Page MenuHomeFreeBSD

mav (Alexander Motin)
User

Projects

User Details

User Since
Jun 3 2014, 6:27 PM (338 w, 5 d)

Recent Activity

Yesterday

mav committed rS368138: Remove alignment requirements for KVA buffer mapping..
Remove alignment requirements for KVA buffer mapping.
Sun, Nov 29, 1:30 AM
mav committed rS368136: Remove aligment requirements for passthrough buffer..
Remove aligment requirements for passthrough buffer.
Sun, Nov 29, 12:57 AM
mav committed rS368134: Remove alignment requirements for KVA buffer mapping..
Remove alignment requirements for KVA buffer mapping.
Sun, Nov 29, 12:49 AM
mav committed rS368133: Mark inline functions static..
Mark inline functions static.
Sun, Nov 29, 12:35 AM
mav committed rS368132: Increase nvme(4) maximum transfer size from 1MB to 2MB..
Increase nvme(4) maximum transfer size from 1MB to 2MB.
Sun, Nov 29, 12:20 AM

Fri, Nov 27

mav committed rS368105: Some code reorganization..
Some code reorganization.
Fri, Nov 27, 3:50 PM
mav committed rS368104: MFC r367862: Microoptimize cam_num_doneqs math in xpt_done()..
MFC r367862: Microoptimize cam_num_doneqs math in xpt_done().
Fri, Nov 27, 1:25 PM
mav committed rS368091: Restore break statement lost in r367857..
Restore break statement lost in r367857.
Fri, Nov 27, 5:14 AM

Thu, Nov 26

mav committed rS368079: More cleanup in response queue and reset code..
More cleanup in response queue and reset code.
Thu, Nov 26, 6:47 PM
mav committed rS368044: Some minor FCoE bits I had lying around..
Some minor FCoE bits I had lying around.
Thu, Nov 26, 2:15 AM
mav committed rS368043: Remove some more dead code from pre-24xx..
Remove some more dead code from pre-24xx.
Thu, Nov 26, 2:00 AM

Tue, Nov 24

mav committed rS368001: Remove more legacy of parallel SCSI..
Remove more legacy of parallel SCSI.
Tue, Nov 24, 10:43 PM
mav committed rS367991: Update RQSFLAG_* definitions..
Update RQSFLAG_* definitions.
Tue, Nov 24, 5:38 PM
mav committed rS367986: Remove concept of mbox_sleep_ok..
Remove concept of mbox_sleep_ok.
Tue, Nov 24, 3:32 PM
mav committed rS367985: Remove unneeded locking around xpt_bus_[de]register()..
Remove unneeded locking around xpt_bus_[de]register().
Tue, Nov 24, 2:06 PM
mav committed rS367982: MFC r367044: Introduce support of SCSI Command Priority..
MFC r367044: Introduce support of SCSI Command Priority.
Tue, Nov 24, 1:17 PM
mav committed rS367979: Implement request queue overflow protection..
Implement request queue overflow protection.
Tue, Nov 24, 4:17 AM
mav committed rS367974: MFC r366922, r367915, r367916:.
MFC r366922, r367915, r367916:
Tue, Nov 24, 12:31 AM

Mon, Nov 23

mav accepted D27330: In g_newprovider_event() return if G_P_WITHER is set.

Looks good. Few places that call it also check for G_PF_WITHER, but they are using g_post_event() for it, that opens a race window.

