bp2->bio_data += off; } else { KASSERT((dp->d_flags & DISKFLAG_UNMAPPED_BIO) != 0, ("unmapped bio not supported by disk %s", dp->d_name)); bp2->bio_ma += off / PAGE_SIZE; bp2->bio_ma_offset += off; bp2->bio_ma_offset %= PAGE_SIZE; bp2->bio_ma_n -= off / PAGE_SIZE; } if (bp2->bio_length > d_maxsize) { /* * XXX: If we have a stripesize we should really * use it here. Care should be taken in the delete * case if this is done as deletes can be very * sensitive to size given how they are processed. */ bp2->bio_length = d_maxsize; if ((bp->bio_flags & BIO_UNMAPPED) != 0) { bp2->bio_ma_n = howmany( bp2->bio_ma_offset + bp2->bio_length, PAGE_SIZE); } off += d_maxsize; /* * To avoid a race, we need to grab the next bio * before we schedule this one. See "notes". */ bp3 = g_clone_bio(bp); if (bp3 == NULL) bp->bio_error = ENOMEM; } bp2->bio_done = g_disk_done; bp2->bio_pblkno = bp2->bio_offset / dp->d_sectorsize; bp2->bio_bcount = bp2->bio_length; bp2->bio_disk = dp; mtx_lock(&sc->start_mtx); devstat_start_transaction_bio(dp->d_devstat, bp2); mtx_unlock(&sc->start_mtx); g_disk_lock_giant(dp); dp->d_strategy(bp2); g_disk_unlock_giant(dp); bp2 = bp3; bp3 = NULL; } while (bp2 != NULL); break; case BIO_GETATTR: /* Give the driver a chance to override */ if (dp->d_getattr != NULL) { if (bp->bio_disk == NULL) bp->bio_disk = dp; error = dp->d_getattr(bp); if (error != -1) break; error = EJUSTRETURN; } if (g_handleattr_int(bp, "GEOM::candelete", (dp->d_flags & DISKFLAG_CANDELETE) != 0)) break; else if (g_handleattr_int(bp, "GEOM::fwsectors", dp->d_fwsectors)) break; else if (g_handleattr_int(bp, "GEOM::fwheads", dp->d_fwheads)) break; else if (g_handleattr_off_t(bp, "GEOM::frontstuff", 0)) break; else if (g_handleattr_str(bp, "GEOM::ident", dp->d_ident)) break; - else if (g_handleattr(bp, "GEOM::hba_vendor", - &dp->d_hba_vendor, 2)) + else if (g_handleattr_uint16_t(bp, "GEOM::hba_vendor", + dp->d_hba_vendor)) break; - else if (g_handleattr(bp, "GEOM::hba_device", - &dp->d_hba_device, 2)) + else if (g_handleattr_uint16_t(bp, "GEOM::hba_device", + dp->d_hba_device)) break; - else if (g_handleattr(bp, "GEOM::hba_subvendor", - &dp->d_hba_subvendor, 2)) + else if (g_handleattr_uint16_t(bp, "GEOM::hba_subvendor", + dp->d_hba_subvendor)) break; - else if (g_handleattr(bp, "GEOM::hba_subdevice", - &dp->d_hba_subdevice, 2)) + else if (g_handleattr_uint16_t(bp, "GEOM::hba_subdevice", + dp->d_hba_subdevice)) break; else if (!strcmp(bp->bio_attribute, "GEOM::kerneldump")) g_disk_kerneldump(bp, dp); else if (!strcmp(bp->bio_attribute, "GEOM::setstate")) g_disk_setstate(bp, sc); - else + else if (!strcmp(bp->bio_attribute, "GEOM::rotation_rate")) { + uint64_t v; + + if ((dp->d_flags & DISKFLAG_LACKS_ROTRATE) == 0) + v = dp->d_rotation_rate; + else + v = 0; /* rate unknown */ + g_handleattr_uint16_t(bp, "GEOM::rotation_rate", v); + break; + } else error = ENOIOCTL; break; case BIO_FLUSH: g_trace(G_T_BIO, "g_disk_flushcache(%s)", bp->bio_to->name); if (!(dp->d_flags & DISKFLAG_CANFLUSHCACHE)) { error = EOPNOTSUPP; break; } bp->bio_disk = dp; bp->bio_to = (void *)bp->bio_done; bp->bio_done = g_disk_done_single; mtx_lock(&sc->start_mtx); devstat_start_transaction_bio(dp->d_devstat, bp); mtx_unlock(&sc->start_mtx); g_disk_lock_giant(dp); dp->d_strategy(bp); g_disk_unlock_giant(dp); break; default: error = EOPNOTSUPP; break; } if (error != EJUSTRETURN) g_io_deliver(bp, error); return; } static void g_disk_dumpconf(struct sbuf *sb, const char *indent, struct g_geom *gp, struct g_consumer *cp, struct g_provider *pp) { struct bio *bp; struct disk *dp; struct g_disk_softc *sc; char *buf; int res = 0; sc = gp->softc; if (sc == NULL || (dp = sc->dp) == NULL) return; if (indent == NULL) { sbuf_printf(sb, " hd %u", dp->d_fwheads); sbuf_printf(sb, " sc %u", dp->d_fwsectors); return; } if (pp != NULL) { sbuf_printf(sb, "%s%u\n", indent, dp->d_fwheads); sbuf_printf(sb, "%s%u\n", indent, dp->d_fwsectors); if (dp->d_getattr != NULL) { buf = g_malloc(DISK_IDENT_SIZE, M_WAITOK); bp = g_alloc_bio(); bp->bio_disk = dp; bp->bio_attribute = "GEOM::ident"; bp->bio_length = DISK_IDENT_SIZE; bp->bio_data = buf; res = dp->d_getattr(bp); sbuf_printf(sb, "%s", indent); g_conf_printf_escaped(sb, "%s", res == 0 ? buf: dp->d_ident); sbuf_printf(sb, "\n"); bp->bio_attribute = "GEOM::lunid"; bp->bio_length = DISK_IDENT_SIZE; bp->bio_data = buf; if (dp->d_getattr(bp) == 0) { sbuf_printf(sb, "%s", indent); g_conf_printf_escaped(sb, "%s", buf); sbuf_printf(sb, "\n"); } bp->bio_attribute = "GEOM::lunname"; bp->bio_length = DISK_IDENT_SIZE; bp->bio_data = buf; if (dp->d_getattr(bp) == 0) { sbuf_printf(sb, "%s", indent); g_conf_printf_escaped(sb, "%s", buf); sbuf_printf(sb, "\n"); } g_destroy_bio(bp); g_free(buf); } else { sbuf_printf(sb, "%s", indent); g_conf_printf_escaped(sb, "%s", dp->d_ident); sbuf_printf(sb, "\n"); } sbuf_printf(sb, "%s", indent); g_conf_printf_escaped(sb, "%s", dp->d_descr); sbuf_printf(sb, "\n"); } } static void g_disk_resize(void *ptr, int flag) { struct disk *dp; struct g_geom *gp; struct g_provider *pp; if (flag == EV_CANCEL) return; g_topology_assert(); dp = ptr; gp = dp->d_geom; if (dp->d_destroyed || gp == NULL) return; LIST_FOREACH(pp, &gp->provider, provider) { if (pp->sectorsize != 0 && pp->sectorsize != dp->d_sectorsize) g_wither_provider(pp, ENXIO); else g_resize_provider(pp, dp->d_mediasize); } } static void g_disk_create(void *arg, int flag) { struct g_geom *gp; struct g_provider *pp; struct disk *dp; struct g_disk_softc *sc; char tmpstr[80]; if (flag == EV_CANCEL) return; g_topology_assert(); dp = arg; sc = g_malloc(sizeof(*sc), M_WAITOK | M_ZERO); mtx_init(&sc->start_mtx, "g_disk_start", NULL, MTX_DEF); mtx_init(&sc->done_mtx, "g_disk_done", NULL, MTX_DEF); sc->dp = dp; gp = g_new_geomf(&g_disk_class, "%s%d", dp->d_name, dp->d_unit); gp->softc = sc; pp = g_new_providerf(gp, "%s", gp->name); devstat_remove_entry(pp->stat); pp->stat = NULL; dp->d_devstat->id = pp; pp->mediasize = dp->d_mediasize; pp->sectorsize = dp->d_sectorsize; pp->stripeoffset = dp->d_stripeoffset; pp->stripesize = dp->d_stripesize; if ((dp->d_flags & DISKFLAG_UNMAPPED_BIO) != 0) pp->flags |= G_PF_ACCEPT_UNMAPPED; if ((dp->d_flags & DISKFLAG_DIRECT_COMPLETION) != 0) pp->flags |= G_PF_DIRECT_SEND; pp->flags |= G_PF_DIRECT_RECEIVE; if (bootverbose) printf("GEOM: new disk %s\n", gp->name); sysctl_ctx_init(&sc->sysctl_ctx); snprintf(tmpstr, sizeof(tmpstr), "GEOM disk %s", gp->name); sc->sysctl_tree = SYSCTL_ADD_NODE(&sc->sysctl_ctx, SYSCTL_STATIC_CHILDREN(_kern_geom_disk), OID_AUTO, gp->name, CTLFLAG_RD, 0, tmpstr); if (sc->sysctl_tree != NULL) { snprintf(tmpstr, sizeof(tmpstr), "kern.geom.disk.%s.led", gp->name); TUNABLE_STR_FETCH(tmpstr, sc->led, sizeof(sc->led)); SYSCTL_ADD_STRING(&sc->sysctl_ctx, SYSCTL_CHILDREN(sc->sysctl_tree), OID_AUTO, "led", CTLFLAG_RW | CTLFLAG_TUN, sc->led, sizeof(sc->led), "LED name"); } pp->private = sc; dp->d_geom = gp; g_error_provider(pp, 0); } /* * We get this callback after all of the consumers have gone away, and just * before the provider is freed. If the disk driver provided a d_gone * callback, let them know that it is okay to free resources -- they won't * be getting any more accesses from GEOM. */ static void g_disk_providergone(struct g_provider *pp) { struct disk *dp; struct g_disk_softc *sc; sc = (struct g_disk_softc *)pp->private; dp = sc->dp; if (dp != NULL && dp->d_gone != NULL) dp->d_gone(dp); if (sc->sysctl_tree != NULL) { sysctl_ctx_free(&sc->sysctl_ctx); sc->sysctl_tree = NULL; } if (sc->led[0] != 0) { led_set(sc->led, "0"); sc->led[0] = 0; } pp->private = NULL; pp->geom->softc = NULL; mtx_destroy(&sc->done_mtx); mtx_destroy(&sc->start_mtx); g_free(sc); } static void g_disk_destroy(void *ptr, int flag) { struct disk *dp; struct g_geom *gp; struct g_disk_softc *sc; g_topology_assert(); dp = ptr; gp = dp->d_geom; if (gp != NULL) { sc = gp->softc; if (sc != NULL) sc->dp = NULL; dp->d_geom = NULL; g_wither_geom(gp, ENXIO); } g_free(dp); } /* * We only allow printable characters in disk ident, * the rest is converted to 'x'. */ static void g_disk_ident_adjust(char *ident, size_t size) { char *p, tmp[4], newid[DISK_IDENT_SIZE]; newid[0] = '\0'; for (p = ident; *p != '\0'; p++) { if (isprint(*p)) { tmp[0] = *p; tmp[1] = '\0'; } else { snprintf(tmp, sizeof(tmp), "x%02hhx", *(unsigned char *)p); } if (strlcat(newid, tmp, sizeof(newid)) >= sizeof(newid)) break; } bzero(ident, size); strlcpy(ident, newid, size); } struct disk * disk_alloc(void) { return (g_malloc(sizeof(struct disk), M_WAITOK | M_ZERO)); } void disk_create(struct disk *dp, int version) { if (version != DISK_VERSION) { printf("WARNING: Attempt to add disk %s%d %s", dp->d_name, dp->d_unit, " using incompatible ABI version of disk(9)\n"); printf("WARNING: Ignoring disk %s%d\n", dp->d_name, dp->d_unit); return; } + if (version < DISK_VERSION_04) + dp->d_flags |= DISKFLAG_LACKS_ROTRATE; KASSERT(dp->d_strategy != NULL, ("disk_create need d_strategy")); KASSERT(dp->d_name != NULL, ("disk_create need d_name")); KASSERT(*dp->d_name != 0, ("disk_create need d_name")); KASSERT(strlen(dp->d_name) < SPECNAMELEN - 4, ("disk name too long")); if (dp->d_devstat == NULL) dp->d_devstat = devstat_new_entry(dp->d_name, dp->d_unit, dp->d_sectorsize, DEVSTAT_ALL_SUPPORTED, DEVSTAT_TYPE_DIRECT, DEVSTAT_PRIORITY_MAX); dp->d_geom = NULL; g_disk_ident_adjust(dp->d_ident, sizeof(dp->d_ident)); g_post_event(g_disk_create, dp, M_WAITOK, dp, NULL); } void disk_destroy(struct disk *dp) { g_cancel_event(dp); dp->d_destroyed = 1; if (dp->d_devstat != NULL) devstat_remove_entry(dp->d_devstat); g_post_event(g_disk_destroy, dp, M_WAITOK, NULL); } void disk_gone(struct disk *dp) { struct g_geom *gp; struct g_provider *pp; gp = dp->d_geom; if (gp != NULL) { pp = LIST_FIRST(&gp->provider); if (pp != NULL) { KASSERT(LIST_NEXT(pp, provider) == NULL, ("geom %p has more than one provider", gp)); g_wither_provider(pp, ENXIO); } } } void disk_attr_changed(struct disk *dp, const char *attr, int flag) { struct g_geom *gp; struct g_provider *pp; gp = dp->d_geom; if (gp != NULL) LIST_FOREACH(pp, &gp->provider, provider) (void)g_attr_changed(pp, attr, flag); } void disk_media_changed(struct disk *dp, int flag) { struct g_geom *gp; struct g_provider *pp; gp = dp->d_geom; if (gp != NULL) { pp = LIST_FIRST(&gp->provider); if (pp != NULL) { KASSERT(LIST_NEXT(pp, provider) == NULL, ("geom %p has more than one provider", gp)); g_media_changed(pp, flag); } } } void disk_media_gone(struct disk *dp, int flag) { struct g_geom *gp; struct g_provider *pp; gp = dp->d_geom; if (gp != NULL) { pp = LIST_FIRST(&gp->provider); if (pp != NULL) { KASSERT(LIST_NEXT(pp, provider) == NULL, ("geom %p has more than one provider", gp)); g_media_gone(pp, flag); } } } int disk_resize(struct disk *dp, int flag) { if (dp->d_destroyed || dp->d_geom == NULL) return (0); return (g_post_event(g_disk_resize, dp, flag, NULL)); } static void g_kern_disks(void *p, int flag __unused) { struct sbuf *sb; struct g_geom *gp; char *sp; sb = p; sp = ""; g_topology_assert(); LIST_FOREACH(gp, &g_disk_class.geom, geom) { sbuf_printf(sb, "%s%s", sp, gp->name); sp = " "; } sbuf_finish(sb); } static int sysctl_disks(SYSCTL_HANDLER_ARGS) { int error; struct sbuf *sb; sb = sbuf_new_auto(); g_waitfor_event(g_kern_disks, sb, M_WAITOK, NULL); error = SYSCTL_OUT(req, sbuf_data(sb), sbuf_len(sb) + 1); sbuf_delete(sb); return error; } SYSCTL_PROC(_kern, OID_AUTO, disks, CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0, sysctl_disks, "A", "names of available disks"); Index: stable/10/sys/geom/geom_disk.h =================================================================== --- stable/10/sys/geom/geom_disk.h (revision 271237) +++ stable/10/sys/geom/geom_disk.h (revision 271238) @@ -1,128 +1,133 @@ /*- * Copyright (c) 2003 Poul-Henning Kamp * All rights reserved. * * This software was developed for the FreeBSD Project by Poul-Henning Kamp * and NAI Labs, the Security Research Division of Network Associates, Inc. * under DARPA/SPAWAR contract N66001-01-C-8035 ("CBOSS"), as part of the * DARPA CHATS research program. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. The names of the authors may not be used to endorse or promote * products derived from this software without specific prior written * permission. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * $FreeBSD$ */ #ifndef _GEOM_GEOM_DISK_H_ #define _GEOM_GEOM_DISK_H_ #ifdef _KERNEL #include #include #include #include #define G_DISK_CLASS_NAME "DISK" struct disk; typedef int disk_open_t(struct disk *); typedef int disk_close_t(struct disk *); typedef void disk_strategy_t(struct bio *bp); typedef int disk_getattr_t(struct bio *bp); typedef void disk_gone_t(struct disk *); typedef int disk_ioctl_t(struct disk *, u_long cmd, void *data, int fflag, struct thread *td); /* NB: disk_ioctl_t SHALL be cast'able to d_ioctl_t */ struct g_geom; struct devstat; struct disk { /* Fields which are private to geom_disk */ struct g_geom *d_geom; struct devstat *d_devstat; int d_destroyed; /* Shared fields */ u_int d_flags; const char *d_name; u_int d_unit; struct bio_queue_head *d_queue; struct mtx *d_lock; /* Disk methods */ disk_open_t *d_open; disk_close_t *d_close; disk_strategy_t *d_strategy; disk_ioctl_t *d_ioctl; dumper_t *d_dump; disk_getattr_t *d_getattr; disk_gone_t *d_gone; /* Info fields from driver to geom_disk.c. Valid when open */ u_int d_sectorsize; off_t d_mediasize; u_int d_fwsectors; u_int d_fwheads; u_int d_maxsize; off_t d_delmaxsize; u_int d_stripeoffset; u_int d_stripesize; char d_ident[DISK_IDENT_SIZE]; char d_descr[DISK_IDENT_SIZE]; uint16_t d_hba_vendor; uint16_t d_hba_device; uint16_t d_hba_subvendor; uint16_t d_hba_subdevice; /* Fields private to the driver */ void *d_drv1; + + /* New field - don't use if DISKFLAG_LACKS_ROTRATE is set */ + uint16_t d_rotation_rate; }; #define DISKFLAG_NEEDSGIANT 0x1 #define DISKFLAG_OPEN 0x2 #define DISKFLAG_CANDELETE 0x4 #define DISKFLAG_CANFLUSHCACHE 0x8 #define DISKFLAG_UNMAPPED_BIO 0x10 #define DISKFLAG_DIRECT_COMPLETION 0x20 +#define DISKFLAG_LACKS_ROTRATE 0x40 struct disk *disk_alloc(void); void disk_create(struct disk *disk, int version); void disk_destroy(struct disk *disk); void disk_gone(struct disk *disk); void disk_attr_changed(struct disk *dp, const char *attr, int flag); void disk_media_changed(struct disk *dp, int flag); void disk_media_gone(struct disk *dp, int flag); int disk_resize(struct disk *dp, int flag); #define DISK_VERSION_00 0x58561059 #define DISK_VERSION_01 0x5856105a #define DISK_VERSION_02 0x5856105b #define DISK_VERSION_03 0x5856105c -#define DISK_VERSION DISK_VERSION_03 +#define DISK_VERSION_04 0x5856105d +#define DISK_VERSION DISK_VERSION_04 #endif /* _KERNEL */ #endif /* _GEOM_GEOM_DISK_H_ */ Index: stable/10/sys/geom/geom_subr.c =================================================================== --- stable/10/sys/geom/geom_subr.c (revision 271237) +++ stable/10/sys/geom/geom_subr.c (revision 271238) @@ -1,1522 +1,1529 @@ /*- * Copyright (c) 2002 Poul-Henning Kamp * Copyright (c) 2002 Networks Associates Technology, Inc. * All rights reserved. * * This software was developed for the FreeBSD Project by Poul-Henning Kamp * and NAI Labs, the Security Research Division of Network Associates, Inc. * under DARPA/SPAWAR contract N66001-01-C-8035 ("CBOSS"), as part of the * DARPA CHATS research program. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. The names of the authors may not be used to endorse or promote * products derived from this software without specific prior written * permission. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include "opt_ddb.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef DDB #include #endif #ifdef KDB #include #endif struct class_list_head g_classes = LIST_HEAD_INITIALIZER(g_classes); static struct g_tailq_head geoms = TAILQ_HEAD_INITIALIZER(geoms); char *g_wait_event, *g_wait_up, *g_wait_down, *g_wait_sim; struct g_hh00 { struct g_class *mp; struct g_provider *pp; off_t size; int error; int post; }; /* * This event offers a new class a chance to taste all preexisting providers. */ static void g_load_class(void *arg, int flag) { struct g_hh00 *hh; struct g_class *mp2, *mp; struct g_geom *gp; struct g_provider *pp; g_topology_assert(); if (flag == EV_CANCEL) /* XXX: can't happen ? */ return; if (g_shutdown) return; hh = arg; mp = hh->mp; hh->error = 0; if (hh->post) { g_free(hh); hh = NULL; } g_trace(G_T_TOPOLOGY, "g_load_class(%s)", mp->name); KASSERT(mp->name != NULL && *mp->name != '\0', ("GEOM class has no name")); LIST_FOREACH(mp2, &g_classes, class) { if (mp2 == mp) { printf("The GEOM class %s is already loaded.\n", mp2->name); if (hh != NULL) hh->error = EEXIST; return; } else if (strcmp(mp2->name, mp->name) == 0) { printf("A GEOM class %s is already loaded.