Index: stable/11/share/man/man9/taskqueue.9 =================================================================== --- stable/11/share/man/man9/taskqueue.9 (revision 323446) +++ stable/11/share/man/man9/taskqueue.9 (revision 323447) @@ -1,492 +1,504 @@ .\" -*- nroff -*- .\" .\" Copyright (c) 2000 Doug Rabson .\" .\" All rights reserved. .\" .\" This program is free software. .\" .\" 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. .\" .\" THIS SOFTWARE IS PROVIDED BY THE DEVELOPERS ``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 DEVELOPERS 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. .\" .\" $FreeBSD$ .\" -.Dd March 1, 2016 +.Dd July 30, 2017 .Dt TASKQUEUE 9 .Os .Sh NAME .Nm taskqueue .Nd asynchronous task execution .Sh SYNOPSIS .In sys/param.h .In sys/kernel.h .In sys/malloc.h .In sys/queue.h .In sys/taskqueue.h .Bd -literal typedef void (*task_fn_t)(void *context, int pending); typedef void (*taskqueue_enqueue_fn)(void *context); struct task { STAILQ_ENTRY(task) ta_link; /* link for queue */ u_short ta_pending; /* count times queued */ u_short ta_priority; /* priority of task in queue */ task_fn_t ta_func; /* task handler */ void *ta_context; /* argument for handler */ }; enum taskqueue_callback_type { TASKQUEUE_CALLBACK_TYPE_INIT, TASKQUEUE_CALLBACK_TYPE_SHUTDOWN, }; typedef void (*taskqueue_callback_fn)(void *context); struct timeout_task; .Ed .Ft struct taskqueue * .Fn taskqueue_create "const char *name" "int mflags" "taskqueue_enqueue_fn enqueue" "void *context" .Ft struct taskqueue * .Fn taskqueue_create_fast "const char *name" "int mflags" "taskqueue_enqueue_fn enqueue" "void *context" .Ft int .Fn taskqueue_start_threads "struct taskqueue **tqp" "int count" "int pri" "const char *name" "..." .Ft int .Fo taskqueue_start_threads_pinned .Fa "struct taskqueue **tqp" "int count" "int pri" "int cpu_id" .Fa "const char *name" "..." .Fc .Ft void .Fn taskqueue_set_callback "struct taskqueue *queue" "enum taskqueue_callback_type cb_type" "taskqueue_callback_fn callback" "void *context" .Ft void .Fn taskqueue_free "struct taskqueue *queue" .Ft int .Fn taskqueue_enqueue "struct taskqueue *queue" "struct task *task" .Ft int .Fn taskqueue_enqueue_timeout "struct taskqueue *queue" "struct timeout_task *timeout_task" "int ticks" .Ft int +.Fn taskqueue_enqueue_timeout_sbt "struct taskqueue *queue" "struct timeout_task *timeout_task" "sbintime_t sbt" "sbintime_t pr" "int flags" +.Ft int .Fn taskqueue_cancel "struct taskqueue *queue" "struct task *task" "u_int *pendp" .Ft int .Fn taskqueue_cancel_timeout "struct taskqueue *queue" "struct timeout_task *timeout_task" "u_int *pendp" .Ft void .Fn taskqueue_drain "struct taskqueue *queue" "struct task *task" .Ft void .Fn taskqueue_drain_timeout "struct taskqueue *queue" "struct timeout_task *timeout_task" .Ft void .Fn taskqueue_drain_all "struct taskqueue *queue" .Ft void .Fn taskqueue_block "struct taskqueue *queue" .Ft void .Fn taskqueue_unblock "struct taskqueue *queue" .Ft int .Fn taskqueue_member "struct taskqueue *queue" "struct thread *td" .Ft void .Fn taskqueue_run "struct taskqueue *queue" .Fn TASK_INIT "struct task *task" "int priority" "task_fn_t func" "void *context" .Fn TASK_INITIALIZER "int priority" "task_fn_t func" "void *context" .Fn TASKQUEUE_DECLARE "name" .Fn TASKQUEUE_DEFINE "name" "taskqueue_enqueue_fn enqueue" "void *context" "init" .Fn TASKQUEUE_FAST_DEFINE "name" "taskqueue_enqueue_fn enqueue" "void *context" "init" .Fn TASKQUEUE_DEFINE_THREAD "name" .Fn TASKQUEUE_FAST_DEFINE_THREAD "name" .Fn TIMEOUT_TASK_INIT "struct taskqueue *queue" "struct timeout_task *timeout_task" "int priority" "task_fn_t func" "void *context" .Sh DESCRIPTION These functions provide a simple interface for asynchronous execution of code. .Pp The function .Fn taskqueue_create is used to create new queues. The arguments to .Fn taskqueue_create include a name that should be unique, a set of .Xr malloc 9 flags that specify whether the call to .Fn malloc is allowed to sleep, a function that is called from .Fn taskqueue_enqueue when a task is added to the queue, and a pointer to the memory location where the identity of the thread that services the queue is recorded. .\" XXX The rest of the sentence gets lots in relation to the first part. The function called from .Fn taskqueue_enqueue must arrange for the queue to be processed (for instance by scheduling a software interrupt or waking a kernel thread). The memory location where the thread identity is recorded is used to signal the service thread(s) to terminate--when this value is set to zero and the thread is signaled it will terminate. If the queue is intended for use in fast interrupt handlers .Fn taskqueue_create_fast should be used in place of .Fn taskqueue_create . .Pp The function .Fn taskqueue_free should be used to free the memory used by the queue. Any tasks that are on the queue will be executed at this time after which the thread servicing the queue will be signaled that it should exit. .Pp Once a taskqueue has been created, its threads should be started using .Fn taskqueue_start_threads or .Fn taskqueue_start_threads_pinned . .Fn taskqueue_start_threads_pinned takes a .Va cpu_id argument which will cause the threads which are started for the taskqueue to be pinned to run on the given CPU. Callbacks may optionally be registered using .Fn taskqueue_set_callback . Currently, callbacks may be registered for the following purposes: .Bl -tag -width TASKQUEUE_CALLBACK_TYPE_SHUTDOWN .It Dv TASKQUEUE_CALLBACK_TYPE_INIT This callback is called by every thread in the taskqueue, before it executes any tasks. This callback must be set before the taskqueue's threads are started. .It Dv TASKQUEUE_CALLBACK_TYPE_SHUTDOWN This callback is called by every thread in the taskqueue, after it executes its last task. This callback will always be called before the taskqueue structure is reclaimed. .El .Pp To add a task to the list of tasks queued on a taskqueue, call .Fn taskqueue_enqueue with pointers to the queue and task. If the task's .Va ta_pending field is non-zero, then it is simply incremented to reflect the number of times the task was enqueued, up to a cap of USHRT_MAX. Otherwise, the task is added to the list before the first task which has a lower .Va ta_priority value or at the end of the list if no tasks have a lower priority. Enqueueing a task does not perform any memory allocation which makes it suitable for calling from an interrupt handler. This function will return .Er EPIPE if the queue is being freed. .Pp When a task is executed, first it is removed from the queue, the value of .Va ta_pending is recorded and then the field is zeroed. The function .Va ta_func from the task structure is called with the value of the field .Va ta_context as its first argument and the value of .Va ta_pending as its second argument. After the function .Va ta_func returns, .Xr wakeup 9 is called on the task pointer passed to .Fn taskqueue_enqueue . .Pp The .Fn taskqueue_enqueue_timeout -is used to schedule the enqueue after the specified amount of +function is used to schedule the enqueue after the specified number of .Va ticks . +The +.Fn taskqueue_enqueue_timeout_sbt +function provides finer control over the scheduling based on +.Va sbt , +.Va pr , +and +.Va flags , +as detailed in +.Xr timeout 9 . Only non-fast task queues can be used for .Va timeout_task scheduling. If the .Va ticks argument is negative, the already scheduled enqueueing is not re-scheduled. Otherwise, the task is scheduled for enqueueing in the future, after the absolute value of .Va ticks is passed. This function returns -1 if the task is being drained. Otherwise, the number of pending calls is returned. .Pp The .Fn taskqueue_cancel function is used to cancel a task. The .Va ta_pending count is cleared, and the old value returned in the reference parameter .Fa pendp , if it is .Pf non- Dv NULL . If the task is currently running, .Dv EBUSY is returned, otherwise 0. To implement a blocking .Fn taskqueue_cancel that waits for a running task to finish, it could look like: .Bd -literal -offset indent while (taskqueue_cancel(tq, task, NULL) != 0) taskqueue_drain(tq, task); .Ed .Pp Note that, as with .Fn taskqueue_drain , the caller is responsible for ensuring that the task is not re-enqueued after being canceled. .Pp Similarly, the .Fn taskqueue_cancel_timeout function is used to cancel the scheduled task execution. .Pp The .Fn taskqueue_drain function is used to wait for the task to finish, and the .Fn taskqueue_drain_timeout function is used to wait for the scheduled task to finish. There is no guarantee that the task will not be enqueued after call to .Fn taskqueue_drain . If the caller wants to put the task into a known state, then before calling .Fn taskqueue_drain the caller should use out-of-band means to ensure that the task would not be enqueued. For example, if the task is enqueued by an interrupt filter, then the interrupt could be disabled. .Pp The .Fn taskqueue_drain_all function is used to wait for all pending and running tasks that are enqueued on the taskqueue to finish. Tasks posted to the taskqueue after .Fn taskqueue_drain_all begins processing, including pending enqueues scheduled by a previous call to .Fn taskqueue_enqueue_timeout , do not extend the wait time of .Fn taskqueue_drain_all and may complete after .Fn taskqueue_drain_all returns. .Pp The .Fn taskqueue_block function blocks the taskqueue. It prevents any enqueued but not running tasks from being executed. Future calls to .Fn taskqueue_enqueue will enqueue tasks, but the tasks will not be run until .Fn taskqueue_unblock is called. Please note that .Fn taskqueue_block does not wait for any currently running tasks to finish. Thus, the .Fn taskqueue_block does not provide a guarantee that .Fn taskqueue_run is not running after .Fn taskqueue_block returns, but it does provide a guarantee that .Fn taskqueue_run will