Mon, Nov 23, 2:42 PM
mav accepted D27329: Free data allocated by getaddrinfo(3).
Mon, Nov 23, 2:37 PM
mav accepted D27328: Fix memory leak by freein the 'addr'.
Mon, Nov 23, 2:32 PM

Sun, Nov 22

mav committed rS367937: Fix debug build after 367926..
Fix debug build after 367926.
Sun, Nov 22, 4:13 PM
mav committed rS367929: Fix build after 367926..
Fix build after 367926.
Sun, Nov 22, 5:43 AM
mav committed rS367926: Make handlers and atpds overflows unlikely..
Make handlers and atpds overflows unlikely.
Sun, Nov 22, 4:30 AM
mav committed rS367925: Do not parent all busdma tags to the payload tag..
Do not parent all busdma tags to the payload tag.
Sun, Nov 22, 4:10 AM
mav committed rS367924: Remove remnants of execthrottle and maxalloc parameters..
Remove remnants of execthrottle and maxalloc parameters.
Sun, Nov 22, 2:52 AM

Sat, Nov 21

mav committed rS367916: Unlucky change....
Unlucky change...
Sat, Nov 21, 12:54 AM
mav committed rS367915: Fix stupid math mistake in r366922..
Fix stupid math mistake in r366922.
Sat, Nov 21, 12:52 AM
mav committed rS367914: MFC r366953: Negotiate iSCSIProtocolLevel of 2 (RFC 7144) in initiator..
MFC r366953: Negotiate iSCSIProtocolLevel of 2 (RFC 7144) in initiator.
Sat, Nov 21, 12:41 AM

Fri, Nov 20

mav committed rS367909: Increase queue depths from 1024/256 to 8192/1024 IOCBs..
Increase queue depths from 1024/256 to 8192/1024 IOCBs.
Fri, Nov 20, 7:36 PM
mav committed rS367906: Cleanup DMA handling..
Cleanup DMA handling.
Fri, Nov 20, 7:09 PM
mav committed rS367862: Microoptimize cam_num_doneqs math in xpt_done()..
Microoptimize cam_num_doneqs math in xpt_done().
Fri, Nov 20, 5:46 AM
mav committed rS367858: Fix r367857 build without ISP_TARGET_MODE..
Fix r367857 build without ISP_TARGET_MODE.
Fri, Nov 20, 2:04 AM
mav committed rS367857: Remove parallel SCSI and 1/2Gb FC support from isp(4)..
Remove parallel SCSI and 1/2Gb FC support from isp(4).
Fri, Nov 20, 1:16 AM
mav committed rS367856: MFC r367630: Improve nvmecontrol error reporting..
MFC r367630: Improve nvmecontrol error reporting.
Fri, Nov 20, 12:31 AM
mav committed rS367855: MFC r366506, r366511:.
MFC r366506, r366511:
Fri, Nov 20, 12:30 AM

Thu, Nov 19

mav committed rS367825: MFC r367625: Fix panic if NVMe is detached before the intrhook call..
MFC r367625: Fix panic if NVMe is detached before the intrhook call.
Thu, Nov 19, 2:44 AM

Wed, Nov 18

mav accepted D27270: Remove NO_EVENTTIMERS support.

I am happy that after all the years the transition is finally done. :)

Wed, Nov 18, 3:49 PM
mav committed rS367784: Move ecmd memory allocation itto separate DMA tag..
Move ecmd memory allocation itto separate DMA tag.
Wed, Nov 18, 3:43 AM
mav committed rS367782: Remove bus_dma locking/sleeping when not needed..
Remove bus_dma locking/sleeping when not needed.
Wed, Nov 18, 2:54 AM
mav committed rS367781: Don't allocate full XCMD_SIZE (512 bytes) on stack..
Don't allocate full XCMD_SIZE (512 bytes) on stack.
Wed, Nov 18, 2:13 AM
mav committed rS367780: MFC r367600: Make CTL nicer to increased MAXPHYS..
MFC r367600: Make CTL nicer to increased MAXPHYS.
Wed, Nov 18, 2:06 AM

Tue, Nov 17

mav added inline comments to D27225: Make MAXPHYS tunable..
Tue, Nov 17, 7:01 PM
mav committed rS367768: Stop using NVME_MAX_XFER_SIZE constant..
Stop using NVME_MAX_XFER_SIZE constant.
Tue, Nov 17, 4:35 PM
mav added inline comments to D27225: Make MAXPHYS tunable..
Tue, Nov 17, 4:05 PM
mav committed rS367739: MFC r367659: Add PMRCAP printing and fix earlier CAP_HI..
MFC r367659: Add PMRCAP printing and fix earlier CAP_HI.
Tue, Nov 17, 12:36 AM
mav added inline comments to D27225: Make MAXPHYS tunable..
Tue, Nov 17, 12:15 AM