\n", mp2->name); if (hh != NULL) hh->error = EEXIST; return; } } LIST_INIT(&mp->geom); LIST_INSERT_HEAD(&g_classes, mp, class); if (mp->init != NULL) mp->init(mp); if (mp->taste == NULL) return; LIST_FOREACH(mp2, &g_classes, class) { if (mp == mp2) continue; LIST_FOREACH(gp, &mp2->geom, geom) { LIST_FOREACH(pp, &gp->provider, provider) { mp->taste(mp, pp, 0); g_topology_assert(); } } } } static int g_unload_class(struct g_class *mp) { struct g_geom *gp; struct g_provider *pp; struct g_consumer *cp; int error; g_topology_lock(); g_trace(G_T_TOPOLOGY, "g_unload_class(%s)", mp->name); retry: G_VALID_CLASS(mp); LIST_FOREACH(gp, &mp->geom, geom) { /* We refuse to unload if anything is open */ LIST_FOREACH(pp, &gp->provider, provider) if (pp->acr || pp->acw || pp->ace) { g_topology_unlock(); return (EBUSY); } LIST_FOREACH(cp, &gp->consumer, consumer) if (cp->acr || cp->acw || cp->ace) { g_topology_unlock(); return (EBUSY); } /* If the geom is withering, wait for it to finish. */ if (gp->flags & G_GEOM_WITHER) { g_topology_sleep(mp, 1); goto retry; } } /* * We allow unloading if we have no geoms, or a class * method we can use to get rid of them. */ if (!LIST_EMPTY(&mp->geom) && mp->destroy_geom == NULL) { g_topology_unlock(); return (EOPNOTSUPP); } /* Bar new entries */ mp->taste = NULL; mp->config = NULL; LIST_FOREACH(gp, &mp->geom, geom) { error = mp->destroy_geom(NULL, mp, gp); if (error != 0) { g_topology_unlock(); return (error); } } /* Wait for withering to finish. */ for (;;) { gp = LIST_FIRST(&mp->geom); if (gp == NULL) break; KASSERT(gp->flags & G_GEOM_WITHER, ("Non-withering geom in class %s", mp->name)); g_topology_sleep(mp, 1); } G_VALID_CLASS(mp); if (mp->fini != NULL) mp->fini(mp); LIST_REMOVE(mp, class); g_topology_unlock(); return (0); } int g_modevent(module_t mod, int type, void *data) { struct g_hh00 *hh; int error; static int g_ignition; struct g_class *mp; mp = data; if (mp->version != G_VERSION) { printf("GEOM class %s has Wrong version %x\n", mp->name, mp->version); return (EINVAL); } if (!g_ignition) { g_ignition++; g_init(); } error = EOPNOTSUPP; switch (type) { case MOD_LOAD: g_trace(G_T_TOPOLOGY, "g_modevent(%s, LOAD)", mp->name); hh = g_malloc(sizeof *hh, M_WAITOK | M_ZERO); hh->mp = mp; /* * Once the system is not cold, MOD_LOAD calls will be * from the userland and the g_event thread will be able * to acknowledge their completion. */ if (cold) { hh->post = 1; error = g_post_event(g_load_class, hh, M_WAITOK, NULL); } else { error = g_waitfor_event(g_load_class, hh, M_WAITOK, NULL); if (error == 0) error = hh->error; g_free(hh); } break; case MOD_UNLOAD: g_trace(G_T_TOPOLOGY, "g_modevent(%s, UNLOAD)", mp->name); DROP_GIANT(); error = g_unload_class(mp); PICKUP_GIANT(); if (error == 0) { KASSERT(LIST_EMPTY(&mp->geom), ("Unloaded class (%s) still has geom", mp->name)); } break; } return (error); } static void g_retaste_event(void *arg, int flag) { struct g_class *mp, *mp2; struct g_geom *gp; struct g_hh00 *hh; struct g_provider *pp; struct g_consumer *cp; g_topology_assert(); if (flag == EV_CANCEL) /* XXX: can't happen ? */ return; if (g_shutdown || g_notaste) return; hh = arg; mp = hh->mp; hh->error = 0; if (hh->post) { g_free(hh); hh = NULL; } g_trace(G_T_TOPOLOGY, "g_retaste(%s)", mp->name); LIST_FOREACH(mp2, &g_classes, class) { LIST_FOREACH(gp, &mp2->geom, geom) { LIST_FOREACH(pp, &gp->provider, provider) { if (pp->acr || pp->acw || pp->ace) continue; LIST_FOREACH(cp, &pp->consumers, consumers) { if (cp->geom->class == mp && (cp->flags & G_CF_ORPHAN) == 0) break; } if (cp != NULL) { cp->flags |= G_CF_ORPHAN; g_wither_geom(cp->geom, ENXIO); } mp->taste(mp, pp, 0); g_topology_assert(); } } } } int g_retaste(struct g_class *mp) { struct g_hh00 *hh; int error; if (mp->taste == NULL) return (EINVAL); hh = g_malloc(sizeof *hh, M_WAITOK | M_ZERO); hh->mp = mp; if (cold) { hh->post = 1; error = g_post_event(g_retaste_event, hh, M_WAITOK, NULL); } else { error = g_waitfor_event(g_retaste_event, hh, M_WAITOK, NULL); if (error == 0) error = hh->error; g_free(hh); } return (error); } struct g_geom * g_new_geomf(struct g_class *mp, const char *fmt, ...) { struct g_geom *gp; va_list ap; struct sbuf *sb; g_topology_assert(); G_VALID_CLASS(mp); sb = sbuf_new_auto(); va_start(ap, fmt); sbuf_vprintf(sb, fmt, ap); va_end(ap); sbuf_finish(sb); gp = g_malloc(sizeof *gp, M_WAITOK | M_ZERO); gp->name = g_malloc(sbuf_len(sb) + 1, M_WAITOK | M_ZERO); gp->class = mp; gp->rank = 1; LIST_INIT(&gp->consumer); LIST_INIT(&gp->provider); LIST_INSERT_HEAD(&mp->geom, gp, geom); TAILQ_INSERT_HEAD(&geoms, gp, geoms); strcpy(gp->name, sbuf_data(sb)); sbuf_delete(sb); /* Fill in defaults from class */ gp->start = mp->start; gp->spoiled = mp->spoiled; gp->attrchanged = mp->attrchanged; gp->providergone = mp->providergone; gp->dumpconf = mp->dumpconf; gp->access = mp->access; gp->orphan = mp->orphan; gp->ioctl = mp->ioctl; gp->resize = mp->resize; return (gp); } void g_destroy_geom(struct g_geom *gp) { g_topology_assert(); G_VALID_GEOM(gp); g_trace(G_T_TOPOLOGY, "g_destroy_geom(%p(%s))", gp, gp->name); KASSERT(LIST_EMPTY(&gp->consumer), ("g_destroy_geom(%s) with consumer(s) [%p]", gp->name, LIST_FIRST(&gp->consumer))); KASSERT(LIST_EMPTY(&gp->provider), ("g_destroy_geom(%s) with provider(s) [%p]", gp->name, LIST_FIRST(&gp->provider))); g_cancel_event(gp); LIST_REMOVE(gp, geom); TAILQ_REMOVE(&geoms, gp, geoms); g_free(gp->name); g_free(gp); } /* * This function is called (repeatedly) until the geom has withered away. */ void g_wither_geom(struct g_geom *gp, int error) { struct g_provider *pp; g_topology_assert(); G_VALID_GEOM(gp); g_trace(G_T_TOPOLOGY, "g_wither_geom(%p(%s))", gp, gp->name); if (!(gp->flags & G_GEOM_WITHER)) { gp->flags |= G_GEOM_WITHER; LIST_FOREACH(pp, &gp->provider, provider) if (!(pp->flags & G_PF_ORPHAN)) g_orphan_provider(pp, error); } g_do_wither(); } /* * Convenience function to destroy a particular provider. */ void g_wither_provider(struct g_provider *pp, int error) { pp->flags |= G_PF_WITHER; if (!(pp->flags & G_PF_ORPHAN)) g_orphan_provider(pp, error); } /* * This function is called (repeatedly) until the has withered away. */ void g_wither_geom_close(struct g_geom *gp, int error) { struct g_consumer *cp; g_topology_assert(); G_VALID_GEOM(gp); g_trace(G_T_TOPOLOGY, "g_wither_geom_close(%p(%s))", gp, gp->name); LIST_FOREACH(cp, &gp->consumer, consumer) if (cp->acr || cp->acw || cp->ace) g_access(cp, -cp->acr, -cp->acw, -cp->ace); g_wither_geom(gp, error); } /* * This function is called (repeatedly) until we cant wash away more * withered bits at present. */ void g_wither_washer() { struct g_class *mp; struct g_geom *gp, *gp2; struct g_provider *pp, *pp2; struct g_consumer *cp, *cp2; g_topology_assert(); LIST_FOREACH(mp, &g_classes, class) { LIST_FOREACH_SAFE(gp, &mp->geom, geom, gp2) { LIST_FOREACH_SAFE(pp, &gp->provider, provider, pp2) { if (!(pp->flags & G_PF_WITHER)) continue; if (LIST_EMPTY(&pp->consumers)) g_destroy_provider(pp); } if (!(gp->flags & G_GEOM_WITHER)) continue; LIST_FOREACH_SAFE(pp, &gp->provider, provider, pp2) { if (LIST_EMPTY(&pp->consumers)) g_destroy_provider(pp); } LIST_FOREACH_SAFE(cp, &gp->consumer, consumer, cp2) { if (cp->acr || cp->acw || cp->ace) continue; if (cp->provider != NULL) g_detach(cp); g_destroy_consumer(cp); } if (LIST_EMPTY(&gp->provider) && LIST_EMPTY(&gp->consumer)) g_destroy_geom(gp); } } } struct g_consumer * g_new_consumer(struct g_geom *gp) { struct g_consumer *cp; g_topology_assert(); G_VALID_GEOM(gp); KASSERT(!(gp->flags & G_GEOM_WITHER), ("g_new_consumer on WITHERing geom(%s) (class %s)", gp->name, gp->class->name)); KASSERT(gp->orphan != NULL, ("g_new_consumer on geom(%s) (class %s) without orphan", gp->name, gp->class->name)); cp = g_malloc(sizeof *cp, M_WAITOK | M_ZERO); cp->geom = gp; cp->stat = devstat_new_entry(cp, -1, 0, DEVSTAT_ALL_SUPPORTED, DEVSTAT_TYPE_DIRECT, DEVSTAT_PRIORITY_MAX); LIST_INSERT_HEAD(&gp->consumer, cp, consumer); return(cp); } void g_destroy_consumer(struct g_consumer *cp) { struct g_geom *gp; g_topology_assert(); G_VALID_CONSUMER(cp); g_trace(G_T_TOPOLOGY, "g_destroy_consumer(%p)", cp); KASSERT (cp->provider == NULL, ("g_destroy_consumer but attached")); KASSERT (cp->acr == 0, ("g_destroy_consumer with acr")); KASSERT (cp->acw == 0, ("g_destroy_consumer with acw")); KASSERT (cp->ace == 0, ("g_destroy_consumer with ace")); g_cancel_event(cp); gp = cp->geom; LIST_REMOVE(cp, consumer); devstat_remove_entry(cp->stat); g_free(cp); if (gp->flags & G_GEOM_WITHER) g_do_wither(); } static void g_new_provider_event(void *arg, int flag) { struct g_class *mp; struct g_provider *pp; struct g_consumer *cp, *next_cp; g_topology_assert(); if (flag == EV_CANCEL) return; if (g_shutdown) return; pp = arg; G_VALID_PROVIDER(pp); KASSERT(!(pp->flags & G_PF_WITHER), ("g_new_provider_event but withered")); LIST_FOREACH_SAFE(cp, &pp->consumers, consumers, next_cp) { if ((cp->flags & G_CF_ORPHAN) == 0 && cp->geom->attrchanged != NULL) cp->geom->attrchanged(cp, "GEOM::media"); } if (g_notaste) return; LIST_FOREACH(mp, &g_classes, class) { if (mp->taste == NULL) continue; LIST_FOREACH(cp, &pp->consumers, consumers) if (cp->geom->class == mp && (cp->flags & G_CF_ORPHAN) == 0) break; if (cp != NULL) continue; mp->taste(mp, pp, 0); g_topology_assert(); } } struct g_provider * g_new_providerf(struct g_geom *gp, const char *fmt, ...) { struct g_provider *pp; struct sbuf *sb; va_list ap; g_topology_assert(); G_VALID_GEOM(gp); KASSERT(gp->access != NULL, ("new provider on geom(%s) without ->access (class %s)", gp->name, gp->class->name)); KASSERT(gp->start != NULL, ("new provider on geom(%s) without ->start (class %s)", gp->name, gp->class->name)); KASSERT(!