not be called again until .Fn taskqueue_unblock is called. If the caller requires a guarantee that .Fn taskqueue_run is not running, then this must be arranged by the caller. Note that if .Fn taskqueue_drain is called on a task that is enqueued on a taskqueue that is blocked by .Fn taskqueue_block , then .Fn taskqueue_drain can not return until the taskqueue is unblocked. This can result in a deadlock if the thread blocked in .Fn taskqueue_drain is the thread that is supposed to call .Fn taskqueue_unblock . Thus, use of .Fn taskqueue_drain after .Fn taskqueue_block is discouraged, because the state of the task can not be known in advance. The same caveat applies to .Fn taskqueue_drain_all . .Pp The .Fn taskqueue_unblock function unblocks the previously blocked taskqueue. All enqueued tasks can be run after this call. .Pp The .Fn taskqueue_member function returns .No 1 if the given thread .Fa td is part of the given taskqueue .Fa queue and .No 0 otherwise. .Pp The .Fn taskqueue_run function will run all pending tasks in the specified .Fa queue . Normally this function is only used internally. .Pp A convenience macro, .Fn TASK_INIT "task" "priority" "func" "context" is provided to initialise a .Va task structure. The .Fn TASK_INITIALIZER macro generates an initializer for a task structure. A macro .Fn TIMEOUT_TASK_INIT "queue" "timeout_task" "priority" "func" "context" initializes the .Va timeout_task structure. The values of .Va priority , .Va func , and .Va context are simply copied into the task structure fields and the .Va ta_pending field is cleared. .Pp Five macros .Fn TASKQUEUE_DECLARE "name" , .Fn TASKQUEUE_DEFINE "name" "enqueue" "context" "init" , .Fn TASKQUEUE_FAST_DEFINE "name" "enqueue" "context" "init" , and .Fn TASKQUEUE_DEFINE_THREAD "name" .Fn TASKQUEUE_FAST_DEFINE_THREAD "name" are used to declare a reference to a global queue, to define the implementation of the queue, and declare a queue that uses its own thread. The .Fn TASKQUEUE_DEFINE macro arranges to call .Fn taskqueue_create with the values of its .Va name , .Va enqueue and .Va context arguments during system initialisation. After calling .Fn taskqueue_create , the .Va init argument to the macro is executed as a C statement, allowing any further initialisation to be performed (such as registering an interrupt handler etc.) .Pp The .Fn TASKQUEUE_DEFINE_THREAD macro defines a new taskqueue with its own kernel thread to serve tasks. The variable .Vt struct taskqueue *taskqueue_name is used to enqueue tasks onto the queue. .Pp .Fn TASKQUEUE_FAST_DEFINE and .Fn TASKQUEUE_FAST_DEFINE_THREAD act just like .Fn TASKQUEUE_DEFINE and .Fn TASKQUEUE_DEFINE_THREAD respectively but taskqueue is created with .Fn taskqueue_create_fast . .Ss Predefined Task Queues The system provides four global taskqueues, .Va taskqueue_fast , .Va taskqueue_swi , .Va taskqueue_swi_giant , and .Va taskqueue_thread . The .Va taskqueue_fast queue is for swi handlers dispatched from fast interrupt handlers, where sleep mutexes cannot be used. The swi taskqueues are run via a software interrupt mechanism. The .Va taskqueue_swi queue runs without the protection of the .Va Giant kernel lock, and the .Va taskqueue_swi_giant queue runs with the protection of the .Va Giant kernel lock. The thread taskqueue .Va taskqueue_thread runs in a kernel thread context, and tasks run from this thread do not run under the .Va Giant kernel lock. If the caller wants to run under .Va Giant , he should explicitly acquire and release .Va Giant in his taskqueue handler routine. .Pp To use these queues, call .Fn taskqueue_enqueue with the value of the global taskqueue variable for the queue you wish to use. .Pp The software interrupt queues can be used, for instance, for implementing interrupt handlers which must perform a significant amount of processing in the handler. The hardware interrupt handler would perform minimal processing of the interrupt and then enqueue a task to finish the work. This reduces to a minimum the amount of time spent with interrupts disabled. .Pp The thread queue can be used, for instance, by interrupt level routines that need to call kernel functions that do things that can only be done from a thread context. (e.g., call malloc with the M_WAITOK flag.) .Pp Note that tasks queued on shared taskqueues such as .Va taskqueue_swi may be delayed an indeterminate amount of time before execution. If queueing delays cannot be tolerated then a private taskqueue should be created with a dedicated processing thread. .Sh SEE ALSO .Xr ithread 9 , .Xr kthread 9 , .Xr swi 9 +.Xr timeout 9 .Sh HISTORY This interface first appeared in .Fx 5.0 . There is a similar facility called work_queue in the Linux kernel. .Sh AUTHORS This manual page was written by .An Doug Rabson . Index: stable/11/sys/amd64/amd64/efirt.c =================================================================== --- stable/11/sys/amd64/amd64/efirt.c (revision 323446) +++ stable/11/sys/amd64/amd64/efirt.c (revision 323447) @@ -1,608 +1,609 @@ /*- * Copyright (c) 2004 Marcel Moolenaar * Copyright (c) 2001 Doug Rabson * Copyright (c) 2016 The FreeBSD Foundation * All rights reserved. * * Portions of this software were developed by Konstantin Belousov * under sponsorship from the FreeBSD Foundation. * * 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. * * 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 #include #include #include #include #include #include #include #include #include #include #include #include +#include #include #include #include #include #include #include #include #include #include #include #include #include static struct efi_systbl *efi_systbl; static struct efi_cfgtbl *efi_cfgtbl; static struct efi_rt *efi_runtime; static int efi_status2err[25] = { 0, /* EFI_SUCCESS */ ENOEXEC, /* EFI_LOAD_ERROR */ EINVAL, /* EFI_INVALID_PARAMETER */ ENOSYS, /* EFI_UNSUPPORTED */ EMSGSIZE, /* EFI_BAD_BUFFER_SIZE */ EOVERFLOW, /* EFI_BUFFER_TOO_SMALL */ EBUSY, /* EFI_NOT_READY */ EIO, /* EFI_DEVICE_ERROR */ EROFS, /* EFI_WRITE_PROTECTED */ EAGAIN, /* EFI_OUT_OF_RESOURCES */ EIO, /* EFI_VOLUME_CORRUPTED */ ENOSPC, /* EFI_VOLUME_FULL */ ENXIO, /* EFI_NO_MEDIA */ ESTALE, /* EFI_MEDIA_CHANGED */ ENOENT, /* EFI_NOT_FOUND */ EACCES, /* EFI_ACCESS_DENIED */ ETIMEDOUT, /* EFI_NO_RESPONSE */ EADDRNOTAVAIL, /* EFI_NO_MAPPING */ ETIMEDOUT, /* EFI_TIMEOUT */ EDOOFUS, /* EFI_NOT_STARTED */ EALREADY, /* EFI_ALREADY_STARTED */ ECANCELED, /* EFI_ABORTED */ EPROTO, /* EFI_ICMP_ERROR */ EPROTO, /* EFI_TFTP_ERROR */ EPROTO /* EFI_PROTOCOL_ERROR */ }; static int efi_status_to_errno(efi_status status) { u_long code; code = status & 0x3ffffffffffffffful; return (code < nitems(efi_status2err) ? efi_status2err[code] : EDOOFUS); } static struct mtx efi_lock; static pml4_entry_t *efi_pml4; static vm_object_t obj_1t1_pt; static vm_page_t efi_pml4_page; static void efi_destroy_1t1_map(void) { vm_page_t m; if (obj_1t1_pt != NULL) { VM_OBJECT_RLOCK(obj_1t1_pt); TAILQ_FOREACH(m, &obj_1t1_pt->memq, listq) m->wire_count = 0; atomic_subtract_int(&vm_cnt.v_wire_count, obj_1t1_pt->resident_page_count); VM_OBJECT_RUNLOCK(obj_1t1_pt); vm_object_deallocate(obj_1t1_pt); } obj_1t1_pt = NULL; efi_pml4 = NULL; efi_pml4_page = NULL; } static vm_page_t efi_1t1_page(vm_pindex_t idx) { return (vm_page_grab(obj_1t1_pt, idx, VM_ALLOC_NOBUSY | VM_ALLOC_WIRED | VM_ALLOC_ZERO)); } static pt_entry_t * efi_1t1_pte(vm_offset_t va) { pml4_entry_t *pml4e; pdp_entry_t *pdpe; pd_entry_t *pde; pt_entry_t *pte; vm_page_t m; vm_pindex_t pml4_idx, pdp_idx, pd_idx; vm_paddr_t mphys; pml4_idx = pmap_pml4e_index(va); pml4e = &efi_pml4[pml4_idx]; if (*pml4e == 0) { m = efi_1t1_page(1 + pml4_idx); mphys = VM_PAGE_TO_PHYS(m); *pml4e = mphys | X86_PG_RW | X86_PG_V; } else { mphys = *pml4e & ~PAGE_MASK; } pdpe = (pdp_entry_t *)PHYS_TO_DMAP(mphys); pdp_idx = pmap_pdpe_index(va); pdpe += pdp_idx; if (*pdpe == 0) { m = efi_1t1_page(1 + NPML4EPG + (pml4_idx + 1) * (pdp_idx + 1)); mphys = VM_PAGE_TO_PHYS(m); *pdpe = mphys | X86_PG_RW | X86_PG_V; } else { mphys = *pdpe & ~PAGE_MASK; } pde = (pd_entry_t *)PHYS_TO_DMAP(mphys); pd_idx = pmap_pde_index(va); pde += pd_idx; if (*pde == 0) { m = efi_1t1_page(1 + NPML4EPG + NPML4EPG * NPDPEPG + (pml4_idx + 1) * (pdp_idx + 1) * (pd_idx + 1)); mphys = VM_PAGE_TO_PHYS(m); *pde = mphys | X86_PG_RW | X86_PG_V; } else { mphys = *pde & ~PAGE_MASK; } pte = (pt_entry_t *)PHYS_TO_DMAP(mphys); pte += pmap_pte_index(va); KASSERT(*pte == 0, ("va %#jx *pt %#jx", va, *pte)); return (pte); } static bool efi_create_1t1_map(struct efi_md *map, int ndesc, int descsz) { struct efi_md *p; pt_entry_t *pte; vm_offset_t va; uint64_t idx; int bits, i, mode; obj_1t1_pt = vm_pager_allocate(OBJT_PHYS, NULL, ptoa(1 + NPML4EPG + NPML4EPG * NPDPEPG + NPML4EPG * NPDPEPG * NPDEPG), VM_PROT_ALL, 0, NULL); VM_OBJECT_WLOCK(obj_1t1_pt); efi_pml4_page = efi_1t1_page(0); VM_OBJECT_WUNLOCK(obj_1t1_pt); efi_pml4 = (pml4_entry_t *)PHYS_TO_DMAP(VM_PAGE_TO_PHYS(efi_pml4_page)); pmap_pinit_pml4(efi_pml4_page); for (i = 0, p = map; i < ndesc; i++, p = efi_next_descriptor(p, descsz)) { if ((p->md_attr & EFI_MD_ATTR_RT) == 0) continue; if (p->md_virt != NULL) { if (bootverbose) printf("EFI Runtime entry %d is mapped\n", i); goto fail; } if ((p->md_phys & EFI_PAGE_MASK) != 0) { if (bootverbose) printf("EFI Runtime entry %d is not aligned\n", i); goto fail; } if (p->md_phys + p->md_pages * EFI_PAGE_SIZE < p->md_phys || p->md_phys + p->md_pages * EFI_PAGE_SIZE >= VM_MAXUSER_ADDRESS) { printf("EFI Runtime entry %d is not in mappable for RT:" "base %#016jx %#jx pages\n", i, (uintmax_t)p->md_phys, (uintmax_t)p->md_pages); goto fail; } if ((p->md_attr & EFI_MD_ATTR_WB) != 0) mode = VM_MEMATTR_WRITE_BACK; else if ((p->md_attr & EFI_MD_ATTR_WT) != 0) mode = VM_MEMATTR_WRITE_THROUGH; else if ((p->md_attr & EFI_MD_ATTR_WC) != 0) mode = VM_MEMATTR_WRITE_COMBINING; else if ((p->md_attr & EFI_MD_ATTR_WP) != 0) mode = VM_MEMATTR_WRITE_PROTECTED; else