Mon, Nov 16

mav added inline comments to D27225: Make MAXPHYS tunable..
Mon, Nov 16, 5:14 PM
mav added inline comments to D27225: Make MAXPHYS tunable..
Mon, Nov 16, 3:36 PM
mav added a comment to D27225: Make MAXPHYS tunable..
In D27225#608119, @kib wrote:

I would worry about reserved KVA misalignment then. Right now, I believe, pbuf' allocated KVA is naturally aligned.

Mon, Nov 16, 2:50 PM
mav accepted D27225: Make MAXPHYS tunable..

It looks great! Thanks!

Mon, Nov 16, 2:42 PM
mav added inline comments to D27225: Make MAXPHYS tunable..
Mon, Nov 16, 3:26 AM

Sun, Nov 15

mav added a comment to D27225: Make MAXPHYS tunable..
In D27225#607878, @imp wrote:

Does ZFS need any tweaking? I know it's mildly wrong today in scheduling 1MB I/Os today, but at least GEOM will break that up so the I/Os from there on down will be legal.

Sun, Nov 15, 8:54 PM

Sat, Nov 14

mav added a comment to D26912: RFC: Disk I/O priority support.

If there is some use high priority, then it works for SATA and it is simple, but since it is absolute priority, the difference between normal and high is too big to use it without very good reason. For low priority though, which would be useful for background operations even with absolute priorities, I haven't found a working implementation so far, unless potentially NVMe. I am hoping to get some comments from ${HDD vendor} about it.

Sat, Nov 14, 11:34 PM · cam
mav committed rS367659: Add PMRCAP printing and fix earlier CAP_HI..
Add PMRCAP printing and fix earlier CAP_HI.
Sat, Nov 14, 1:45 AM

Fri, Nov 13

mav committed rS367630: Improve nvmecontrol error reporting..
Improve nvmecontrol error reporting.
Fri, Nov 13, 2:06 AM

Thu, Nov 12

mav committed rS367625: Fix panic if NVMe is detached before the intrhook call..
Fix panic if NVMe is detached before the intrhook call.
Thu, Nov 12, 8:21 PM

Wed, Nov 11

mav committed rS367600: Make CTL nicer to increased MAXPHYS..
Make CTL nicer to increased MAXPHYS.
Wed, Nov 11, 9:59 PM

Mon, Nov 9

mav committed rS367506: MFC r367261: Add icc (Isochronous Command Completion) ccb_ataio field..
MFC r367261: Add icc (Isochronous Command Completion) ccb_ataio field.
Mon, Nov 9, 1:15 AM
mav committed rS367505: MFC r367052: Enable bioq 'car limit' added at r335066 at 128 bios..
MFC r367052: Enable bioq 'car limit' added at r335066 at 128 bios.
Mon, Nov 9, 1:14 AM

Sun, Nov 8

mav committed rS367469: MFC r367041: Fix incorrect constants of target tag action..
MFC r367041: Fix incorrect constants of target tag action.
Sun, Nov 8, 12:32 AM
mav committed rS367468: MFC r367041: Fix incorrect constants of target tag action..
MFC r367041: Fix incorrect constants of target tag action.
Sun, Nov 8, 12:31 AM

Sat, Nov 7

mav committed rS367446: MFC r367022: Fix asymmetry in devstat(9) calls by GEOM..
MFC r367022: Fix asymmetry in devstat(9) calls by GEOM.
Sat, Nov 7, 1:42 PM

Thu, Nov 5

mav committed rS367359: MFC r364038: Enable hw.pci.enable_aspm tunable by default..
MFC r364038: Enable hw.pci.enable_aspm tunable by default.
Thu, Nov 5, 2:58 AM

Wed, Nov 4

mav committed rS367330: MFC r367112: Use proper variable for device path..
MFC r367112: Use proper variable for device path.
Wed, Nov 4, 2:13 PM
mav committed rS367329: MFC r367109, r367113: Print NVMe controller capabilities in verbose dmesg..
MFC r367109, r367113: Print NVMe controller capabilities in verbose dmesg.
Wed, Nov 4, 2:12 PM

Mon, Nov 2

mav added a comment to D26912: RFC: Disk I/O priority support.