(gp->flags & G_GEOM_WITHER), ("new provider on WITHERing geom(%s) (class %s)", gp->name, gp->class->name)); sb = sbuf_new_auto(); va_start(ap, fmt); sbuf_vprintf(sb, fmt, ap); va_end(ap); sbuf_finish(sb); pp = g_malloc(sizeof *pp + sbuf_len(sb) + 1, M_WAITOK | M_ZERO); pp->name = (char *)(pp + 1); strcpy(pp->name, sbuf_data(sb)); sbuf_delete(sb); LIST_INIT(&pp->consumers); pp->error = ENXIO; pp->geom = gp; pp->stat = devstat_new_entry(pp, -1, 0, DEVSTAT_ALL_SUPPORTED, DEVSTAT_TYPE_DIRECT, DEVSTAT_PRIORITY_MAX); LIST_INSERT_HEAD(&gp->provider, pp, provider); g_post_event(g_new_provider_event, pp, M_WAITOK, pp, gp, NULL); return (pp); } void g_error_provider(struct g_provider *pp, int error) { /* G_VALID_PROVIDER(pp); We may not have g_topology */ pp->error = error; } static void g_resize_provider_event(void *arg, int flag) { struct g_hh00 *hh; struct g_class *mp; struct g_geom *gp; struct g_provider *pp; struct g_consumer *cp, *cp2; off_t size; g_topology_assert(); if (g_shutdown) return; hh = arg; pp = hh->pp; size = hh->size; g_free(hh); G_VALID_PROVIDER(pp); g_trace(G_T_TOPOLOGY, "g_resize_provider_event(%p)", pp); LIST_FOREACH_SAFE(cp, &pp->consumers, consumers, cp2) { gp = cp->geom; if (gp->resize == NULL && size < pp->mediasize) { cp->flags |= G_CF_ORPHAN; cp->geom->orphan(cp); } } pp->mediasize = size; LIST_FOREACH_SAFE(cp, &pp->consumers, consumers, cp2) { gp = cp->geom; if (gp->resize != NULL) gp->resize(cp); } /* * After resizing, the previously invalid GEOM class metadata * might become valid. This means we should retaste. */ LIST_FOREACH(mp, &g_classes, class) { if (mp->taste == NULL) continue; LIST_FOREACH(cp, &pp->consumers, consumers) if (cp->geom->class == mp && (cp->flags & G_CF_ORPHAN) == 0) break; if (cp != NULL) continue; mp->taste(mp, pp, 0); g_topology_assert(); } } void g_resize_provider(struct g_provider *pp, off_t size) { struct g_hh00 *hh; G_VALID_PROVIDER(pp); if (size == pp->mediasize) return; hh = g_malloc(sizeof *hh, M_WAITOK | M_ZERO); hh->pp = pp; hh->size = size; g_post_event(g_resize_provider_event, hh, M_WAITOK, NULL); } #ifndef _PATH_DEV #define _PATH_DEV "/dev/" #endif struct g_provider * g_provider_by_name(char const *arg) { struct g_class *cp; struct g_geom *gp; struct g_provider *pp; if (strncmp(arg, _PATH_DEV, sizeof(_PATH_DEV) - 1) == 0) arg += sizeof(_PATH_DEV) - 1; LIST_FOREACH(cp, &g_classes, class) { LIST_FOREACH(gp, &cp->geom, geom) { LIST_FOREACH(pp, &gp->provider, provider) { if (!strcmp(arg, pp->name)) return (pp); } } } return (NULL); } void g_destroy_provider(struct g_provider *pp) { struct g_geom *gp; g_topology_assert(); G_VALID_PROVIDER(pp); KASSERT(LIST_EMPTY(&pp->consumers), ("g_destroy_provider but attached")); KASSERT (pp->acr == 0, ("g_destroy_provider with acr")); KASSERT (pp->acw == 0, ("g_destroy_provider with acw")); KASSERT (pp->ace == 0, ("g_destroy_provider with ace")); g_cancel_event(pp); LIST_REMOVE(pp, provider); gp = pp->geom; devstat_remove_entry(pp->stat); /* * If a callback was provided, send notification that the provider * is now gone. */ if (gp->providergone != NULL) gp->providergone(pp); g_free(pp); if ((gp->flags & G_GEOM_WITHER)) g_do_wither(); } /* * We keep the "geoms" list sorted by topological order (== increasing * numerical rank) at all times. * When an attach is done, the attaching geoms rank is invalidated * and it is moved to the tail of the list. * All geoms later in the sequence has their ranks reevaluated in * sequence. If we cannot assign rank to a geom because it's * prerequisites do not have rank, we move that element to the tail * of the sequence with invalid rank as well. * At some point we encounter our original geom and if we stil fail * to assign it a rank, there must be a loop and we fail back to * g_attach() which detach again and calls redo_rank again * to fix up the damage. * It would be much simpler code wise to do it recursively, but we * can't risk that on the kernel stack. */ static int redo_rank(struct g_geom *gp) { struct g_consumer *cp; struct g_geom *gp1, *gp2; int n, m; g_topology_assert(); G_VALID_GEOM(gp); /* Invalidate this geoms rank and move it to the tail */ gp1 = TAILQ_NEXT(gp, geoms); if (gp1 != NULL) { gp->rank = 0; TAILQ_REMOVE(&geoms, gp, geoms); TAILQ_INSERT_TAIL(&geoms, gp, geoms); } else { gp1 = gp; } /* re-rank the rest of the sequence */ for (; gp1 != NULL; gp1 = gp2) { gp1->rank = 0; m = 1; LIST_FOREACH(cp, &gp1->consumer, consumer) { if (cp->provider == NULL) continue; n = cp->provider->geom->rank; if (n == 0) { m = 0; break; } else if (n >= m) m = n + 1; } gp1->rank = m; gp2 = TAILQ_NEXT(gp1, geoms); /* got a rank, moving on */ if (m != 0) continue; /* no rank to original geom means loop */ if (gp == gp1) return (ELOOP); /* no rank, put it at the end move on */ TAILQ_REMOVE(&geoms, gp1, geoms); TAILQ_INSERT_TAIL(&geoms, gp1, geoms); } return (0); } int g_attach(struct g_consumer *cp, struct g_provider *pp) { int error; g_topology_assert(); G_VALID_CONSUMER(cp); G_VALID_PROVIDER(pp); g_trace(G_T_TOPOLOGY, "g_attach(%p, %p)", cp, pp); KASSERT(cp->provider == NULL, ("attach but attached")); cp->provider = pp; LIST_INSERT_HEAD(&pp->consumers, cp, consumers); error = redo_rank(cp->geom); if (error) { LIST_REMOVE(cp, consumers); cp->provider = NULL; redo_rank(cp->geom); } return (error); } void g_detach(struct g_consumer *cp) { struct g_provider *pp; g_topology_assert(); G_VALID_CONSUMER(cp); g_trace(G_T_TOPOLOGY, "g_detach(%p)", cp); KASSERT(cp->provider != NULL, ("detach but not attached")); KASSERT(cp->acr == 0, ("detach but nonzero acr")); KASSERT(cp->acw == 0, ("detach but nonzero acw")); KASSERT(cp->ace == 0, ("detach but nonzero ace")); KASSERT(cp->nstart == cp->nend, ("detach with active requests")); pp = cp->provider; LIST_REMOVE(cp, consumers); cp->provider = NULL; if ((cp->geom->flags & G_GEOM_WITHER) || (pp->geom->flags & G_GEOM_WITHER) || (pp->flags & G_PF_WITHER)) g_do_wither(); redo_rank(cp->geom); } /* * g_access() * * Access-check with delta values. The question asked is "can provider * "cp" change the access counters by the relative amounts dc[rwe] ?" */ int g_access(struct g_consumer *cp, int dcr, int dcw, int dce) { struct g_provider *pp; int pr,pw,pe; int error; g_topology_assert(); G_VALID_CONSUMER(cp); pp = cp->provider; KASSERT(pp != NULL, ("access but not attached")); G_VALID_PROVIDER(pp); g_trace(G_T_ACCESS, "g_access(%p(%s), %d, %d, %d)", cp, pp->name, dcr, dcw, dce); KASSERT(cp->acr + dcr >= 0, ("access resulting in negative acr")); KASSERT(cp->acw + dcw >= 0, ("access resulting in negative acw")); KASSERT(cp->ace + dce >= 0, ("access resulting in negative ace")); KASSERT(dcr != 0 || dcw != 0 || dce != 0, ("NOP access request")); KASSERT(pp->geom->access != NULL, ("NULL geom->access")); /* * If our class cares about being spoiled, and we have been, we * are probably just ahead of the event telling us that. Fail * now rather than having to unravel this later. */ if (cp->geom->spoiled != NULL && (cp->flags & G_CF_SPOILED) && (dcr > 0 || dcw > 0 || dce > 0)) return (ENXIO); /* * Figure out what counts the provider would have had, if this * consumer had (r0w0e0) at this time. */ pr = pp->acr - cp->acr; pw = pp->acw - cp->acw; pe = pp->ace - cp->ace; g_trace(G_T_ACCESS, "open delta:[r%dw%de%d] old:[r%dw%de%d] provider:[r%dw%de%d] %p(%s)", dcr, dcw, dce, cp->acr, cp->acw, cp->ace, pp->acr, pp->acw, pp->ace, pp, pp->name); /* If foot-shooting is enabled, any open on rank#1 is OK */ if ((g_debugflags & 16) && pp->geom->rank == 1) ; /* If we try exclusive but already write: fail */ else if (dce > 0 && pw > 0) return (EPERM); /* If we try write but already exclusive: fail */ else if (dcw > 0 && pe > 0) return (EPERM); /* If we try to open more but provider is error'ed: fail */ else if ((dcr > 0 || dcw > 0 || dce > 0) && pp->error != 0) return (pp->error); /* Ok then... */ error = pp->geom->access(pp, dcr, dcw, dce); KASSERT(dcr > 0 || dcw > 0 || dce > 0 || error == 0, ("Geom provider %s::%s dcr=%d dcw=%d dce=%d error=%d failed " "closing ->access()", pp->geom->class->name, pp->name, dcr, dcw, dce, error)); if (!error) { /* * If we open first write, spoil any partner consumers. * If we close last write and provider is not errored, * trigger re-taste. */ if (pp->acw == 0 && dcw != 0) g_spoil(pp, cp); else if (pp->acw != 0 && pp->acw == -dcw && pp->error == 0 && !