if ((p->md_attr & EFI_MD_ATTR_UC) != 0) mode = VM_MEMATTR_UNCACHEABLE; else { if (bootverbose) printf("EFI Runtime entry %d mapping " "attributes unsupported\n", i); mode = VM_MEMATTR_UNCACHEABLE; } bits = pmap_cache_bits(kernel_pmap, mode, FALSE) | X86_PG_RW | X86_PG_V; VM_OBJECT_WLOCK(obj_1t1_pt); for (va = p->md_phys, idx = 0; idx < p->md_pages; idx++, va += PAGE_SIZE) { pte = efi_1t1_pte(va); pte_store(pte, va | bits); } VM_OBJECT_WUNLOCK(obj_1t1_pt); } return (true); fail: efi_destroy_1t1_map(); return (false); } /* * Create an environment for the EFI runtime code call. The most * important part is creating the required 1:1 physical->virtual * mappings for the runtime segments. To do that, we manually create * page table which unmap userspace but gives correct kernel mapping. * The 1:1 mappings for runtime segments usually occupy low 4G of the * physical address map. * * The 1:1 mappings were chosen over the SetVirtualAddressMap() EFI RT * service, because there are some BIOSes which fail to correctly * relocate itself on the call, requiring both 1:1 and virtual * mapping. As result, we must provide 1:1 mapping anyway, so no * reason to bother with the virtual map, and no need to add a * complexity into loader. * * The fpu_kern_enter() call allows firmware to use FPU, as mandated * by the specification. In particular, CR0.TS bit is cleared. Also * it enters critical section, giving us neccessary protection against * context switch. * * There is no need to disable interrupts around the change of %cr3, * the kernel mappings are correct, while we only grabbed the * userspace portion of VA. Interrupts handlers must not access * userspace. Having interrupts enabled fixes the issue with * firmware/SMM long operation, which would negatively affect IPIs, * esp. TLB shootdown requests. */ static int efi_enter(void) { pmap_t curpmap; int error; if (efi_runtime == NULL) return (ENXIO); curpmap = PCPU_GET(curpmap); PMAP_LOCK(curpmap); mtx_lock(&efi_lock); error = fpu_kern_enter(curthread, NULL, FPU_KERN_NOCTX); if (error != 0) { PMAP_UNLOCK(curpmap); return (error); } /* * IPI TLB shootdown handler invltlb_pcid_handler() reloads * %cr3 from the curpmap->pm_cr3, which would disable runtime * segments mappings. Block the handler's action by setting * curpmap to impossible value. See also comment in * pmap.c:pmap_activate_sw(). */ if (pmap_pcid_enabled && !invpcid_works) PCPU_SET(curpmap, NULL); load_cr3(VM_PAGE_TO_PHYS(efi_pml4_page) | (pmap_pcid_enabled ? curpmap->pm_pcids[PCPU_GET(cpuid)].pm_pcid : 0)); /* * If PCID is enabled, the clear CR3_PCID_SAVE bit in the loaded %cr3 * causes TLB invalidation. */ if (!pmap_pcid_enabled) invltlb(); return (0); } static void efi_leave(void) { pmap_t curpmap; curpmap = &curproc->p_vmspace->vm_pmap; if (pmap_pcid_enabled && !invpcid_works) PCPU_SET(curpmap, curpmap); load_cr3(curpmap->pm_cr3 | (pmap_pcid_enabled ? curpmap->pm_pcids[PCPU_GET(cpuid)].pm_pcid : 0)); if (!pmap_pcid_enabled) invltlb(); fpu_kern_leave(curthread, NULL); mtx_unlock(&efi_lock); PMAP_UNLOCK(curpmap); } static int efi_init(void) { struct efi_map_header *efihdr; struct efi_md *map; caddr_t kmdp; size_t efisz; mtx_init(&efi_lock, "efi", NULL, MTX_DEF); if (efi_systbl_phys == 0) { if (bootverbose) printf("EFI systbl not available\n"); return (0); } efi_systbl = (struct efi_systbl *)PHYS_TO_DMAP(efi_systbl_phys); if (efi_systbl->st_hdr.th_sig != EFI_SYSTBL_SIG) { efi_systbl = NULL; if (bootverbose) printf("EFI systbl signature invalid\n"); return (0); } efi_cfgtbl = (efi_systbl->st_cfgtbl == 0) ? NULL : (struct efi_cfgtbl *)efi_systbl->st_cfgtbl; if (efi_cfgtbl == NULL) { if (bootverbose) printf("EFI config table is not present\n"); } kmdp = preload_search_by_type("elf kernel"); if (kmdp == NULL) kmdp = preload_search_by_type("elf64 kernel"); efihdr = (struct efi_map_header *)preload_search_info(kmdp, MODINFO_METADATA | MODINFOMD_EFI_MAP); if (efihdr == NULL) { if (bootverbose) printf("EFI map is not present\n"); return (0); } efisz = (sizeof(struct efi_map_header) + 0xf) & ~0xf; map = (struct efi_md *)((uint8_t *)efihdr + efisz); if (efihdr->descriptor_size == 0) return (ENOMEM); if (!efi_create_1t1_map(map, efihdr->memory_size / efihdr->descriptor_size, efihdr->descriptor_size)) { if (bootverbose) printf("EFI cannot create runtime map\n"); return (ENOMEM); } efi_runtime = (efi_systbl->st_rt == 0) ? NULL : (struct efi_rt *)efi_systbl->st_rt; if (efi_runtime == NULL) { if (bootverbose) printf("EFI runtime services table is not present\n"); efi_destroy_1t1_map(); return (ENXIO); } return (0); } static void efi_uninit(void) { efi_destroy_1t1_map(); efi_systbl = NULL; efi_cfgtbl = NULL; efi_runtime = NULL; mtx_destroy(&efi_lock); } int efi_get_table(struct uuid *uuid, void **ptr) { struct efi_cfgtbl *ct; u_long count; if (efi_cfgtbl == NULL) return (ENXIO); count = efi_systbl->st_entries; ct = efi_cfgtbl; while (count--) { if (!bcmp(&ct->ct_uuid, uuid, sizeof(*uuid))) { *ptr = (void *)PHYS_TO_DMAP(ct->ct_data); return (0); } ct++; } return (ENOENT); } int efi_get_time_locked(struct efi_tm *tm) { efi_status status; int error; - mtx_assert(&resettodr_lock, MA_OWNED); + mtx_assert(&atrtc_time_lock, MA_OWNED); error = efi_enter(); if (error != 0) return (error); status = efi_runtime->rt_gettime(tm, NULL); efi_leave(); error = efi_status_to_errno(status); return (error); } int efi_get_time(struct efi_tm *tm) { int error; if (efi_runtime == NULL) return (ENXIO); - mtx_lock(&resettodr_lock); + mtx_lock(&atrtc_time_lock); error = efi_get_time_locked(tm); - mtx_unlock(&resettodr_lock); + mtx_unlock(&atrtc_time_lock); return (error); } int efi_reset_system(void) { int error; error = efi_enter(); if (error != 0) return (error); efi_runtime->rt_reset(EFI_RESET_WARM, 0, 0, NULL); efi_leave(); return (EIO); } int efi_set_time_locked(struct efi_tm *tm) { efi_status status; int error; - mtx_assert(&resettodr_lock, MA_OWNED); + mtx_assert(&atrtc_time_lock, MA_OWNED); error = efi_enter(); if (error != 0) return (error); status = efi_runtime->rt_settime(tm); efi_leave(); error = efi_status_to_errno(status); return (error); } int efi_set_time(struct efi_tm *tm) { int error; if (efi_runtime == NULL) return (ENXIO); - mtx_lock(&resettodr_lock); + mtx_lock(&atrtc_time_lock); error = efi_set_time_locked(tm); - mtx_unlock(&resettodr_lock); + mtx_unlock(&atrtc_time_lock); return (error); } int efi_var_get(efi_char *name, struct uuid *vendor, uint32_t *attrib, size_t *datasize, void *data) { efi_status status; int error; error = efi_enter(); if (error != 0) return (error); status = efi_runtime->rt_getvar(name, vendor, attrib, datasize, data); efi_leave(); error = efi_status_to_errno(status); return (error); } int efi_var_nextname(size_t *namesize, efi_char *name, struct uuid *vendor) { efi_status status; int error; error = efi_enter(); if (error != 0) return (error); status = efi_runtime->rt_scanvar(namesize, name, vendor); efi_leave(); error = efi_status_to_errno(status); return (error); } int efi_var_set(efi_char *name, struct uuid *vendor, uint32_t attrib, size_t datasize, void *data) { efi_status status; int error; error = efi_enter(); if (error != 0) return (error); status = efi_runtime->rt_setvar(name, vendor, attrib, datasize, data); efi_leave(); error = efi_status_to_errno(status); return (error); } static int efirt_modevents(module_t m, int event, void *arg __unused) { switch (event) { case MOD_LOAD: return (efi_init()); case MOD_UNLOAD: efi_uninit(); return (0); case MOD_SHUTDOWN: return (0); default: return (EOPNOTSUPP); } } static moduledata_t efirt_moddata = { .name = "efirt", .evhand = efirt_modevents, .priv = NULL, }; DECLARE_MODULE(efirt, efirt_moddata, SI_SUB_VM_CONF, SI_ORDER_ANY); MODULE_VERSION(efirt, 1); /* XXX debug stuff */ static int efi_time_sysctl_handler(SYSCTL_HANDLER_ARGS) { struct efi_tm tm; int error, val; val = 0; error = sysctl_handle_int(oidp, &val, 0, req); if (error != 0 || req->newptr == NULL) return (error); error = efi_get_time(&tm); if (error == 0) { uprintf("EFI reports: Year %d Month %d Day %d Hour %d Min %d " "Sec %d\n", tm.tm_year, tm.tm_mon, tm.tm_mday, tm.tm_hour, tm.tm_min, tm.tm_sec); } return (error); } SYSCTL_PROC(_debug, OID_AUTO, efi_time, CTLTYPE_INT | CTLFLAG_RW, NULL, 0, efi_time_sysctl_handler, "I", ""); Index: stable/11/sys/isa/rtc.h =================================================================== --- stable/11/sys/isa/rtc.h (revision 323446) +++ stable/11/sys/isa/rtc.h (revision 323447) @@ -1,123 +1,124 @@ /*- * Copyright (c) 1990 The Regents of the University of California. * All rights reserved. * * This code is derived from software contributed to Berkeley by * William Jolitz. * * 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. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS 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 REGENTS 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. * * from: @(#)rtc.h 7.1 (Berkeley) 5/12/91 * $FreeBSD$ */ #ifndef _I386_ISA_RTC_H_ #define _I386_ISA_RTC_H_ 1 /* * MC146818 RTC Register locations */ #define RTC_SEC 0x00 /* seconds */ #define RTC_SECALRM 0x01 /* seconds alarm */ #define RTC_MIN 0x02 /* minutes */ #define RTC_MINALRM 0x03 /* minutes alarm */ #define RTC_HRS 0x04 /* hours */ #define RTC_HRSALRM 0x05 /* hours alarm */ #define RTC_WDAY 0x06 /* week day */ #define RTC_DAY 0x07 /* day of month */ #define RTC_MONTH 0x08 /* month of year */ #define RTC_YEAR 0x09 /* month of year */ #define RTC_STATUSA 0x0a /* status register A */ #define RTCSA_TUP 0x80 /* time update, don't look now */ #define RTCSA_RESET 0x70 /* reset divider */ #define RTCSA_DIVIDER 0x20 /* divider correct for 32768 Hz */ #define RTCSA_8192 0x03 /* 8192 Hz interrupt */ #define RTCSA_4096 0x04 #define RTCSA_2048 0x05 #define RTCSA_1024 0x06 /* default for profiling */ #define RTCSA_PROF RTCSA_1024 #define RTC_PROFRATE 1024 #define RTCSA_512 0x07 #define RTCSA_256 0x08 #define RTCSA_128 0x09 #define RTCSA_NOPROF RTCSA_128 #define RTC_NOPROFRATE 128 #define RTCSA_64 0x0a #define RTCSA_32 0x0b /* 32 Hz interrupt */ #define RTC_STATUSB 0x0b /* status register B */ #define RTCSB_DST 0x01 /* USA Daylight Savings Time enable */ #define RTCSB_24HR 0x02 /* 0 = 12 hours, 1 = 24 hours */ #define RTCSB_BCD 0x04 /* 0 = BCD, 1 = Binary coded time */ #define RTCSB_SQWE 0x08 /* 1 = output sqare wave at SQW pin */ #define RTCSB_UINTR 0x10 /* 1 = enable update-ended interrupt */ #define RTCSB_AINTR 0x20 /* 1 = enable alarm interrupt */ #define RTCSB_PINTR 0x40 /* 1 = enable periodic clock interrupt */ #define RTCSB_HALT 0x80 /* stop clock updates */ #define RTC_INTR 0x0c /* status register C (R) interrupt source */ #define RTCIR_UPDATE 0x10 /* update intr */ #define RTCIR_ALARM 0x20 /* alarm intr */ #define RTCIR_PERIOD 0x40 /* periodic intr */ #define RTCIR_INT 0x80 /* interrupt output signal */ #define RTC_STATUSD 0x0d /* status register D (R) Lost Power */ #define RTCSD_PWR 0x80 /* clock power OK */ #define RTC_DIAG 0x0e /* status register E - bios diagnostic */ #define RTCDG_BITS "\020\010clock_battery\007ROM_cksum\006config_unit\005memory_size\004fixed_disk\003invalid_time" #define RTC_RESET 0x0f /* status register F - reset code byte */ #define RTCRS_RST 0x00 /* normal reset */ #define RTCRS_LOAD 0x04 /* load system */ #define RTC_FDISKETTE 0x10 /* diskette drive type in upper/lower nibble */ #define RTCFDT_NONE 0 /* none present */ #define RTCFDT_360K 0x10 /* 360K */ #define RTCFDT_12M 0x20 /* 1.2M */ #define RTCFDT_720K 0x30 /* 720K */ #define RTCFDT_144M 0x40 /* 1.44M */ #define RTCFDT_288M_1 0x50 /* 2.88M, some BIOSes */ #define RTCFDT_288M 0x60 /* 2.88M */ #define RTC_BASELO 0x15 /* low byte of basemem size */ #define RTC_BASEHI 0x16 /* high byte of basemem size */ #define RTC_EXTLO 0x17 /* low byte of extended mem size */ #define RTC_EXTHI 0x18 /* low byte of extended mem size */ #define RTC_CENTURY 0x32 /* current century */ #ifdef _KERNEL extern struct mtx clock_lock; +extern struct mtx atrtc_time_lock; extern int atrtcclock_disable; int rtcin(int reg); void atrtc_restore(void); void writertc(int reg, u_char val); void atrtc_set(struct timespec *ts); #endif #endif /* _I386_ISA_RTC_H_ */ Index: stable/11/sys/kern/subr_clock.c =================================================================== --- stable/11/sys/kern/subr_clock.c (revision 323446) +++ stable/11/sys/kern/subr_clock.c (revision 323447) @@ -1,224 +1,252 @@ /*- * Copyright (c) 1988 University of Utah. * Copyright (c) 1982, 1990, 1993 * The Regents of the University of California. All rights reserved. * * This code is derived from software contributed to Berkeley by * the Systems Programming Group of the University of Utah Computer * Science Department. * * 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. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS 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 REGENTS 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. * * from: Utah $Hdr: clock.c 1.18 91/01/21$ * from: @(#)clock.c 8.2 (Berkeley) 1/12/94 * from: NetBSD: clock_subr.c,v 1.6 2001/07/07 17:04:02 thorpej Exp * and * from: src/sys/i386/isa/clock.c,v 1.176 2001/09/04 */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include int tz_minuteswest; int tz_dsttime; /* * The adjkerntz and wall_cmos_clock sysctls are in the "machdep" sysctl * namespace because they were misplaced there originally. */ static int adjkerntz; static int sysctl_machdep_adjkerntz(SYSCTL_HANDLER_ARGS) { int error; error = sysctl_handle_int(oidp, oidp->oid_arg1, oidp->oid_arg2, req); if (!error && req->newptr) resettodr(); return (error); } SYSCTL_PROC(_machdep, OID_AUTO, adjkerntz, CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, &adjkerntz, 0, sysctl_machdep_adjkerntz, "I", "Local offset from UTC in seconds"); static int ct_debug; -SYSCTL_INT(_debug, OID_AUTO, clocktime, CTLFLAG_RW, +SYSCTL_INT(_debug, OID_AUTO, clocktime, CTLFLAG_RWTUN, &ct_debug, 0, "Enable printing of clocktime debugging"); static int wall_cmos_clock; SYSCTL_INT(_machdep, OID_AUTO, wall_cmos_clock, CTLFLAG_RW, &wall_cmos_clock, 0, "Enables application of machdep.adjkerntz"); /*--------------------------------------------------------------------* * Generic routines to convert between a POSIX date * (seconds since 1/1/1970) and yr/mo/day/hr/min/sec * Derived from NetBSD arch/hp300/hp300/clock.c */ #define FEBRUARY 2 #define days_in_year(y) (leapyear(y) ? 366 : 365) #define days_in_month(y, m) \ (month_days[(m) - 1] + (m == FEBRUARY ? leapyear(y) : 0)) /* Day of week. Days are counted from 1/1/1970, which was a Thursday */ #define day_of_week(days) (((days) + 4) % 7) static const int month_days[12] = { 31, 28, 31, 30, 31, 30, 31, 31, 30, 31, 30, 31 }; +/* + * Optimization: using a precomputed count of days between POSIX_BASE_YEAR and + * some recent year avoids lots of unnecessary loop iterations in conversion. + * recent_base_days is the number of days before the start of recent_base_year. + */ +static const int recent_base_year = 2017; +static const int recent_base_days = 17167; /* * This inline avoids some unnecessary modulo operations * as compared with the usual macro: * ( ((year % 4) == 0 && * (year % 100) != 0) || * ((year % 400) == 0) ) * It is otherwise equivalent. */ static int leapyear(int year) { int rv = 0; if ((year & 3) == 0) { rv = 1; if ((year % 100) == 0) { rv = 0; if ((year % 400) == 0) rv = 1; } } return (rv); } static void print_ct(struct clocktime *ct) { printf("[%04d-%02d-%02d %02d:%02d:%02d]", ct->year, ct->mon, ct->day, ct->hour, ct->min, ct->sec); } int clock_ct_to_ts(struct clocktime *ct, struct timespec *ts) { int i, year, days; - year = ct->year; - if (ct_debug) { printf("ct_to_ts("); print_ct(ct); printf(")"); } + /* + * Many realtime clocks store the year as 2-digit BCD; pivot on 70 to + * determine century. Some clocks have a "century bit" and drivers do + * year += 100, so interpret values between 70-199 as relative to 1900. + */ + year = ct->year; + if (year < 70) + year += 2000; + else if (year < 200) + year += 1900; + /* Sanity checks. */ if (ct->mon < 1 || ct->mon > 12 || ct->day < 1 || ct->day > days_in_month(year, ct->mon) || - ct->hour > 23 || ct->min > 59 || ct->sec > 59 || + ct->hour > 23 || ct->min > 59 || ct->sec > 59 || year < 1970 || (sizeof(time_t) == 4 && year > 2037)) { /* time_t overflow */ if (ct_debug) printf(" = EINVAL\n"); return (EINVAL); } /* * Compute days since start of time * First from years, then from months. */ - days = 0; - for (i = POSIX_BASE_YEAR; i < year; i++) + if (year >= recent_base_year) { + i = recent_base_year; + days = recent_base_days; + } else { + i = POSIX_BASE_YEAR; + days = 0; + } + for (; i < year; i++) days += days_in_year(i); /* Months */ for (i = 1; i < ct->mon; i++) days += days_in_month(year, i); days += (ct->day - 1); ts->tv_sec = (((time_t)days * 24 + ct->hour) * 60 + ct->min) * 60 + ct->sec; ts->tv_nsec = ct->nsec; if (ct_debug) - printf(" = %ld.%09ld\n", (long)ts->tv_sec, (long)ts->tv_nsec); + printf(" = %jd.%09ld\n", (intmax_t)ts->tv_sec, ts->tv_nsec); return (0); } void clock_ts_to_ct(struct timespec *ts, struct clocktime *ct) { int i, year, days; time_t rsec; /* remainder seconds */ time_t secs; secs = ts->tv_sec; days = secs / SECDAY; rsec = secs % SECDAY; ct->dow = day_of_week(days); - /* Subtract out whole years, counting them in i. */ - for (year = POSIX_BASE_YEAR; days >= days_in_year(year); year++) + /* Subtract out whole years. */ + if (days >= recent_base_days) { + year = recent_base_year; + days -= recent_base_days; + } else { + year = POSIX_BASE_YEAR; + } + for (; days >= days_in_year(year); year++) days -= days_in_year(year); ct->year = year; /* Subtract out whole months, counting them in i. */ for (i = 1; days >= days_in_month(year, i); i++) days -= days_in_month(year, i); ct->mon = i; /* Days are what is left over (+1) from all that. */ ct->day = days + 1; /* Hours, minutes, seconds are easy */ ct->hour = rsec / 3600; rsec = rsec % 3600; ct->min = rsec / 60; rsec = rsec % 60; ct->sec = rsec; ct->nsec = ts->tv_nsec; if (ct_debug) { - printf("ts_to_ct(%ld.%09ld) = ", - (long)ts->tv_sec, (long)ts->tv_nsec); + printf("ts_to_ct(%jd.%09ld) = ", + (intmax_t)ts->tv_sec, ts->tv_nsec); print_ct(ct); printf("\n"); } } int utc_offset(void) { return (tz_minuteswest * 60 + (wall_cmos_clock ? adjkerntz : 0)); } Index: stable/11/sys/kern/subr_rtc.c =================================================================== --- stable/11/sys/kern/subr_rtc.c (revision 323446) +++ stable/11/sys/kern/subr_rtc.c (revision 323447) @@ -1,187 +1,322 @@ /*- * Copyright (c) 1988 University of Utah. * Copyright (c) 1982, 1990, 1993 * The Regents of the University of California. * Copyright (c) 2011 The FreeBSD Foundation * All rights reserved. * * This code is derived from software contributed to Berkeley by * the Systems Programming Group of the University of Utah Computer * Science Department. * * Portions of this software were developed by Julien Ridoux at the University * of Melbourne under sponsorship from the FreeBSD Foundation. * * 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. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS 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 REGENTS 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. * * from: Utah $Hdr: clock.c 1.18 91/01/21$ * from: @(#)clock.c 8.2 (Berkeley) 1/12/94 * from: NetBSD: clock_subr.c,v 1.6 2001/07/07 17:04:02 thorpej Exp * and * from: src/sys/i386/isa/clock.c,v 1.176 2001/09/04 */ /* * Helpers for time-of-day clocks. This is useful for architectures that need * support multiple models of such clocks, and generally serves to make the * code more machine-independent. * If the clock in question can also be used as a time counter, the driver * needs to initiate this. * This code is not yet used by all architectures. */ #include __FBSDID("$FreeBSD$"); #include "opt_ffclock.h" #include #include #include #include #include #include -#include +#include +#include #include +#include #ifdef FFCLOCK #include #endif #include #include "clock_if.h" -static device_t clock_dev = NULL; -static long clock_res; -static struct timespec clock_adj; -struct mtx resettodr_lock; -MTX_SYSINIT(resettodr_init, &resettodr_lock, "tod2rl", MTX_DEF); - /* XXX: should be kern. now, it's no longer machdep. */ static int disable_rtc_set; SYSCTL_INT(_machdep, OID_AUTO, disable_rtc_set, CTLFLAG_RW, &disable_rtc_set, 0, "Disallow adjusting time-of-day clock"); +/* + * An instance of a realtime clock. A list of these tracks all the registered + * clocks in the system. + * + * The resadj member is used to apply a "resolution adjustment" equal to half + * the clock's resolution, which is useful mainly on clocks with a whole-second + * resolution. Because the clock truncates the fractional part, adding half the + * resolution performs 4/5 rounding. The same adjustment is applied to the + * times returned from clock_gettime(), because the fraction returned will + * always be zero, but on average the actual fraction at the time of the call + * should be about .5. + */ +struct rtc_instance { + device_t clockdev; + int resolution; + int flags; + u_int schedns; + struct timespec resadj; + struct timeout_task + stask; + LIST_ENTRY(rtc_instance) + rtc_entries; +}; + +/* + * Clocks are updated using a task running on taskqueue_thread. + */ +static void settime_task_func(void *arg, int pending); + +/* + * Registered clocks are kept in a list which is sorted by resolution; the more + * accurate clocks get the first shot at providing the time. + */ +LIST_HEAD(rtc_listhead, rtc_instance); +static struct rtc_listhead rtc_list = LIST_HEAD_INITIALIZER(rtc_list); +static struct sx rtc_list_lock; +SX_SYSINIT(rtc_list_lock_init, &rtc_list_lock, "rtc list"); + +/* + * On the task thread, invoke the clock_settime() method of the clock. Do so + * holding no locks, so that clock drivers are free to do whatever kind of + * locking or sleeping they need to. + */ +static void +settime_task_func(void *arg, int pending) +{ + struct timespec ts; + struct rtc_instance *rtc; + + rtc = arg; + if (!(rtc->flags & CLOCKF_SETTIME_NO_TS)) { + getnanotime(&ts); + if (!(rtc->flags & CLOCKF_SETTIME_NO_ADJ)) { + ts.tv_sec -= utc_offset(); + timespecadd(&ts, &rtc->resadj); + } + } else { + ts.tv_sec = 0; + ts.tv_nsec = 0; + } + CLOCK_SETTIME(rtc->clockdev, &ts); +} + void -clock_register(device_t dev, long res) /* res has units of microseconds */ +clock_register_flags(device_t clockdev, long resolution, int flags) { + struct rtc_instance *rtc, *newrtc; - if (clock_dev != NULL) { - if (clock_res <= res) { - if (bootverbose) - device_printf(dev, "not installed as " - "time-of-day clock: clock %s has higher " - "resolution\n", device_get_name(clock_dev)); - return; + newrtc = malloc(sizeof(*newrtc), M_DEVBUF, M_WAITOK); + newrtc->clockdev = clockdev; + newrtc->resolution = (int)resolution; + newrtc->flags = flags; + newrtc->schedns = 0; + newrtc->resadj.tv_sec = newrtc->resolution / 2 / 1000000; + newrtc->resadj.tv_nsec = newrtc->resolution / 2 % 1000000 * 1000; + TIMEOUT_TASK_INIT(taskqueue_thread, &newrtc->stask, 0, + settime_task_func, newrtc); + + sx_xlock(&rtc_list_lock); + if (LIST_EMPTY(&rtc_list)) { + LIST_INSERT_HEAD(&rtc_list, newrtc, rtc_entries); + } else { + LIST_FOREACH(rtc, &rtc_list, rtc_entries) { + if (rtc->resolution > newrtc->resolution) { + LIST_INSERT_BEFORE(rtc, newrtc, rtc_entries); + break; + } else if (LIST_NEXT(rtc, rtc_entries) == NULL) { + LIST_INSERT_AFTER(rtc, newrtc, rtc_entries); + break; + } } - if (bootverbose) - device_printf(clock_dev, "removed as " - "time-of-day clock: clock %s has higher " - "resolution\n", device_get_name(dev)); } - clock_dev = dev; - clock_res = res; - clock_adj.tv_sec = res / 2 / 1000000; - clock_adj.tv_nsec = res / 2 % 1000000 * 1000; - if (bootverbose) - device_printf(dev, "registered as a time-of-day clock " - "(resolution %ldus, adjustment %jd.%09jds)\n", res, - (intmax_t)clock_adj.tv_sec, (intmax_t)clock_adj.tv_nsec); + sx_xunlock(&rtc_list_lock); + + device_printf(clockdev, + "registered as a time-of-day clock, resolution %d.%6.6ds\n", + newrtc->resolution / 1000000, newrtc->resolution % 1000000); } -/* - * inittodr and settodr derived from the i386 versions written - * by Christoph Robitschko , reintroduced and - * updated by Chris Stenton 8/10/94 - */ +void +clock_register(device_t dev, long res) +{ + clock_register_flags(dev, res, 0); +} + +void +clock_unregister(device_t clockdev) +{ + struct rtc_instance *rtc, *tmp; + + sx_xlock(&rtc_list_lock); + LIST_FOREACH_SAFE(rtc, &rtc_list, rtc_entries, tmp) { + if (rtc->clockdev == clockdev) { + LIST_REMOVE(rtc, rtc_entries); + break; + } + } + sx_xunlock(&rtc_list_lock); + if (rtc != NULL) { + taskqueue_cancel_timeout(taskqueue_thread, &rtc->stask, NULL); + taskqueue_drain_timeout(taskqueue_thread, &rtc->stask); + free(rtc, M_DEVBUF); + } +} + +void +clock_schedule(device_t clockdev, u_int offsetns) +{ + struct rtc_instance *rtc; + + sx_xlock(&rtc_list_lock); + LIST_FOREACH(rtc, &rtc_list, rtc_entries) { + if (rtc->clockdev == clockdev) { + rtc->schedns = offsetns; + break; + } + } + sx_xunlock(&rtc_list_lock); +} + /* - * Initialize the time of day register, based on the time base which is, e.g. - * from a filesystem. + * Initialize the system time. Must be called from a context which does not + * restrict any locking or sleeping that clock drivers may need to do. + * + * First attempt to get the time from a registered realtime clock. The clocks + * are queried in order of resolution until one provides the time. If no clock + * can provide the current time, use the 'base' time provided by the caller, if + * non-zero. The 'base' time is potentially highly inaccurate, such as the last + * known good value of the system clock, or even a filesystem last-updated + * timestamp. It is used to prevent system time from appearing to move + * backwards in logs. */ void inittodr(time_t base) { struct timespec ts; + struct rtc_instance *rtc; int error; - if (clock_dev == NULL) { - printf("warning: no time-of-day clock registered, system time " - "will not be set accurately\n"); - goto wrong_time; + error = ENXIO; + sx_xlock(&rtc_list_lock); + LIST_FOREACH(rtc, &rtc_list, rtc_entries) { + if ((error = CLOCK_GETTIME(rtc->clockdev, &ts)) != 0) + continue; + if (ts.tv_sec < 0 || ts.tv_nsec < 0) { + error = EINVAL; + continue; + } + if (!(rtc->flags & CLOCKF_GETTIME_NO_ADJ)) { + timespecadd(&ts, &rtc->resadj); + ts.tv_sec += utc_offset(); + } + if (bootverbose) + device_printf(rtc->clockdev, + "providing initial system time\n"); + break; } - /* XXX: We should poll all registered RTCs in case of failure */ - mtx_lock(&resettodr_lock); - error = CLOCK_GETTIME(clock_dev, &ts); - mtx_unlock(&resettodr_lock); - if (error != 0 && error != EINVAL) { - printf("warning: clock_gettime failed (%d), the system time " - "will not be set accurately\n", error); - goto wrong_time; + sx_xunlock(&rtc_list_lock); + + /* + * Do not report errors from each clock; it is expected that some clocks + * cannot provide results in some situations. Only report problems when + * no clocks could provide the time. + */ + if (error != 0) { + switch (error) { + case ENXIO: + printf("Warning: no time-of-day clock registered, "); + break; + case EINVAL: + printf("Warning: bad time from time-of-day clock, "); + break; + default: + printf("Error reading time-of-day clock (%d), ", error); + break; + } + printf("system time will not be set accurately\n"); + ts.tv_sec = (base > 0) ? base : -1; + ts.tv_nsec = 0; } - if (error == EINVAL || ts.tv_sec < 0) { - printf("Invalid time in real time clock.\n" - "Check and reset the date immediately!\n"); - goto wrong_time; - } - ts.tv_sec += utc_offset(); - timespecadd(&ts, &clock_adj); - tc_setclock(&ts); + if (ts.tv_sec >= 0) { + tc_setclock(&ts); #ifdef FFCLOCK - ffclock_reset_clock(&ts); + ffclock_reset_clock(&ts); #endif - return; - -wrong_time: - if (base > 0) { - ts.tv_sec = base; - ts.tv_nsec = 0; - tc_setclock(&ts); } } /* - * Write system time back to RTC + * Write system time back to all registered clocks, unless disabled by admin. + * This can be called from a context that restricts locking and/or sleeping; the + * actual updating is done asynchronously on a task thread. */ void resettodr(void) { - struct timespec ts; - int error; + struct timespec now; + struct rtc_instance *rtc; + sbintime_t sbt; + long waitns; - if (disable_rtc_set || clock_dev == NULL) + if (disable_rtc_set) return; - getnanotime(&ts); - timespecadd(&ts, &clock_adj); - ts.tv_sec -= utc_offset(); - /* XXX: We should really set all registered RTCs */ - mtx_lock(&resettodr_lock); - error = CLOCK_SETTIME(clock_dev, &ts); - mtx_unlock(&resettodr_lock); - if (error != 0) - printf("warning: clock_settime failed (%d), time-of-day clock " - "not adjusted to system time\n", error); + sx_xlock(&rtc_list_lock); + LIST_FOREACH(rtc, &rtc_list, rtc_entries) { + if (rtc->schedns != 0) { + getnanotime(&now); + waitns = rtc->schedns - now.tv_nsec; + if (waitns < 0) + waitns += 1000000000; + sbt = nstosbt(waitns); + } else + sbt = 0; + taskqueue_enqueue_timeout_sbt(taskqueue_thread, + &rtc->stask, -sbt, 0, C_PREL(31)); + } + sx_xunlock(&rtc_list_lock); } Index: stable/11/sys/kern/subr_taskqueue.c =================================================================== --- stable/11/sys/kern/subr_taskqueue.c (revision 323446) +++ stable/11/sys/kern/subr_taskqueue.c (revision 323447) @@ -1,830 +1,839 @@ /*- * Copyright (c) 2000 Doug Rabson * 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. * 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 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 #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include static MALLOC_DEFINE(M_TASKQUEUE, "taskqueue", "Task