Just for information, I've also experimented with isochronous NCQ priority (AKA NCQ streaming). I hoped that setting large timeout would reduce the request priority. But at least on WD Red I see no any priority effects until the timeout is reached, and I see priority increase (again with the IOPS problem) when it is. It is good to see that the feature is really working, but unfortunately I see no usage for it in this shape. I see plenty of use cases for low priority (that SATA/SAS drives don't provide), but not really for high priority (that they do, but not very efficiently). NVMe seems to have usable priority concept and some devices support it, just not sure how important is the priority for pretty fast NVMe's.

Mon, Nov 2, 2:12 PM · cam
mav committed rS367261: Add icc (Isochronous Command Completion) ccb_ataio field..
Add icc (Isochronous Command Completion) ccb_ataio field.
Mon, Nov 2, 1:02 AM

Oct 29 2020

mav added a comment to D26912: RFC: Disk I/O priority support.

I've tried opposite approach of adding LOWPRIO flag instead and using it only for background operations in few places, and marking BIOs without it high-priority in ATA/SCSI. But while testing it I've noticed that disk random IOPS drop to almost non-NCQ level on a mix of different priorities. And I am measuring the same on both WD and HGST. I don't understand what is going on there, may be I am missing something, but that is unacceptable trade-off to me. I've uploaded my present patch in case somebody wish to play, but probably won't commit it in this state.

Oct 29 2020, 5:44 PM · cam
mav updated the diff for D26912: RFC: Disk I/O priority support.
Oct 29 2020, 5:22 PM · cam
mav committed rS367124: MFC r366733: Drop unsolicited responses to the still attaching CODECs..
MFC r366733: Drop unsolicited responses to the still attaching CODECs.
Oct 29 2020, 12:30 AM

Oct 28 2020

mav added a comment to D26912: RFC: Disk I/O priority support.

Priority is working on top of tag, affecting specifically only commands tagged as SIMPLE . The ORDERED and HEAD tags still have their function as they are mandatory in their fencing semantics, while priority is a softer hint for a schduler.

Oct 28 2020, 8:26 PM · cam
mav added a comment to D26912: RFC: Disk I/O priority support.

More experiments with SATA WD REDs show that priorities there more like absolute with deadline. On WD20EFRX-68E on heavy random workload I see low-priority requests in presence of high-priority are all delayed for about a second, while on WD80EFZX-68U they are all delayed for about 5 seconds. So big difference makes me think it is unusable for differentiation of sync vs async requests, but should still be good for read/write vs scrub/initialization/etc differentiation. Unfortunately I still haven't found any capable SAS drive to check there, but considering SATL directly map one into another I suppose they should have the same (absolute) semantics.

Oct 28 2020, 8:01 PM · cam
mav committed rS367113: Fix unintentional constant rename in r367109..
Fix unintentional constant rename in r367109.
Oct 28 2020, 6:22 PM
mav committed rS367112: Use proper variable for device path..
Use proper variable for device path.
Oct 28 2020, 5:28 PM
mav committed rS367109: Print NVMe controller capabilities in verbose dmesg..
Print NVMe controller capabilities in verbose dmesg.
Oct 28 2020, 3:43 PM

Oct 26 2020

mav committed rS367052: Enable bioq 'car limit' added at r335066 at 128 bios..
Enable bioq 'car limit' added at r335066 at 128 bios.
Oct 26 2020, 4:04 AM

Oct 25 2020

mav committed rS367044: Introduce support of SCSI Command Priority..
Introduce support of SCSI Command Priority.
Oct 25 2020, 7:34 PM
mav committed rS367041: Fix incorrect constants of target tag action..
Fix incorrect constants of target tag action.
Oct 25 2020, 4:59 PM

Oct 24 2020

mav committed rS367022: Fix asymmetry in devstat(9) calls by GEOM..
Fix asymmetry in devstat(9) calls by GEOM.
Oct 24 2020, 9:07 PM
mav accepted D26627: Make use of IP_VLAN_PCP setsockopt in ping and ping6..