(pp->geom->flags & G_GEOM_WITHER)) g_post_event(g_new_provider_event, pp, M_WAITOK, pp, NULL); pp->acr += dcr; pp->acw += dcw; pp->ace += dce; cp->acr += dcr; cp->acw += dcw; cp->ace += dce; if (pp->acr != 0 || pp->acw != 0 || pp->ace != 0) KASSERT(pp->sectorsize > 0, ("Provider %s lacks sectorsize", pp->name)); if ((cp->geom->flags & G_GEOM_WITHER) && cp->acr == 0 && cp->acw == 0 && cp->ace == 0) g_do_wither(); } return (error); } int g_handleattr_int(struct bio *bp, const char *attribute, int val) { return (g_handleattr(bp, attribute, &val, sizeof val)); } int +g_handleattr_uint16_t(struct bio *bp, const char *attribute, uint16_t val) +{ + + return (g_handleattr(bp, attribute, &val, sizeof val)); +} + +int g_handleattr_off_t(struct bio *bp, const char *attribute, off_t val) { return (g_handleattr(bp, attribute, &val, sizeof val)); } int g_handleattr_str(struct bio *bp, const char *attribute, const char *str) { return (g_handleattr(bp, attribute, str, 0)); } int g_handleattr(struct bio *bp, const char *attribute, const void *val, int len) { int error = 0; if (strcmp(bp->bio_attribute, attribute)) return (0); if (len == 0) { bzero(bp->bio_data, bp->bio_length); if (strlcpy(bp->bio_data, val, bp->bio_length) >= bp->bio_length) { printf("%s: %s bio_length %jd len %zu -> EFAULT\n", __func__, bp->bio_to->name, (intmax_t)bp->bio_length, strlen(val)); error = EFAULT; } } else if (bp->bio_length == len) { bcopy(val, bp->bio_data, len); } else { printf("%s: %s bio_length %jd len %d -> EFAULT\n", __func__, bp->bio_to->name, (intmax_t)bp->bio_length, len); error = EFAULT; } if (error == 0) bp->bio_completed = bp->bio_length; g_io_deliver(bp, error); return (1); } int g_std_access(struct g_provider *pp, int dr __unused, int dw __unused, int de __unused) { g_topology_assert(); G_VALID_PROVIDER(pp); return (0); } void g_std_done(struct bio *bp) { struct bio *bp2; bp2 = bp->bio_parent; if (bp2->bio_error == 0) bp2->bio_error = bp->bio_error; bp2->bio_completed += bp->bio_completed; g_destroy_bio(bp); bp2->bio_inbed++; if (bp2->bio_children == bp2->bio_inbed) g_io_deliver(bp2, bp2->bio_error); } /* XXX: maybe this is only g_slice_spoiled */ void g_std_spoiled(struct g_consumer *cp) { struct g_geom *gp; struct g_provider *pp; g_topology_assert(); G_VALID_CONSUMER(cp); g_trace(G_T_TOPOLOGY, "g_std_spoiled(%p)", cp); cp->flags |= G_CF_ORPHAN; g_detach(cp); gp = cp->geom; LIST_FOREACH(pp, &gp->provider, provider) g_orphan_provider(pp, ENXIO); g_destroy_consumer(cp); if (LIST_EMPTY(&gp->provider) && LIST_EMPTY(&gp->consumer)) g_destroy_geom(gp); else gp->flags |= G_GEOM_WITHER; } /* * Spoiling happens when a provider is opened for writing, but consumers * which are configured by in-band data are attached (slicers for instance). * Since the write might potentially change the in-band data, such consumers * need to re-evaluate their existence after the writing session closes. * We do this by (offering to) tear them down when the open for write happens * in return for a re-taste when it closes again. * Together with the fact that such consumers grab an 'e' bit whenever they * are open, regardless of mode, this ends up DTRT. */ static void g_spoil_event(void *arg, int flag) { struct g_provider *pp; struct g_consumer *cp, *cp2; g_topology_assert(); if (flag == EV_CANCEL) return; pp = arg; G_VALID_PROVIDER(pp); g_trace(G_T_TOPOLOGY, "%s %p(%s:%s:%s)", __func__, pp, pp->geom->class->name, pp->geom->name, pp->name); for (cp = LIST_FIRST(&pp->consumers); cp != NULL; cp = cp2) { cp2 = LIST_NEXT(cp, consumers); if ((cp->flags & G_CF_SPOILED) == 0) continue; cp->flags &= ~G_CF_SPOILED; if (cp->geom->spoiled == NULL) continue; cp->geom->spoiled(cp); g_topology_assert(); } } void g_spoil(struct g_provider *pp, struct g_consumer *cp) { struct g_consumer *cp2; g_topology_assert(); G_VALID_PROVIDER(pp); G_VALID_CONSUMER(cp); LIST_FOREACH(cp2, &pp->consumers, consumers) { if (cp2 == cp) continue; /* KASSERT(cp2->acr == 0, ("spoiling cp->acr = %d", cp2->acr)); KASSERT(cp2->acw == 0, ("spoiling cp->acw = %d", cp2->acw)); */ KASSERT(cp2->ace == 0, ("spoiling cp->ace = %d", cp2->ace)); cp2->flags |= G_CF_SPOILED; } g_post_event(g_spoil_event, pp, M_WAITOK, pp, NULL); } static void g_media_changed_event(void *arg, int flag) { struct g_provider *pp; int retaste; g_topology_assert(); if (flag == EV_CANCEL) return; pp = arg; G_VALID_PROVIDER(pp); /* * If provider was not open for writing, queue retaste after spoiling. * If it was, retaste will happen automatically on close. */ retaste = (pp->acw == 0 && pp->error == 0 && !(pp->geom->flags & G_GEOM_WITHER)); g_spoil_event(arg, flag); if (retaste) g_post_event(g_new_provider_event, pp, M_WAITOK, pp, NULL); } int g_media_changed(struct g_provider *pp, int flag) { struct g_consumer *cp; LIST_FOREACH(cp, &pp->consumers, consumers) cp->flags |= G_CF_SPOILED; return (g_post_event(g_media_changed_event, pp, flag, pp, NULL)); } int g_media_gone(struct g_provider *pp, int flag) { struct g_consumer *cp; LIST_FOREACH(cp, &pp->consumers, consumers) cp->flags |= G_CF_SPOILED; return (g_post_event(g_spoil_event, pp, flag, pp, NULL)); } int g_getattr__(const char *attr, struct g_consumer *cp, void *var, int len) { int error, i; i = len; error = g_io_getattr(attr, cp, &i, var); if (error) return (error); if (i != len) return (EINVAL); return (0); } static int g_get_device_prefix_len(const char *name) { int len; if (strncmp(name, "ada", 3) == 0) len = 3; else if (strncmp(name, "ad", 2) == 0) len = 2; else return (0); if (name[len] < '0' || name[len] > '9') return (0); do { len++; } while (name[len] >= '0' && name[len] <= '9'); return (len); } int g_compare_names(const char *namea, const char *nameb) { int deva, devb; if (strcmp(namea, nameb) == 0) return (1); deva = g_get_device_prefix_len(namea); if (deva == 0) return (0); devb = g_get_device_prefix_len(nameb); if (devb == 0) return (0); if (strcmp(namea + deva, nameb + devb) == 0) return (1); return (0); } #if defined(DIAGNOSTIC) || defined(DDB) /* * This function walks the mesh and returns a non-zero integer if it * finds the argument pointer is an object. The return value indicates * which type of object it is believed to be. If topology is not locked, * this function is potentially dangerous, but we don't assert that the * topology lock is held when called from debugger. */ int g_valid_obj(void const *ptr) { struct g_class *mp; struct g_geom *gp; struct g_consumer *cp; struct g_provider *pp; #ifdef KDB if (kdb_active == 0) #endif g_topology_assert(); LIST_FOREACH(mp, &g_classes, class) { if (ptr == mp) return (1); LIST_FOREACH(gp, &mp->geom, geom) { if (ptr == gp) return (2); LIST_FOREACH(cp, &gp->consumer, consumer) if (ptr == cp) return (3); LIST_FOREACH(pp, &gp->provider, provider) if (ptr == pp) return (4); } } return(0); } #endif #ifdef DDB #define gprintf(...) do { \ db_printf("%*s", indent, ""); \ db_printf(__VA_ARGS__); \ } while (0) #define gprintln(...) do { \ gprintf(__VA_ARGS__); \ db_printf("\n"); \ } while (0) #define ADDFLAG(obj, flag, sflag) do { \ if ((obj)->flags & (flag)) { \ if (comma) \ strlcat(str, ",", size); \ strlcat(str, (sflag), size); \ comma = 1; \ } \ } while (0) static char * provider_flags_to_string(struct g_provider *pp, char *str, size_t size) { int comma = 0; bzero(str, size); if (pp->flags == 0) { strlcpy(str, "NONE", size); return (str); } ADDFLAG(pp, G_PF_WITHER, "G_PF_WITHER"); ADDFLAG(pp, G_PF_ORPHAN, "G_PF_ORPHAN"); return (str); } static char * geom_flags_to_string(struct g_geom *gp, char *str, size_t size) { int comma = 0; bzero(str, size); if (gp->flags == 0) { strlcpy(str, "NONE", size); return (str); } ADDFLAG(gp, G_GEOM_WITHER, "G_GEOM_WITHER"); return (str); } static void db_show_geom_consumer(int indent, struct g_consumer *cp) { if (indent == 0) { gprintln("consumer: %p", cp); gprintln(" class: %s (%p)", cp->geom->class->name, cp->geom->class); gprintln(" geom: %s (%p)", cp->geom->name, cp->geom); if (cp->provider == NULL) gprintln(" provider: none"); else { gprintln(" provider: %s (%p)", cp->provider->name, cp->provider); } gprintln(" access: r%dw%de%d", cp->acr, cp->acw, cp->ace); gprintln(" flags: 0x%04x", cp->flags); gprintln(" nstart: %u", cp->nstart); gprintln(" nend: %u", cp->nend); } else { gprintf("consumer: %p (%s), access=r%dw%de%d", cp, cp->provider != NULL ? cp->provider->name : "none", cp->acr, cp->acw, cp->ace); if (cp->flags) db_printf(", flags=0x%04x", cp->flags); db_printf("\n"); } } static void db_show_geom_provider(int indent, struct g_provider *pp) { struct g_consumer *cp; char flags[64]; if (indent == 0) { gprintln("provider: %s (%p)", pp->name, pp); gprintln(" class: %s (%p)", pp->geom->class->name, pp->geom->class); gprintln(" geom: %s (%p)", pp->geom->name, pp->geom); gprintln(" mediasize: %jd", (intmax_t)pp->mediasize); gprintln(" sectorsize: %u", pp->sectorsize); gprintln(" stripesize: %u", pp->stripesize); gprintln(" stripeoffset: %u", pp->stripeoffset); gprintln(" access: r%dw%de%d", pp->acr, pp->acw, pp->ace); gprintln(" flags: %s (0x%04x)", provider_flags_to_string(pp, flags, sizeof(flags)), pp->flags); gprintln(" error: %d", pp->error); gprintln(" nstart: %u", pp->nstart); gprintln(" nend: %u", pp->nend); if (LIST_EMPTY(&pp->consumers)) gprintln(" consumers: none"); } else { gprintf("provider: %s (%p), access=r%dw%de%d", pp->name, pp, pp->acr, pp->acw, pp->ace); if (pp->flags != 0) { db_printf(", flags=%s (0x%04x)", provider_flags_to_string(pp, flags, sizeof(flags)), pp->flags); } db_printf("\n"); } if (!LIST_EMPTY(&pp->consumers)) { LIST_FOREACH(cp, &pp->consumers, consumers) { db_show_geom_consumer(indent + 2, cp); if (db_pager_quit) break; } } } static void db_show_geom_geom(int indent, struct g_geom *gp) { struct g_provider *pp; struct g_consumer *cp; char flags[64]; if (indent == 0) { gprintln("geom: %s (%p)", gp->name, gp); gprintln(" class: %s (%p)", gp->class->name, gp->class); gprintln(" flags: %s (0x%04x)", geom_flags_to_string(gp, flags, sizeof(flags)), gp->flags); gprintln(" rank: %d", gp->rank); if (LIST_EMPTY(&gp->provider)) gprintln(" providers: none"); if (LIST_EMPTY(&gp->consumer)) gprintln(" consumers: none"); } else { gprintf("geom: %s (%p), rank=%d", gp->name, gp, gp->rank); if (gp->flags != 0) { db_printf(", flags=%s (0x%04x)", geom_flags_to_string(gp, flags, sizeof(flags)), gp->flags); } db_printf("\n"); } if (!LIST_EMPTY(&gp->provider)) { LIST_FOREACH(pp, &gp->provider, provider) { db_show_geom_provider(indent + 2, pp); if (db_pager_quit) break; } } if (!LIST_EMPTY(&gp->consumer)) { LIST_FOREACH(cp, &gp->consumer, consumer) { db_show_geom_consumer(indent + 2, cp); if (db_pager_quit) break; } } } static void db_show_geom_class(struct g_class *mp) { struct g_geom *gp; db_printf("class: %s (%p)\n", mp->name, mp); LIST_FOREACH(gp, &mp->geom, geom) { db_show_geom_geom(2, gp); if (db_pager_quit) break; } } /* * Print the GEOM topology or the given object. */ DB_SHOW_COMMAND(geom, db_show_geom) { struct g_class *mp; if (!have_addr) { /* No address given, print the entire topology. */ LIST_FOREACH(mp, &g_classes, class) { db_show_geom_class(mp); db_printf("\n"); if (db_pager_quit) break; } } else { switch (g_valid_obj((void *)addr)) { case 1: db_show_geom_class((struct g_class *)addr); break; case 2: db_show_geom_geom(0, (struct g_geom *)addr); break; case 3: db_show_geom_consumer(0, (struct g_consumer *)addr); break; case 4: db_show_geom_provider(0, (struct g_provider *)addr); break; default: db_printf("Not a GEOM object.\n"); break; } } } static void db_print_bio_cmd(struct bio *bp) { db_printf(" cmd: "); switch (bp->bio_cmd) { case BIO_READ: db_printf("BIO_READ"); break; case BIO_WRITE: db_printf("BIO_WRITE"); break; case BIO_DELETE: db_printf("BIO_DELETE"); break; case BIO_GETATTR: db_printf("BIO_GETATTR"); break; case BIO_FLUSH: db_printf("BIO_FLUSH"); break; case BIO_CMD0: db_printf("BIO_CMD0"); break; case BIO_CMD1: db_printf("BIO_CMD1"); break; case BIO_CMD2: db_printf("BIO_CMD2"); break; default: db_printf("UNKNOWN"); break; } db_printf("\n"); } static void db_print_bio_flags(struct bio *bp) { int comma; comma = 0; db_printf(" flags: "); if (bp->bio_flags & BIO_ERROR) { db_printf("BIO_ERROR"); comma = 1; } if (bp->bio_flags & BIO_DONE) { db_printf("%sBIO_DONE", (comma ? ", " : "")); comma = 1; } if (bp->bio_flags & BIO_ONQUEUE) db_printf("%sBIO_ONQUEUE", (comma ? ", " : "")); db_printf("\n"); } /* * Print useful information in a BIO */ DB_SHOW_COMMAND(bio, db_show_bio) { struct bio *bp; if (have_addr) { bp = (struct bio *)addr; db_printf("BIO %p\n", bp); db_print_bio_cmd(bp); db_print_bio_flags(bp); db_printf(" cflags: 0x%hhx\n", bp->bio_cflags); db_printf(" pflags: 0x%hhx\n", bp->bio_pflags); db_printf(" offset: %jd\n", (intmax_t)bp->bio_offset); db_printf(" length: %jd\n", (intmax_t)bp->bio_length); db_printf(" bcount: %ld\n", bp->bio_bcount); db_printf(" resid: %ld\n", bp->bio_resid); db_printf(" completed: %jd\n", (intmax_t)bp->bio_completed); db_printf(" children: %u\n", bp->bio_children); db_printf(" inbed: %u\n", bp->bio_inbed); db_printf(" error: %d\n", bp->bio_error); db_printf(" parent: %p\n", bp->bio_parent); db_printf(" driver1: %p\n", bp->bio_driver1); db_printf(" driver2: %p\n", bp->bio_driver2); db_printf(" caller1: %p\n", bp->bio_caller1); db_printf(" caller2: %p\n", bp->bio_caller2); db_printf(" bio_from: %p\n", bp->bio_from); db_printf(" bio_to: %p\n", bp->bio_to); } } #undef gprintf #undef gprintln #undef ADDFLAG #endif /* DDB */ Index: stable/10/sys/sys/ata.h =================================================================== --- stable/10/sys/sys/ata.h (revision 271237) +++ stable/10/sys/sys/ata.h (revision 271238) @@ -1,633 +1,635 @@ /*- * Copyright (c) 2000 - 2008 Søren Schmidt * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer, * without modification, immediately at the beginning of the file. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. * * $FreeBSD$ */ #ifndef _SYS_ATA_H_ #define _SYS_ATA_H_ #include /* ATA/ATAPI device parameters */ struct ata_params { /*000*/ u_int16_t config; /* configuration info */ #define ATA_PROTO_MASK 0x8003 #define ATA_PROTO_ATAPI 0x8000 #define ATA_PROTO_ATAPI_12 0x8000 #define ATA_PROTO_ATAPI_16 0x8001 #define ATA_PROTO_CFA 0x848a #define ATA_ATAPI_TYPE_MASK 0x1f00 #define ATA_ATAPI_TYPE_DIRECT 0x0000 /* disk/floppy */ #define ATA_ATAPI_TYPE_TAPE 0x0100 /* streaming tape */ #define ATA_ATAPI_TYPE_CDROM 0x0500 /* CD-ROM device */ #define ATA_ATAPI_TYPE_OPTICAL 0x0700 /* optical disk */ #define ATA_DRQ_MASK 0x0060 #define ATA_DRQ_SLOW 0x0000 /* cpu 3 ms delay */ #define ATA_DRQ_INTR 0x0020 /* interrupt 10 ms delay */ #define ATA_DRQ_FAST 0x0040 /* accel 50 us delay */ #define ATA_RESP_INCOMPLETE 0x0004 /*001*/ u_int16_t cylinders; /* # of cylinders */ /*002*/ u_int16_t specconf; /* specific configuration */ /*003*/ u_int16_t heads; /* # heads */ u_int16_t obsolete4; u_int16_t obsolete5; /*006*/ u_int16_t sectors; /* # sectors/track */ /*007*/ u_int16_t vendor7[3]; /*010*/ u_int8_t serial[20]; /* serial number */ /*020*/ u_int16_t retired20; u_int16_t retired21; u_int16_t obsolete22; /*023*/ u_int8_t revision[8]; /* firmware revision */ /*027*/ u_int8_t model[40]; /* model name */ /*047*/ u_int16_t sectors_intr; /* sectors per interrupt */ /*048*/ u_int16_t usedmovsd; /* double word read/write? */ /*049*/ u_int16_t capabilities1; #define ATA_SUPPORT_DMA 0x0100 #define ATA_SUPPORT_LBA 0x0200 #define ATA_SUPPORT_IORDY 0x0400 #define ATA_SUPPORT_IORDYDIS 0x0800 #define ATA_SUPPORT_OVERLAP 0x4000 /*050*/ u_int16_t capabilities2; /*051*/ u_int16_t retired_piomode; /* PIO modes 0-2 */ #define ATA_RETIRED_PIO_MASK 0x0300 /*052*/ u_int16_t retired_dmamode; /* DMA modes */ #define ATA_RETIRED_DMA_MASK 0x0003 /*053*/ u_int16_t atavalid; /* fields valid */ #define ATA_FLAG_54_58 0x0001 /* words 54-58 valid */ #define ATA_FLAG_64_70 0x0002 /* words 64-70 valid */ #define ATA_FLAG_88 0x0004 /* word 88 valid */ /*054*/ u_int16_t current_cylinders; /*055*/ u_int16_t current_heads; /*056*/ u_int16_t current_sectors; /*057*/ u_int16_t current_size_1; /*058*/ u_int16_t current_size_2; /*059*/ u_int16_t multi; #define ATA_MULTI_VALID 0x0100 /*060*/ u_int16_t lba_size_1; u_int16_t lba_size_2; u_int16_t obsolete62; /*063*/ u_int16_t mwdmamodes; /* multiword DMA modes */ /*064*/ u_int16_t apiomodes; /* advanced PIO modes */ /*065*/ u_int16_t mwdmamin; /* min. M/W DMA time/word ns */ /*066*/ u_int16_t mwdmarec; /* rec. M/W DMA time ns */ /*067*/ u_int16_t pioblind; /* min. PIO cycle w/o flow */ /*068*/ u_int16_t pioiordy; /* min. PIO cycle IORDY flow */ /*069*/ u_int16_t support3; #define ATA_SUPPORT_RZAT 0x0020 #define ATA_SUPPORT_DRAT 0x4000 u_int16_t reserved70; /*071*/ u_int16_t rlsovlap; /* rel time (us) for overlap */ /*072*/ u_int16_t rlsservice; /* rel time (us) for service */ u_int16_t reserved73; u_int16_t reserved74; /*075*/ u_int16_t queue; #define ATA_QUEUE_LEN(x) ((x) & 0x001f) /*76*/ u_int16_t satacapabilities; #define ATA_SATA_GEN1 0x0002 #define ATA_SATA_GEN2 0x0004 #define ATA_SATA_GEN3 0x0008 #define ATA_SUPPORT_NCQ 0x0100 #define ATA_SUPPORT_IFPWRMNGTRCV 0x0200 #define ATA_SUPPORT_PHYEVENTCNT 0x0400 #define ATA_SUPPORT_NCQ_UNLOAD 0x0800 #define ATA_SUPPORT_NCQ_PRIO 0x1000 #define ATA_SUPPORT_HAPST 0x2000 #define ATA_SUPPORT_DAPST 0x4000 #define ATA_SUPPORT_READLOGDMAEXT 0x8000 /*77*/ u_int16_t satacapabilities2; #define ATA_SATA_CURR_GEN_MASK 0x0006 #define ATA_SUPPORT_NCQ_STREAM 0x0010 #define ATA_SUPPORT_NCQ_QMANAGEMENT 0x0020 #define ATA_SUPPORT_RCVSND_FPDMA_QUEUED 0x0040 /*78*/ u_int16_t satasupport; #define ATA_SUPPORT_NONZERO 0x0002 #define ATA_SUPPORT_AUTOACTIVATE 0x0004 #define ATA_SUPPORT_IFPWRMNGT 0x0008 #define ATA_SUPPORT_INORDERDATA 0x0010 #define ATA_SUPPORT_ASYNCNOTIF 0x0020 #define ATA_SUPPORT_SOFTSETPRESERVE 0x0040 /*79*/ u_int16_t sataenabled; #define