Queues"); static void *taskqueue_giant_ih; static void *taskqueue_ih; static void taskqueue_fast_enqueue(void *); static void taskqueue_swi_enqueue(void *); static void taskqueue_swi_giant_enqueue(void *); struct taskqueue_busy { struct task *tb_running; TAILQ_ENTRY(taskqueue_busy) tb_link; }; struct task * const TB_DRAIN_WAITER = (struct task *)0x1; struct taskqueue { STAILQ_HEAD(, task) tq_queue; taskqueue_enqueue_fn tq_enqueue; void *tq_context; char *tq_name; TAILQ_HEAD(, taskqueue_busy) tq_active; struct mtx tq_mutex; struct thread **tq_threads; int tq_tcount; int tq_spin; int tq_flags; int tq_callouts; taskqueue_callback_fn tq_callbacks[TASKQUEUE_NUM_CALLBACKS]; void *tq_cb_contexts[TASKQUEUE_NUM_CALLBACKS]; }; #define TQ_FLAGS_ACTIVE (1 << 0) #define TQ_FLAGS_BLOCKED (1 << 1) #define TQ_FLAGS_UNLOCKED_ENQUEUE (1 << 2) #define DT_CALLOUT_ARMED (1 << 0) #define DT_DRAIN_IN_PROGRESS (1 << 1) #define TQ_LOCK(tq) \ do { \ if ((tq)->tq_spin) \ mtx_lock_spin(&(tq)->tq_mutex); \ else \ mtx_lock(&(tq)->tq_mutex); \ } while (0) #define TQ_ASSERT_LOCKED(tq) mtx_assert(&(tq)->tq_mutex, MA_OWNED) #define TQ_UNLOCK(tq) \ do { \ if ((tq)->tq_spin) \ mtx_unlock_spin(&(tq)->tq_mutex); \ else \ mtx_unlock(&(tq)->tq_mutex); \ } while (0) #define TQ_ASSERT_UNLOCKED(tq) mtx_assert(&(tq)->tq_mutex, MA_NOTOWNED) void _timeout_task_init(struct taskqueue *queue, struct timeout_task *timeout_task, int priority, task_fn_t func, void *context) { TASK_INIT(&timeout_task->t, priority, func, context); callout_init_mtx(&timeout_task->c, &queue->tq_mutex, CALLOUT_RETURNUNLOCKED); timeout_task->q = queue; timeout_task->f = 0; } static __inline int TQ_SLEEP(struct taskqueue *tq, void *p, struct mtx *m, int pri, const char *wm, int t) { if (tq->tq_spin) return (msleep_spin(p, m, wm, t)); return (msleep(p, m, pri, wm, t)); } static struct taskqueue * _taskqueue_create(const char *name, int mflags, taskqueue_enqueue_fn enqueue, void *context, int mtxflags, const char *mtxname __unused) { struct taskqueue *queue; char *tq_name; tq_name = malloc(TASKQUEUE_NAMELEN, M_TASKQUEUE, mflags | M_ZERO); if (tq_name == NULL) return (NULL); queue = malloc(sizeof(struct taskqueue), M_TASKQUEUE, mflags | M_ZERO); if (queue == NULL) { free(tq_name, M_TASKQUEUE); return (NULL); } snprintf(tq_name, TASKQUEUE_NAMELEN, "%s", (name) ? name : "taskqueue"); STAILQ_INIT(&queue->tq_queue); TAILQ_INIT(&queue->tq_active); queue->tq_enqueue = enqueue; queue->tq_context = context; queue->tq_name = tq_name; queue->tq_spin = (mtxflags & MTX_SPIN) != 0; queue->tq_flags |= TQ_FLAGS_ACTIVE; if (enqueue == taskqueue_fast_enqueue || enqueue == taskqueue_swi_enqueue || enqueue == taskqueue_swi_giant_enqueue || enqueue == taskqueue_thread_enqueue) queue->tq_flags |= TQ_FLAGS_UNLOCKED_ENQUEUE; mtx_init(&queue->tq_mutex, tq_name, NULL, mtxflags); return (queue); } struct taskqueue * taskqueue_create(const char *name, int mflags, taskqueue_enqueue_fn enqueue, void *context) { return _taskqueue_create(name, mflags, enqueue, context, MTX_DEF, name); } void taskqueue_set_callback(struct taskqueue *queue, enum taskqueue_callback_type cb_type, taskqueue_callback_fn callback, void *context) { KASSERT(((cb_type >= TASKQUEUE_CALLBACK_TYPE_MIN) && (cb_type <= TASKQUEUE_CALLBACK_TYPE_MAX)), ("Callback type %d not valid, must be %d-%d", cb_type, TASKQUEUE_CALLBACK_TYPE_MIN, TASKQUEUE_CALLBACK_TYPE_MAX)); KASSERT((queue->tq_callbacks[cb_type] == NULL), ("Re-initialization of taskqueue callback?")); queue->tq_callbacks[cb_type] = callback; queue->tq_cb_contexts[cb_type] = context; } /* * Signal a taskqueue thread to terminate. */ static void taskqueue_terminate(struct thread **pp, struct taskqueue *tq) { while (tq->tq_tcount > 0 || tq->tq_callouts > 0) { wakeup(tq); TQ_SLEEP(tq, pp, &tq->tq_mutex, PWAIT, "taskqueue_destroy", 0); } } void taskqueue_free(struct taskqueue *queue) { TQ_LOCK(queue); queue->tq_flags &= ~TQ_FLAGS_ACTIVE; taskqueue_terminate(queue->tq_threads, queue); KASSERT(TAILQ_EMPTY(&queue->tq_active), ("Tasks still running?")); KASSERT(queue->tq_callouts == 0, ("Armed timeout tasks")); mtx_destroy(&queue->tq_mutex); free(queue->tq_threads, M_TASKQUEUE); free(queue->tq_name, M_TASKQUEUE); free(queue, M_TASKQUEUE); } static int taskqueue_enqueue_locked(struct taskqueue *queue, struct task *task) { struct task *ins; struct task *prev; KASSERT(task->ta_func != NULL, ("enqueueing task with NULL func")); /* * Count multiple enqueues. */ if (task->ta_pending) { if (task->ta_pending < USHRT_MAX) task->ta_pending++; TQ_UNLOCK(queue); return (0); } /* * Optimise the case when all tasks have the same priority. */ prev = STAILQ_LAST(&queue->tq_queue, task, ta_link); if (!prev || prev->ta_priority >= task->ta_priority) { STAILQ_INSERT_TAIL(&queue->tq_queue, task, ta_link); } else { prev = NULL; for (ins = STAILQ_FIRST(&queue->tq_queue); ins; prev = ins, ins = STAILQ_NEXT(ins, ta_link)) if (ins->ta_priority < task->ta_priority) break; if (prev) STAILQ_INSERT_AFTER(&queue->tq_queue, prev, task, ta_link); else STAILQ_INSERT_HEAD(&queue->tq_queue, task, ta_link); } task->ta_pending = 1; if ((queue->tq_flags & TQ_FLAGS_UNLOCKED_ENQUEUE) != 0) TQ_UNLOCK(queue); if ((queue->tq_flags & TQ_FLAGS_BLOCKED) == 0) queue->tq_enqueue(queue->tq_context); if ((queue->tq_flags & TQ_FLAGS_UNLOCKED_ENQUEUE) == 0) TQ_UNLOCK(queue); /* Return with lock released. */ return (0); } int taskqueue_enqueue(struct taskqueue *queue, struct task *task) { int res; TQ_LOCK(queue); res = taskqueue_enqueue_locked(queue, task); /* The lock is released inside. */ return (res); } static void taskqueue_timeout_func(void *arg) { struct taskqueue *queue; struct timeout_task *timeout_task; timeout_task = arg; queue = timeout_task->q; KASSERT((timeout_task->f & DT_CALLOUT_ARMED) != 0, ("Stray timeout")); timeout_task->f &= ~DT_CALLOUT_ARMED; queue->tq_callouts--; taskqueue_enqueue_locked(timeout_task->q, &timeout_task->t); /* The lock is released inside. */ } int -taskqueue_enqueue_timeout(struct taskqueue *queue, - struct timeout_task *timeout_task, int ticks) +taskqueue_enqueue_timeout_sbt(struct taskqueue *queue, + struct timeout_task *timeout_task, sbintime_t sbt, sbintime_t pr, int flags) { int res; TQ_LOCK(queue); KASSERT(timeout_task->q == NULL || timeout_task->q == queue, ("Migrated queue")); KASSERT(!queue->tq_spin, ("Timeout for spin-queue")); timeout_task->q = queue; res = timeout_task->t.ta_pending; if (timeout_task->f & DT_DRAIN_IN_PROGRESS) { /* Do nothing */ TQ_UNLOCK(queue); res = -1; - } else if (ticks == 0) { + } else if (sbt == 0) { taskqueue_enqueue_locked(queue, &timeout_task->t); /* The lock is released inside. */ } else { if ((timeout_task->f & DT_CALLOUT_ARMED) != 0) { res++; } else { queue->tq_callouts++; timeout_task->f |= DT_CALLOUT_ARMED; - if (ticks < 0) - ticks = -ticks; /* Ignore overflow. */ + if (sbt < 0) + sbt = -sbt; /* Ignore overflow. */ } - if (ticks > 0) { - callout_reset(&timeout_task->c, ticks, - taskqueue_timeout_func, timeout_task); + if (sbt > 0) { + callout_reset_sbt(&timeout_task->c, sbt, pr, + taskqueue_timeout_func, timeout_task, flags); } TQ_UNLOCK(queue); } return (res); +} + +int +taskqueue_enqueue_timeout(struct taskqueue *queue, + struct timeout_task *ttask, int ticks) +{ + + return (taskqueue_enqueue_timeout_sbt(queue, ttask, ticks * tick_sbt, + 0, 0)); } static void taskqueue_task_nop_fn(void *context, int pending) { } /* * Block until all currently queued tasks in this taskqueue * have begun execution. Tasks queued during execution of * this function are ignored. */ static void taskqueue_drain_tq_queue(struct taskqueue *queue) { struct task t_barrier; if (STAILQ_EMPTY(&queue->tq_queue)) return; /* * Enqueue our barrier after all current tasks, but with * the highest priority so that newly queued tasks cannot * pass it. Because of the high priority, we can not use * taskqueue_enqueue_locked directly (which drops the lock * anyway) so just insert it at tail while we have the * queue lock. */ TASK_INIT(&t_barrier, USHRT_MAX, taskqueue_task_nop_fn, &t_barrier); STAILQ_INSERT_TAIL(&queue->tq_queue, &t_barrier, ta_link); t_barrier.ta_pending = 1; /* * Once the barrier has executed, all previously queued tasks * have completed or are currently executing. */ while (t_barrier.ta_pending != 0) TQ_SLEEP(queue, &t_barrier, &queue->tq_mutex, PWAIT, "-", 0); } /* * Block until all currently executing tasks for this taskqueue * complete. Tasks that begin execution during the execution * of this function are ignored. */ static void taskqueue_drain_tq_active(struct taskqueue *queue) { struct taskqueue_busy tb_marker, *tb_first; if (TAILQ_EMPTY(&queue->tq_active)) return; /* Block taskq_terminate().