Looks good to me.

Oct 24 2020, 8:41 PM

Oct 22 2020

mav requested review of D26912: RFC: Disk I/O priority support.
Oct 22 2020, 9:23 PM · cam
mav committed rS366953: Negotiate iSCSIProtocolLevel of 2 (RFC 7144) in initiator..
Negotiate iSCSIProtocolLevel of 2 (RFC 7144) in initiator.
Oct 22 2020, 8:26 PM
mav committed rS366922: Pass lower 3 bits of sector_count for FPDMA commands..
Pass lower 3 bits of sector_count for FPDMA commands.
Oct 22 2020, 3:31 AM

Oct 21 2020

mav accepted D26893: ubuf is too small to hold a human readable 64 bit value.
Oct 21 2020, 5:06 PM
mav committed rS366905: MFC r366707: Use RTD3 Entry Latency value as shutdown timeout..
MFC r366707: Use RTD3 Entry Latency value as shutdown timeout.
Oct 21 2020, 12:47 AM

Oct 20 2020

mav committed rS366868: MFC r366689: Fix sbuf_finish() error code check in user-space..
MFC r366689: Fix sbuf_finish() error code check in user-space.
Oct 20 2020, 2:32 AM
mav committed rS366867: MFC r366689: Fix sbuf_finish() error code check in user-space..
MFC r366689: Fix sbuf_finish() error code check in user-space.
Oct 20 2020, 2:28 AM

Oct 19 2020

mav committed rS366860: MFC r360546, r360547 (by imp): Various improvements to this man page:.
MFC r360546, r360547 (by imp): Various improvements to this man page:
Oct 19 2020, 8:42 PM
mav committed rS366859: MFC r352671 (by imp): Size is unsigned, so remove the test entirely..
MFC r352671 (by imp): Size is unsigned, so remove the test entirely.
Oct 19 2020, 8:40 PM
mav committed rS366858: MFC r352665 (by imp):.
MFC r352665 (by imp):
Oct 19 2020, 8:39 PM
mav committed rS366856: MFC r352212 (by imp):.
MFC r352212 (by imp):
Oct 19 2020, 8:37 PM
mav accepted D26857: IOMMU quirks added.

Thanks!

Oct 19 2020, 8:22 PM
mav added a comment to D26857: IOMMU quirks added.

I have no objections, but the link specified lead me to Linux' drivers/pci/quirks.c file, where I've found the same done for 12 different Marvell HBAs. Why not do the same?

Oct 19 2020, 5:36 PM
mav accepted D26855: Fix fallout from r366811.
Oct 19 2020, 3:30 PM
mav added inline comments to D26855: Fix fallout from r366811.
Oct 19 2020, 1:52 PM

Oct 18 2020

mav added inline comments to D26842: hptmv(4): use xpt_alloc_ccb/xpt_free_ccb instead of malloc/free..
Oct 18 2020, 7:34 PM
mav added inline comments to D26844: Use UMA for CCBs..
Oct 18 2020, 7:32 PM
mav accepted D26843: sbp(4): use xpt_alloc_ccb/xpt_free_ccb instead of malloc/free.
Oct 18 2020, 7:14 PM
mav accepted D26842: hptmv(4): use xpt_alloc_ccb/xpt_free_ccb instead of malloc/free..

The core around this looks pretty weird to me, considering hpt_async() immediately calls panic(). But I don't expect troubles from this change.

Oct 18 2020, 7:08 PM