ATA_ENABLED_DAPST 0x0080 /*080*/ u_int16_t version_major; /*081*/ u_int16_t version_minor; struct { /*082/085*/ u_int16_t command1; #define ATA_SUPPORT_SMART 0x0001 #define ATA_SUPPORT_SECURITY 0x0002 #define ATA_SUPPORT_REMOVABLE 0x0004 #define ATA_SUPPORT_POWERMGT 0x0008 #define ATA_SUPPORT_PACKET 0x0010 #define ATA_SUPPORT_WRITECACHE 0x0020 #define ATA_SUPPORT_LOOKAHEAD 0x0040 #define ATA_SUPPORT_RELEASEIRQ 0x0080 #define ATA_SUPPORT_SERVICEIRQ 0x0100 #define ATA_SUPPORT_RESET 0x0200 #define ATA_SUPPORT_PROTECTED 0x0400 #define ATA_SUPPORT_WRITEBUFFER 0x1000 #define ATA_SUPPORT_READBUFFER 0x2000 #define ATA_SUPPORT_NOP 0x4000 /*083/086*/ u_int16_t command2; #define ATA_SUPPORT_MICROCODE 0x0001 #define ATA_SUPPORT_QUEUED 0x0002 #define ATA_SUPPORT_CFA 0x0004 #define ATA_SUPPORT_APM 0x0008 #define ATA_SUPPORT_NOTIFY 0x0010 #define ATA_SUPPORT_STANDBY 0x0020 #define ATA_SUPPORT_SPINUP 0x0040 #define ATA_SUPPORT_MAXSECURITY 0x0100 #define ATA_SUPPORT_AUTOACOUSTIC 0x0200 #define ATA_SUPPORT_ADDRESS48 0x0400 #define ATA_SUPPORT_OVERLAY 0x0800 #define ATA_SUPPORT_FLUSHCACHE 0x1000 #define ATA_SUPPORT_FLUSHCACHE48 0x2000 /*084/087*/ u_int16_t extension; #define ATA_SUPPORT_SMARTLOG 0x0001 #define ATA_SUPPORT_SMARTTEST 0x0002 #define ATA_SUPPORT_MEDIASN 0x0004 #define ATA_SUPPORT_MEDIAPASS 0x0008 #define ATA_SUPPORT_STREAMING 0x0010 #define ATA_SUPPORT_GENLOG 0x0020 #define ATA_SUPPORT_WRITEDMAFUAEXT 0x0040 #define ATA_SUPPORT_WRITEDMAQFUAEXT 0x0080 #define ATA_SUPPORT_64BITWWN 0x0100 #define ATA_SUPPORT_UNLOAD 0x2000 } __packed support, enabled; /*088*/ u_int16_t udmamodes; /* UltraDMA modes */ /*089*/ u_int16_t erase_time; /* time req'd in 2min units */ /*090*/ u_int16_t enhanced_erase_time; /* time req'd in 2min units */ /*091*/ u_int16_t apm_value; /*092*/ u_int16_t master_passwd_revision; /* password revision code */ /*093*/ u_int16_t hwres; #define ATA_CABLE_ID 0x2000 /*094*/ u_int16_t acoustic; #define ATA_ACOUSTIC_CURRENT(x) ((x) & 0x00ff) #define ATA_ACOUSTIC_VENDOR(x) (((x) & 0xff00) >> 8) /*095*/ u_int16_t stream_min_req_size; /*096*/ u_int16_t stream_transfer_time; /*097*/ u_int16_t stream_access_latency; /*098*/ u_int32_t stream_granularity; /*100*/ u_int16_t lba_size48_1; u_int16_t lba_size48_2; u_int16_t lba_size48_3; u_int16_t lba_size48_4; u_int16_t reserved104; /*105*/ u_int16_t max_dsm_blocks; /*106*/ u_int16_t pss; #define ATA_PSS_LSPPS 0x000F #define ATA_PSS_LSSABOVE512 0x1000 #define ATA_PSS_MULTLS 0x2000 #define ATA_PSS_VALID_MASK 0xC000 #define ATA_PSS_VALID_VALUE 0x4000 /*107*/ u_int16_t isd; /*108*/ u_int16_t wwn[4]; u_int16_t reserved112[5]; /*117*/ u_int16_t lss_1; /*118*/ u_int16_t lss_2; /*119*/ u_int16_t support2; #define ATA_SUPPORT_WRITEREADVERIFY 0x0002 #define ATA_SUPPORT_WRITEUNCORREXT 0x0004 #define ATA_SUPPORT_RWLOGDMAEXT 0x0008 #define ATA_SUPPORT_MICROCODE3 0x0010 #define ATA_SUPPORT_FREEFALL 0x0020 /*120*/ u_int16_t enabled2; u_int16_t reserved121[6]; /*127*/ u_int16_t removable_status; /*128*/ u_int16_t security_status; #define ATA_SECURITY_LEVEL 0x0100 /* 0: high, 1: maximum */ #define ATA_SECURITY_ENH_SUPP 0x0020 /* enhanced erase supported */ #define ATA_SECURITY_COUNT_EXP 0x0010 /* count expired */ #define ATA_SECURITY_FROZEN 0x0008 /* security config is frozen */ #define ATA_SECURITY_LOCKED 0x0004 /* drive is locked */ #define ATA_SECURITY_ENABLED 0x0002 /* ATA Security is enabled */ #define ATA_SECURITY_SUPPORTED 0x0001 /* ATA Security is supported */ u_int16_t reserved129[31]; /*160*/ u_int16_t cfa_powermode1; u_int16_t reserved161; /*162*/ u_int16_t cfa_kms_support; /*163*/ u_int16_t cfa_trueide_modes; /*164*/ u_int16_t cfa_memory_modes; u_int16_t reserved165[4]; /*169*/ u_int16_t support_dsm; #define ATA_SUPPORT_DSM_TRIM 0x0001 u_int16_t reserved170[6]; /*176*/ u_int8_t media_serial[60]; /*206*/ u_int16_t sct; u_int16_t reserved206[2]; /*209*/ u_int16_t lsalign; /*210*/ u_int16_t wrv_sectors_m3_1; u_int16_t wrv_sectors_m3_2; /*212*/ u_int16_t wrv_sectors_m2_1; u_int16_t wrv_sectors_m2_2; /*214*/ u_int16_t nv_cache_caps; /*215*/ u_int16_t nv_cache_size_1; u_int16_t nv_cache_size_2; /*217*/ u_int16_t media_rotation_rate; +#define ATA_RATE_NOT_REPORTED 0x0000 +#define ATA_RATE_NON_ROTATING 0x0001 u_int16_t reserved218; /*219*/ u_int16_t nv_cache_opt; /*220*/ u_int16_t wrv_mode; u_int16_t reserved221; /*222*/ u_int16_t transport_major; /*223*/ u_int16_t transport_minor; u_int16_t reserved224[31]; /*255*/ u_int16_t integrity; } __packed; /* ATA Dataset Management */ #define ATA_DSM_BLK_SIZE 512 #define ATA_DSM_BLK_RANGES 64 #define ATA_DSM_RANGE_SIZE 8 #define ATA_DSM_RANGE_MAX 65535 /* * ATA Device Register * * bit 7 Obsolete (was 1 in early ATA specs) * bit 6 Sets LBA/CHS mode. 1=LBA, 0=CHS * bit 5 Obsolete (was 1 in early ATA specs) * bit 4 1 = Slave Drive, 0 = Master Drive * bit 3-0 In LBA mode, 27-24 of address. In CHS mode, head number */ #define ATA_DEV_MASTER 0x00 #define ATA_DEV_SLAVE 0x10 #define ATA_DEV_LBA 0x40 /* ATA limits */ #define ATA_MAX_28BIT_LBA 268435455UL /* ATA Status Register */ #define ATA_STATUS_ERROR 0x01 #define ATA_STATUS_DEVICE_FAULT 0x20 /* ATA Error Register */ #define ATA_ERROR_ABORT 0x04 #define ATA_ERROR_ID_NOT_FOUND 0x10 /* ATA HPA Features */ #define ATA_HPA_FEAT_MAX_ADDR 0x00 #define ATA_HPA_FEAT_SET_PWD 0x01 #define ATA_HPA_FEAT_LOCK 0x02 #define ATA_HPA_FEAT_UNLOCK 0x03 #define ATA_HPA_FEAT_FREEZE 0x04 /* ATA transfer modes */ #define ATA_MODE_MASK 0x0f #define ATA_DMA_MASK 0xf0 #define ATA_PIO 0x00 #define ATA_PIO0 0x08 #define ATA_PIO1 0x09 #define ATA_PIO2 0x0a #define ATA_PIO3 0x0b #define ATA_PIO4 0x0c #define ATA_PIO_MAX 0x0f #define ATA_DMA 0x10 #define ATA_WDMA0 0x20 #define ATA_WDMA1 0x21 #define ATA_WDMA2 0x22 #define ATA_UDMA0 0x40 #define ATA_UDMA1 0x41 #define ATA_UDMA2 0x42 #define ATA_UDMA3 0x43 #define ATA_UDMA4 0x44 #define ATA_UDMA5 0x45 #define ATA_UDMA6 0x46 #define ATA_SA150 0x47 #define ATA_SA300 0x48 #define ATA_DMA_MAX 0x4f /* ATA commands */ #define ATA_NOP 0x00 /* NOP */ #define ATA_NF_FLUSHQUEUE 0x00 /* flush queued cmd's */ #define ATA_NF_AUTOPOLL 0x01 /* start autopoll function */ #define ATA_DATA_SET_MANAGEMENT 0x06 #define ATA_DSM_TRIM 0x01 #define ATA_DEVICE_RESET 0x08 /* reset device */ #define ATA_READ 0x20 /* read */ #define ATA_READ48 0x24 /* read 48bit LBA */ #define ATA_READ_DMA48 0x25 /* read DMA 48bit LBA */ #define ATA_READ_DMA_QUEUED48 0x26 /* read DMA QUEUED 48bit LBA */ #define ATA_READ_NATIVE_MAX_ADDRESS48 0x27 /* read native max addr 48bit */ #define ATA_READ_MUL48 0x29 /* read multi 48bit LBA */ #define ATA_READ_STREAM_DMA48 0x2a /* read DMA stream 48bit LBA */ #define ATA_READ_LOG_EXT 0x2f /* read log ext - PIO Data-In */ #define ATA_READ_STREAM48 0x2b /* read stream 48bit LBA */ #define ATA_WRITE 0x30 /* write */ #define ATA_WRITE48 0x34 /* write 48bit LBA */ #define ATA_WRITE_DMA48 0x35 /* write DMA 48bit LBA */ #define ATA_WRITE_DMA_QUEUED48 0x36 /* write DMA QUEUED 48bit LBA*/ #define ATA_SET_MAX_ADDRESS48 0x37 /* set max address 48bit */ #define ATA_WRITE_MUL48 0x39 /* write multi 48bit LBA */ #define ATA_WRITE_STREAM_DMA48 0x3a #define ATA_WRITE_STREAM48 0x3b #define ATA_WRITE_DMA_FUA48 0x3d #define ATA_WRITE_DMA_QUEUED_FUA48 0x3e #define ATA_WRITE_LOG_EXT 0x3f #define ATA_READ_VERIFY 0x40 #define ATA_READ_VERIFY48 0x42 #define ATA_READ_LOG_DMA_EXT 0x47 /* read log DMA ext - PIO Data-In */ #define ATA_READ_FPDMA_QUEUED 0x60 /* read DMA NCQ */ #define ATA_WRITE_FPDMA_QUEUED 0x61 /* write DMA NCQ */ #define ATA_SEND_FPDMA_QUEUED 0x64 /* send DMA NCQ */ #define ATA_RECV_FPDMA_QUEUED 0x65 /* recieve DMA NCQ */ #define ATA_SEP_ATTN 0x67 /* SEP request */ #define ATA_SEEK 0x70 /* seek */ #define ATA_PACKET_CMD 0xa0 /* packet command */ #define ATA_ATAPI_IDENTIFY 0xa1 /* get ATAPI params*/ #define ATA_SERVICE 0xa2 /* service command */ #define ATA_SMART_CMD 0xb0 /* SMART command */ #define ATA_CFA_ERASE 0xc0 /* CFA erase */ #define ATA_READ_MUL 0xc4 /* read multi */ #define ATA_WRITE_MUL 0xc5 /* write multi */ #define ATA_SET_MULTI 0xc6 /* set multi size */ #define ATA_READ_DMA_QUEUED 0xc7 /* read DMA QUEUED */ #define ATA_READ_DMA 0xc8 /* read DMA */ #define ATA_WRITE_DMA 0xca /* write DMA */ #define ATA_WRITE_DMA_QUEUED 0xcc /* write DMA QUEUED */ #define ATA_WRITE_MUL_FUA48 0xce #define ATA_STANDBY_IMMEDIATE 0xe0 /* standby immediate */ #define ATA_IDLE_IMMEDIATE 0xe1 /* idle immediate */ #define ATA_STANDBY_CMD 0xe2 /* standby */ #define ATA_IDLE_CMD 0xe3 /* idle */ #define ATA_READ_BUFFER 0xe4 /* read buffer */ #define ATA_READ_PM 0xe4 /* read portmultiplier */ #define ATA_SLEEP 0xe6 /* sleep */ #define ATA_FLUSHCACHE 0xe7 /* flush cache to disk */ #define ATA_WRITE_PM 0xe8 /* write portmultiplier */ #define ATA_FLUSHCACHE48 0xea /* flush cache to disk */ #define ATA_ATA_IDENTIFY 