*/ queue->tq_callouts++; /* * Wait for all currently executing taskqueue threads * to go idle. */ tb_marker.tb_running = TB_DRAIN_WAITER; TAILQ_INSERT_TAIL(&queue->tq_active, &tb_marker, tb_link); while (TAILQ_FIRST(&queue->tq_active) != &tb_marker) TQ_SLEEP(queue, &tb_marker, &queue->tq_mutex, PWAIT, "-", 0); TAILQ_REMOVE(&queue->tq_active, &tb_marker, tb_link); /* * Wakeup any other drain waiter that happened to queue up * without any intervening active thread. */ tb_first = TAILQ_FIRST(&queue->tq_active); if (tb_first != NULL && tb_first->tb_running == TB_DRAIN_WAITER) wakeup(tb_first); /* Release taskqueue_terminate(). */ queue->tq_callouts--; if ((queue->tq_flags & TQ_FLAGS_ACTIVE) == 0) wakeup_one(queue->tq_threads); } void taskqueue_block(struct taskqueue *queue) { TQ_LOCK(queue); queue->tq_flags |= TQ_FLAGS_BLOCKED; TQ_UNLOCK(queue); } void taskqueue_unblock(struct taskqueue *queue) { TQ_LOCK(queue); queue->tq_flags &= ~TQ_FLAGS_BLOCKED; if (!STAILQ_EMPTY(&queue->tq_queue)) queue->tq_enqueue(queue->tq_context); TQ_UNLOCK(queue); } static void taskqueue_run_locked(struct taskqueue *queue) { struct taskqueue_busy tb; struct taskqueue_busy *tb_first; struct task *task; int pending; KASSERT(queue != NULL, ("tq is NULL")); TQ_ASSERT_LOCKED(queue); tb.tb_running = NULL; while (STAILQ_FIRST(&queue->tq_queue)) { TAILQ_INSERT_TAIL(&queue->tq_active, &tb, tb_link); /* * Carefully remove the first task from the queue and * zero its pending count. */ task = STAILQ_FIRST(&queue->tq_queue); KASSERT(task != NULL, ("task is NULL")); STAILQ_REMOVE_HEAD(&queue->tq_queue, ta_link); pending = task->ta_pending; task->ta_pending = 0; tb.tb_running = task; TQ_UNLOCK(queue); KASSERT(task->ta_func != NULL, ("task->ta_func is NULL")); task->ta_func(task->ta_context, pending); TQ_LOCK(queue); tb.tb_running = NULL; wakeup(task); TAILQ_REMOVE(&queue->tq_active, &tb, tb_link); tb_first = TAILQ_FIRST(&queue->tq_active); if (tb_first != NULL && tb_first->tb_running == TB_DRAIN_WAITER) wakeup(tb_first); } } void taskqueue_run(struct taskqueue *queue) { TQ_LOCK(queue); taskqueue_run_locked(queue); TQ_UNLOCK(queue); } static int task_is_running(struct taskqueue *queue, struct task *task) { struct taskqueue_busy *tb; TQ_ASSERT_LOCKED(queue); TAILQ_FOREACH(tb, &queue->tq_active, tb_link) { if (tb->tb_running == task) return (1); } return (0); } /* * Only use this function in single threaded contexts. It returns * non-zero if the given task is either pending or running. Else the * task is idle and can be queued again or freed. */ int taskqueue_poll_is_busy(struct taskqueue *queue, struct task *task) { int retval; TQ_LOCK(queue); retval = task->ta_pending > 0 || task_is_running(queue, task); TQ_UNLOCK(queue); return (retval); } static int taskqueue_cancel_locked(struct taskqueue *queue, struct task *task, u_int *pendp) { if (task->ta_pending > 0) STAILQ_REMOVE(&queue->tq_queue, task, task, ta_link); if (pendp != NULL) *pendp = task->ta_pending; task->ta_pending = 0; return (task_is_running(queue, task) ? EBUSY : 0); } int taskqueue_cancel(struct taskqueue *queue, struct task *task, u_int *pendp) { int error; TQ_LOCK(queue); error = taskqueue_cancel_locked(queue, task, pendp); TQ_UNLOCK(queue); return (error); } int taskqueue_cancel_timeout(struct taskqueue *queue, struct timeout_task *timeout_task, u_int *pendp) { u_int pending, pending1; int error; TQ_LOCK(queue); pending = !!(callout_stop(&timeout_task->c) > 0); error = taskqueue_cancel_locked(queue, &timeout_task->t, &pending1); if ((timeout_task->f & DT_CALLOUT_ARMED) != 0) { timeout_task->f &= ~DT_CALLOUT_ARMED; queue->tq_callouts--; } TQ_UNLOCK(queue); if (pendp != NULL) *pendp = pending + pending1; return (error); } void taskqueue_drain(struct taskqueue *queue, struct task *task) { if (!queue->tq_spin) WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, __func__); TQ_LOCK(queue); while (task->ta_pending != 0 || task_is_running(queue, task)) TQ_SLEEP(queue, task, &queue->tq_mutex, PWAIT, "-", 0); TQ_UNLOCK(queue); } void taskqueue_drain_all(struct taskqueue *queue) { if (!queue->tq_spin) WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL, __func__); TQ_LOCK(queue); taskqueue_drain_tq_queue(queue); taskqueue_drain_tq_active(queue); TQ_UNLOCK(queue); } void taskqueue_drain_timeout(struct taskqueue *queue, struct timeout_task *timeout_task) { /* * Set flag to prevent timer from re-starting during drain: */ TQ_LOCK(queue); KASSERT((timeout_task->f & DT_DRAIN_IN_PROGRESS) == 0, ("Drain already in progress")); timeout_task->f |= DT_DRAIN_IN_PROGRESS; TQ_UNLOCK(queue); callout_drain(&timeout_task->c); taskqueue_drain(queue, &timeout_task->t); /* * Clear flag to allow timer to re-start: */ TQ_LOCK(queue); timeout_task->f &= ~DT_DRAIN_IN_PROGRESS; TQ_UNLOCK(queue); } static void taskqueue_swi_enqueue(void *context) { swi_sched(taskqueue_ih, 0); } static void taskqueue_swi_run(void *dummy) { taskqueue_run(taskqueue_swi); } static void taskqueue_swi_giant_enqueue(void *context) { swi_sched(taskqueue_giant_ih, 0); } static void taskqueue_swi_giant_run(void *dummy) { taskqueue_run(taskqueue_swi_giant); } static int _taskqueue_start_threads(struct taskqueue **tqp, int count, int pri, cpuset_t *mask, const char *name, va_list ap) { char ktname[MAXCOMLEN + 1]; struct thread *td; struct taskqueue *tq; int i, error; if (count <= 0) return (EINVAL); vsnprintf(ktname, sizeof(ktname), name, ap); tq = *tqp; tq->tq_threads = malloc(sizeof(struct thread *) * count, M_TASKQUEUE, M_NOWAIT | M_ZERO); if (tq->tq_threads == NULL) { printf("%s: no memory for %s threads\n", __func__, ktname); return (ENOMEM); } for (i = 0; i < count; i++) { if (count == 1) error = kthread_add(taskqueue_thread_loop, tqp, NULL, &tq->tq_threads[i], RFSTOPPED, 0, "%s", ktname); else error = kthread_add(taskqueue_thread_loop, tqp, NULL, &tq->tq_threads[i], RFSTOPPED, 0, "%s_%d", ktname, i); if (error) { /* should be ok to continue, taskqueue_free will dtrt */ printf("%s: kthread_add(%s): error %d", __func__, ktname, error); tq->tq_threads[i] = NULL; /* paranoid */ } else tq->tq_tcount++; } for (i = 0; i < count; i++) { if (tq->tq_threads[i] == NULL) continue; td = tq->tq_threads[i]; if (mask) { error = cpuset_setthread(td->td_tid, mask); /* * Failing to pin is rarely an actual fatal error; * it'll just affect performance. */ if (error) printf("%s: curthread=%llu: can't pin; " "error=%d\n", __func__, (unsigned long long) td->td_tid, error); } thread_lock(td); sched_prio(td, pri); sched_add(td, SRQ_BORING); thread_unlock(td); } return (0); } int taskqueue_start_threads(struct taskqueue **tqp, int count, int pri, const char *name, ...) { va_list ap; int error; va_start(ap, name); error = _taskqueue_start_threads(tqp, count, pri, NULL, name, ap); va_end(ap); return (error); } int taskqueue_start_threads_cpuset(struct taskqueue **tqp, int count, int pri, cpuset_t *mask, const char *name, ...) { va_list ap; int error; va_start(ap, name); error = _taskqueue_start_threads(tqp, count, pri, mask, name, ap); va_end(ap); return (error); } static inline void taskqueue_run_callback(struct taskqueue *tq, enum taskqueue_callback_type cb_type) { taskqueue_callback_fn tq_callback; TQ_ASSERT_UNLOCKED(tq); tq_callback = tq->tq_callbacks[cb_type]; if (tq_callback != NULL) tq_callback(tq->tq_cb_contexts[cb_type]); } void taskqueue_thread_loop(void *arg) { struct taskqueue **tqp, *tq; tqp = arg; tq = *tqp; taskqueue_run_callback(tq, TASKQUEUE_CALLBACK_TYPE_INIT); TQ_LOCK(tq); while ((tq->tq_flags & TQ_FLAGS_ACTIVE) != 0) { /* XXX ? */ taskqueue_run_locked(tq); /* * Because taskqueue_run() can drop tq_mutex, we need to * check if the TQ_FLAGS_ACTIVE flag wasn't removed in the * meantime, which means we missed a wakeup. */ if ((tq->tq_flags & TQ_FLAGS_ACTIVE) == 0) break; TQ_SLEEP(tq, tq, &tq->tq_mutex, 0, "-", 0); } taskqueue_run_locked(tq); /* * This thread is on its way out, so just drop the lock temporarily * in order to call the shutdown callback. This allows the callback * to look at the taskqueue, even just before it dies. */ TQ_UNLOCK(tq); taskqueue_run_callback(tq, TASKQUEUE_CALLBACK_TYPE_SHUTDOWN); TQ_LOCK(tq); /* rendezvous with thread that asked us to terminate */ tq->tq_tcount--; wakeup_one(tq->tq_threads); TQ_UNLOCK(tq); kthread_exit(); } void taskqueue_thread_enqueue(void *context) { struct taskqueue **tqp, *tq; tqp = context; tq = *tqp; wakeup_one(tq); } TASKQUEUE_DEFINE(swi, taskqueue_swi_enqueue, NULL, swi_add(NULL, "task queue", taskqueue_swi_run, NULL, SWI_TQ, INTR_MPSAFE, &taskqueue_ih)); TASKQUEUE_DEFINE(swi_giant, taskqueue_swi_giant_enqueue, NULL, swi_add(NULL, "Giant taskq", taskqueue_swi_giant_run, NULL, SWI_TQ_GIANT, 0, &taskqueue_giant_ih)); TASKQUEUE_DEFINE_THREAD(thread); struct taskqueue * taskqueue_create_fast(const char *name, int mflags, taskqueue_enqueue_fn enqueue, void *context) { return _taskqueue_create(name, mflags, enqueue, context, MTX_SPIN, "fast_taskqueue"); } static void *taskqueue_fast_ih; static void taskqueue_fast_enqueue(void *context) { swi_sched(taskqueue_fast_ih, 0); } static void taskqueue_fast_run(void *dummy) { taskqueue_run(taskqueue_fast); } TASKQUEUE_FAST_DEFINE(fast, taskqueue_fast_enqueue, NULL, swi_add(NULL, "fast taskq", taskqueue_fast_run, NULL, SWI_TQ_FAST, INTR_MPSAFE, &taskqueue_fast_ih)); int taskqueue_member(struct taskqueue *queue, struct thread *td) { int i, j, ret = 0; for (i = 0, j = 0; ; i++) { if (queue->tq_threads[i] == NULL) continue; if (queue->tq_threads[i] == td) { ret = 1; break; } if (++j >= queue->tq_tcount) break; } return (ret); } Index: stable/11/sys/sys/clock.h =================================================================== --- stable/11/sys/sys/clock.h (revision 323446) +++ stable/11/sys/sys/clock.h (revision 323447) @@ -1,99 +1,133 @@ /*- * Copyright (c) 1996 The NetBSD Foundation, Inc. * All rights reserved. * * This code is derived from software contributed to The NetBSD Foundation * by Gordon W. Ross * * 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. * * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. 