0xec /* get ATA params */ #define ATA_SETFEATURES 0xef /* features command */ #define ATA_SF_SETXFER 0x03 /* set transfer mode */ #define ATA_SF_ENAB_WCACHE 0x02 /* enable write cache */ #define ATA_SF_DIS_WCACHE 0x82 /* disable write cache */ #define ATA_SF_ENAB_PUIS 0x06 /* enable PUIS */ #define ATA_SF_DIS_PUIS 0x86 /* disable PUIS */ #define ATA_SF_PUIS_SPINUP 0x07 /* PUIS spin-up */ #define ATA_SF_ENAB_RCACHE 0xaa /* enable readahead cache */ #define ATA_SF_DIS_RCACHE 0x55 /* disable readahead cache */ #define ATA_SF_ENAB_RELIRQ 0x5d /* enable release interrupt */ #define ATA_SF_DIS_RELIRQ 0xdd /* disable release interrupt */ #define ATA_SF_ENAB_SRVIRQ 0x5e /* enable service interrupt */ #define ATA_SF_DIS_SRVIRQ 0xde /* disable service interrupt */ #define ATA_SECURITY_SET_PASSWORD 0xf1 /* set drive password */ #define ATA_SECURITY_UNLOCK 0xf2 /* unlock drive using passwd */ #define ATA_SECURITY_ERASE_PREPARE 0xf3 /* prepare to erase drive */ #define ATA_SECURITY_ERASE_UNIT 0xf4 /* erase all blocks on drive */ #define ATA_SECURITY_FREEZE_LOCK 0xf5 /* freeze security config */ #define ATA_SECURITY_DISABLE_PASSWORD 0xf6 /* disable drive password */ #define ATA_READ_NATIVE_MAX_ADDRESS 0xf8 /* read native max address */ #define ATA_SET_MAX_ADDRESS 0xf9 /* set max address */ /* ATAPI commands */ #define ATAPI_TEST_UNIT_READY 0x00 /* check if device is ready */ #define ATAPI_REZERO 0x01 /* rewind */ #define ATAPI_REQUEST_SENSE 0x03 /* get sense data */ #define ATAPI_FORMAT 0x04 /* format unit */ #define ATAPI_READ 0x08 /* read data */ #define ATAPI_WRITE 0x0a /* write data */ #define ATAPI_WEOF 0x10 /* write filemark */ #define ATAPI_WF_WRITE 0x01 #define ATAPI_SPACE 0x11 /* space command */ #define ATAPI_SP_FM 0x01 #define ATAPI_SP_EOD 0x03 #define ATAPI_INQUIRY 0x12 /* get inquiry data */ #define ATAPI_MODE_SELECT 0x15 /* mode select */ #define ATAPI_ERASE 0x19 /* erase */ #define ATAPI_MODE_SENSE 0x1a /* mode sense */ #define ATAPI_START_STOP 0x1b /* start/stop unit */ #define ATAPI_SS_LOAD 0x01 #define ATAPI_SS_RETENSION 0x02 #define ATAPI_SS_EJECT 0x04 #define ATAPI_PREVENT_ALLOW 0x1e /* media removal */ #define ATAPI_READ_FORMAT_CAPACITIES 0x23 /* get format capacities */ #define ATAPI_READ_CAPACITY 0x25 /* get volume capacity */ #define ATAPI_READ_BIG 0x28 /* read data */ #define ATAPI_WRITE_BIG 0x2a /* write data */ #define ATAPI_LOCATE 0x2b /* locate to position */ #define ATAPI_READ_POSITION 0x34 /* read position */ #define ATAPI_SYNCHRONIZE_CACHE 0x35 /* flush buf, close channel */ #define ATAPI_WRITE_BUFFER 0x3b /* write device buffer */ #define ATAPI_READ_BUFFER 0x3c /* read device buffer */ #define ATAPI_READ_SUBCHANNEL 0x42 /* get subchannel info */ #define ATAPI_READ_TOC 0x43 /* get table of contents */ #define ATAPI_PLAY_10 0x45 /* play by lba */ #define ATAPI_PLAY_MSF 0x47 /* play by MSF address */ #define ATAPI_PLAY_TRACK 0x48 /* play by track number */ #define ATAPI_PAUSE 0x4b /* pause audio operation */ #define ATAPI_READ_DISK_INFO 0x51 /* get disk info structure */ #define ATAPI_READ_TRACK_INFO 0x52 /* get track info structure */ #define ATAPI_RESERVE_TRACK 0x53 /* reserve track */ #define ATAPI_SEND_OPC_INFO 0x54 /* send OPC structurek */ #define ATAPI_MODE_SELECT_BIG 0x55 /* set device parameters */ #define ATAPI_REPAIR_TRACK 0x58 /* repair track */ #define ATAPI_READ_MASTER_CUE 0x59 /* read master CUE info */ #define ATAPI_MODE_SENSE_BIG 0x5a /* get device parameters */ #define ATAPI_CLOSE_TRACK 0x5b /* close track/session */ #define ATAPI_READ_BUFFER_CAPACITY 0x5c /* get buffer capicity */ #define ATAPI_SEND_CUE_SHEET 0x5d /* send CUE sheet */ #define ATAPI_SERVICE_ACTION_IN 0x96 /* get service data */ #define ATAPI_BLANK 0xa1 /* blank the media */ #define ATAPI_SEND_KEY 0xa3 /* send DVD key structure */ #define ATAPI_REPORT_KEY 0xa4 /* get DVD key structure */ #define ATAPI_PLAY_12 0xa5 /* play by lba */ #define ATAPI_LOAD_UNLOAD 0xa6 /* changer control command */ #define ATAPI_READ_STRUCTURE 0xad /* get DVD structure */ #define ATAPI_PLAY_CD 0xb4 /* universal play command */ #define ATAPI_SET_SPEED 0xbb /* set drive speed */ #define ATAPI_MECH_STATUS 0xbd /* get changer status */ #define ATAPI_READ_CD 0xbe /* read data */ #define ATAPI_POLL_DSC 0xff /* poll DSC status bit */ struct ata_ioc_devices { int channel; char name[2][32]; struct ata_params params[2]; }; /* pr channel ATA ioctl calls */ #define IOCATAGMAXCHANNEL _IOR('a', 1, int) #define IOCATAREINIT _IOW('a', 2, int) #define IOCATAATTACH _IOW('a', 3, int) #define IOCATADETACH _IOW('a', 4, int) #define IOCATADEVICES _IOWR('a', 5, struct ata_ioc_devices) /* ATAPI request sense structure */ struct atapi_sense { u_int8_t error; /* current or deferred errors */ #define ATA_SENSE_VALID 0x80 u_int8_t segment; /* segment number */ u_int8_t key; /* sense key */ #define ATA_SENSE_KEY_MASK 0x0f /* sense key mask */ #define ATA_SENSE_NO_SENSE 0x00 /* no specific sense key info */ #define ATA_SENSE_RECOVERED_ERROR 0x01 /* command OK, data recovered */ #define ATA_SENSE_NOT_READY 0x02 /* no access to drive */ #define ATA_SENSE_MEDIUM_ERROR 0x03 /* non-recovered data error */ #define ATA_SENSE_HARDWARE_ERROR 0x04 /* non-recoverable HW failure */ #define ATA_SENSE_ILLEGAL_REQUEST 0x05 /* invalid command param(s) */ #define ATA_SENSE_UNIT_ATTENTION 0x06 /* media changed */ #define ATA_SENSE_DATA_PROTECT 0x07 /* write protect */ #define ATA_SENSE_BLANK_CHECK 0x08 /* blank check */ #define ATA_SENSE_VENDOR_SPECIFIC 0x09 /* vendor specific skey */ #define ATA_SENSE_COPY_ABORTED 0x0a /* copy aborted */ #define ATA_SENSE_ABORTED_COMMAND 0x0b /* command aborted, try again */ #define ATA_SENSE_EQUAL 0x0c /* equal */ #define ATA_SENSE_VOLUME_OVERFLOW 0x0d /* volume overflow */ #define ATA_SENSE_MISCOMPARE 0x0e /* data dont match the medium */ #define ATA_SENSE_RESERVED 0x0f #define ATA_SENSE_ILI 0x20; #define ATA_SENSE_EOM 0x40; #define ATA_SENSE_FILEMARK 0x80; u_int32_t cmd_info; /* cmd information */ u_int8_t sense_length; /* additional sense len (n-7) */ u_int32_t cmd_specific_info; /* additional cmd spec info */ u_int8_t asc; /* additional sense code */ u_int8_t ascq; /* additional sense code qual */ u_int8_t replaceable_unit_code; /* replaceable unit code */ u_int8_t specific; /* sense key specific */ #define ATA_SENSE_SPEC_VALID 0x80 #define ATA_SENSE_SPEC_MASK 0x7f u_int8_t specific1; /* sense key specific */ u_int8_t specific2; /* sense key specific */ } __packed; struct ata_ioc_request { union { struct { u_int8_t command; u_int8_t feature; u_int64_t lba; u_int16_t count; } ata; struct { char ccb[16]; struct atapi_sense sense; } atapi; } u; caddr_t data; int count; int flags; #define ATA_CMD_CONTROL 0x01 #define ATA_CMD_READ 0x02 #define ATA_CMD_WRITE 0x04 #define ATA_CMD_ATAPI 0x08 int timeout; int error; }; struct ata_security_password { u_int16_t ctrl; #define ATA_SECURITY_PASSWORD_USER 0x0000 #define ATA_SECURITY_PASSWORD_MASTER 0x0001 #define ATA_SECURITY_ERASE_NORMAL 0x0000 #define ATA_SECURITY_ERASE_ENHANCED 0x0002 #define ATA_SECURITY_LEVEL_HIGH 0x0000 #define ATA_SECURITY_LEVEL_MAXIMUM 0x0100 u_int8_t password[32]; u_int16_t revision; u_int16_t reserved[238]; }; /* pr device ATA ioctl calls */ #define IOCATAREQUEST _IOWR('a', 100, struct ata_ioc_request) #define IOCATAGPARM _IOR('a', 101, struct ata_params) #define IOCATAGMODE _IOR('a', 102, int) #define IOCATASMODE _IOW('a', 103, int) #define IOCATAGSPINDOWN _IOR('a', 104, int) #define IOCATASSPINDOWN _IOW('a', 105, int) struct ata_ioc_raid_config { int lun; int type; #define AR_JBOD 0x0001 #define AR_SPAN 0x0002 #define AR_RAID0 0x0004 #define AR_RAID1 0x0008 #define AR_RAID01 0x0010 #define AR_RAID3 0x0020 #define AR_RAID4 0x0040 #define AR_RAID5 0x0080 int interleave; int status; #define AR_READY 1 #define AR_DEGRADED 2 #define AR_REBUILDING 4 int progress; int total_disks; int disks[16]; }; struct ata_ioc_raid_status { int lun; int type; int interleave; int status; int progress; int total_disks; struct { int state; #define AR_DISK_ONLINE 0x01 #define AR_DISK_PRESENT 0x02 #define AR_DISK_SPARE 0x04 int lun; } disks[16]; }; /* ATA RAID ioctl calls */ #define IOCATARAIDCREATE _IOWR('a', 200, struct ata_ioc_raid_config) #define IOCATARAIDDELETE _IOW('a', 201, int) #define IOCATARAIDSTATUS _IOWR('a', 202, struct ata_ioc_raid_status) #define IOCATARAIDADDSPARE _IOW('a', 203, struct ata_ioc_raid_config) #define IOCATARAIDREBUILD _IOW('a', 204, int) #endif /* _SYS_ATA_H_ */ Index: stable/10 =================================================================== --- stable/10 (revision 271237) +++ stable/10 (revision 271238) Property changes on: stable/10 ___________________________________________________________________ Modified: svn:mergeinfo ## -0,0 +0,1 ## Merged /head:r256956,260713