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 FOUNDATION 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. * * $NetBSD: clock_subr.h,v 1.7 2000/10/03 13:41:07 tsutsui Exp $ * * * This file is the central clearing-house for calendrical issues. * * In general the kernel does not know about minutes, hours, days, timezones, * daylight savings time, leap-years and such. All that is theoretically a * matter for userland only. * * Parts of kernel code does however care: badly designed filesystems store * timestamps in local time and RTC chips sometimes track time in a local * timezone instead of UTC and so on. * * All that code should go here for service. * * $FreeBSD$ */ #ifndef _SYS_CLOCK_H_ #define _SYS_CLOCK_H_ #ifdef _KERNEL /* No user serviceable parts */ /* * Timezone info from settimeofday(2), usually not used */ extern int tz_minuteswest; extern int tz_dsttime; -extern struct mtx resettodr_lock; int utc_offset(void); /* * Structure to hold the values typically reported by time-of-day clocks. * This can be passed to the generic conversion functions to be converted * to a struct timespec. */ struct clocktime { int year; /* year (4 digit year) */ int mon; /* month (1 - 12) */ int day; /* day (1 - 31) */ int hour; /* hour (0 - 23) */ int min; /* minute (0 - 59) */ int sec; /* second (0 - 59) */ int dow; /* day of week (0 - 6; 0 = Sunday) */ long nsec; /* nano seconds */ }; int clock_ct_to_ts(struct clocktime *, struct timespec *); void clock_ts_to_ct(struct timespec *, struct clocktime *); -void clock_register(device_t, long); + +/* + * Time-of-day clock functions and flags. These functions might sleep. + * + * clock_register and clock_unregister() do what they say. Upon return from + * unregister, the clock's methods are not running and will not be called again. + * + * clock_schedule() requests that a registered clock's clock_settime() calls + * happen at the given offset into the second. The default is 0, meaning no + * specific scheduling. To schedule the call as soon after top-of-second as + * possible, specify 1. Each clock has its own schedule, but taskqueue_thread + * is shared by many tasks; the timing of the call is not guaranteed. + * + * Flags: + * + * CLOCKF_SETTIME_NO_TS + * Do not pass a timespec to clock_settime(), the driver obtains its own time + * and applies its own adjustments (this flag implies CLOCKF_SETTIME_NO_ADJ). + * + * CLOCKF_SETTIME_NO_ADJ + * Do not apply utc offset and resolution/accuracy adjustments to the value + * passed to clock_settime(), the driver applies them itself. + * + * CLOCKF_GETTIME_NO_ADJ + * Do not apply utc offset and resolution/accuracy adjustments to the value + * returned from clock_gettime(), the driver has already applied them. + */ + +#define CLOCKF_SETTIME_NO_TS 0x00000001 +#define CLOCKF_SETTIME_NO_ADJ 0x00000002 +#define CLOCKF_GETTIME_NO_ADJ 0x00000004 + +void clock_register(device_t _clockdev, long _resolution_us); +void clock_register_flags(device_t _clockdev, long _resolution_us, int _flags); +void clock_schedule(device_t clockdev, u_int _offsetns); +void clock_unregister(device_t _clockdev); /* * BCD to decimal and decimal to BCD. */ #define FROMBCD(x) bcd2bin(x) #define TOBCD(x) bin2bcd(x) /* Some handy constants. */ #define SECDAY (24 * 60 * 60) #define SECYR (SECDAY * 365) /* Traditional POSIX base year */ #define POSIX_BASE_YEAR 1970 void timespec2fattime(struct timespec *tsp, int utc, u_int16_t *ddp, u_int16_t *dtp, u_int8_t *dhp); void fattime2timespec(unsigned dd, unsigned dt, unsigned dh, int utc, struct timespec *tsp); #endif /* _KERNEL */ #endif /* !_SYS_CLOCK_H_ */ Index: stable/11/sys/sys/taskqueue.h =================================================================== --- stable/11/sys/sys/taskqueue.h (revision 323446) +++ stable/11/sys/sys/taskqueue.h (revision 323447) @@ -1,208 +1,211 @@ /*- * Copyright (c) 2000 Doug Rabson * 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. * 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 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 _SYS_TASKQUEUE_H_ #define _SYS_TASKQUEUE_H_ #ifndef _KERNEL #error "no user-servicable parts inside" #endif #include #include #include #include struct taskqueue; struct taskqgroup; struct thread; struct timeout_task { struct taskqueue *q; struct task t; struct callout c; int f; }; enum taskqueue_callback_type { TASKQUEUE_CALLBACK_TYPE_INIT, TASKQUEUE_CALLBACK_TYPE_SHUTDOWN, }; #define TASKQUEUE_CALLBACK_TYPE_MIN TASKQUEUE_CALLBACK_TYPE_INIT #define TASKQUEUE_CALLBACK_TYPE_MAX TASKQUEUE_CALLBACK_TYPE_SHUTDOWN #define TASKQUEUE_NUM_CALLBACKS TASKQUEUE_CALLBACK_TYPE_MAX + 1 #define TASKQUEUE_NAMELEN 32 typedef void (*taskqueue_callback_fn)(void *context); /* * A notification callback function which is called from * taskqueue_enqueue(). The context argument is given in the call to * taskqueue_create(). This function would normally be used to allow the * queue to arrange to run itself later (e.g., by scheduling a software * interrupt or waking a kernel thread). */ typedef void (*taskqueue_enqueue_fn)(void *context); struct taskqueue *taskqueue_create(const char *name, int mflags, taskqueue_enqueue_fn enqueue, void *context); int taskqueue_start_threads(struct taskqueue **tqp, int count, int pri, const char *name, ...) __printflike(4, 5); int taskqueue_start_threads_cpuset(struct taskqueue **tqp, int count, int pri, cpuset_t *mask, const char *name, ...) __printflike(5, 6); int taskqueue_enqueue(struct taskqueue *queue, struct task *task); int taskqueue_enqueue_timeout(struct taskqueue *queue, struct timeout_task *timeout_task, int ticks); +int taskqueue_enqueue_timeout_sbt(struct taskqueue *queue, + struct timeout_task *timeout_task, sbintime_t sbt, sbintime_t pr, + int flags); int taskqueue_poll_is_busy(struct taskqueue *queue, struct task *task); int taskqueue_cancel(struct taskqueue *queue, struct task *task, u_int *pendp); int taskqueue_cancel_timeout(struct taskqueue *queue, struct timeout_task *timeout_task, u_int *pendp); void taskqueue_drain(struct taskqueue *queue, struct task *task); void taskqueue_drain_timeout(struct taskqueue *queue, struct timeout_task *timeout_task); void taskqueue_drain_all(struct taskqueue *queue); void taskqueue_free(struct taskqueue *queue); void taskqueue_run(struct taskqueue *queue); void taskqueue_block(struct taskqueue *queue); void taskqueue_unblock(struct taskqueue *queue); int taskqueue_member(struct taskqueue *queue, struct thread *td); void taskqueue_set_callback(struct taskqueue *queue, enum taskqueue_callback_type cb_type, taskqueue_callback_fn callback, void *context); #define TASK_INITIALIZER(priority, func, context) \ { .ta_pending = 0, \ .ta_priority = (priority), \ .ta_func = (func), \ .ta_context = (context) } /* * Functions for dedicated thread taskqueues */ void taskqueue_thread_loop(void *arg); void taskqueue_thread_enqueue(void *context); /* * Initialise a task structure. */ #define TASK_INIT(task, priority, func, context) do { \ (task)->ta_pending = 0; \ (task)->ta_priority = (priority); \ (task)->ta_func = (func); \ (task)->ta_context = (context); \ } while (0) void _timeout_task_init(struct taskqueue *queue, struct timeout_task *timeout_task, int priority, task_fn_t func, void *context); #define TIMEOUT_TASK_INIT(queue, timeout_task, priority, func, context) \ _timeout_task_init(queue, timeout_task, priority, func, context); /* * Declare a reference to a taskqueue. */ #define TASKQUEUE_DECLARE(name) \ extern struct taskqueue *taskqueue_##name /* * Define and initialise a global taskqueue that uses sleep mutexes. */ #define TASKQUEUE_DEFINE(name, enqueue, context, init) \ \ struct taskqueue *taskqueue_##name; \ \ static void \ taskqueue_define_##name(void *arg) \ { \ taskqueue_##name = \ taskqueue_create(#name, M_WAITOK, (enqueue), (context)); \ init; \ } \ \ SYSINIT(taskqueue_##name, SI_SUB_INIT_IF, SI_ORDER_SECOND, \ taskqueue_define_##name, NULL); \ \ struct __hack #define TASKQUEUE_DEFINE_THREAD(name) \ TASKQUEUE_DEFINE(name, taskqueue_thread_enqueue, &taskqueue_##name, \ taskqueue_start_threads(&taskqueue_##name, 1, PWAIT, \ "%s taskq", #name)) /* * Define and initialise a global taskqueue that uses spin mutexes. */ #define TASKQUEUE_FAST_DEFINE(name, enqueue, context, init) \ \ struct taskqueue *taskqueue_##name; \ \ static void \ taskqueue_define_##name(void *arg) \ { \ taskqueue_##name = \ taskqueue_create_fast(#name, M_WAITOK, (enqueue), \ (context)); \ init; \ } \ \ SYSINIT(taskqueue_##name, SI_SUB_INIT_IF, SI_ORDER_SECOND, \ taskqueue_define_##name, NULL); \ \ struct __hack #define TASKQUEUE_FAST_DEFINE_THREAD(name) \ TASKQUEUE_FAST_DEFINE(name, taskqueue_thread_enqueue, \ &taskqueue_##name, taskqueue_start_threads(&taskqueue_##name \ 1, PWAIT, "%s taskq", #name)) /* * These queues are serviced by software interrupt handlers. To enqueue * a task, call taskqueue_enqueue(taskqueue_swi, &task) or * taskqueue_enqueue(taskqueue_swi_giant, &task). */ TASKQUEUE_DECLARE(swi_giant); TASKQUEUE_DECLARE(swi); /* * This queue is serviced by a kernel thread. To enqueue a task, call * taskqueue_enqueue(taskqueue_thread, &task). */ TASKQUEUE_DECLARE(thread); /* * Queue for swi handlers dispatched from fast interrupt handlers. * These are necessarily different from the above because the queue * must be locked with spinlocks since sleep mutex's cannot be used * from a fast interrupt handler context. */ TASKQUEUE_DECLARE(fast); struct taskqueue *taskqueue_create_fast(const char *name, int mflags, taskqueue_enqueue_fn enqueue, void *context); #endif /* !_SYS_TASKQUEUE_H_ */ Index: stable/11/sys/x86/isa/atrtc.c =================================================================== --- stable/11/sys/x86/isa/atrtc.c (revision 323446) +++ stable/11/sys/x86/isa/atrtc.c (revision 323447) @@ -1,411 +1,425 @@ /*- * Copyright (c) 2008 Poul-Henning Kamp * Copyright (c) 2010 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. * 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 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. * * $FreeBSD$ */ #include __FBSDID("$FreeBSD$"); #include "opt_isa.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef DEV_ISA #include #include #endif #include #include "clock_if.h" +/* + * clock_lock protects low-level access to individual hardware registers. + * atrtc_time_lock protects the entire sequence of accessing multiple registers + * to read or write the date and time. + */ #define RTC_LOCK do { if (!kdb_active) mtx_lock_spin(&clock_lock); } while (0) #define RTC_UNLOCK do { if (!kdb_active) mtx_unlock_spin(&clock_lock); } while (0) +struct mtx atrtc_time_lock; +MTX_SYSINIT(atrtc_lock_init, &atrtc_time_lock, "atrtc", MTX_DEF); + int atrtcclock_disable = 0; static int rtc_reg = -1; static u_char rtc_statusa = RTCSA_DIVIDER | RTCSA_NOPROF; static u_char rtc_statusb = RTCSB_24HR; /* * RTC support routines */ int rtcin(int reg) { u_char val; RTC_LOCK; if (rtc_reg != reg) { inb(0x84); outb(IO_RTC, reg); rtc_reg = reg; inb(0x84); } val = inb(IO_RTC + 1); RTC_UNLOCK; return (val); } void writertc(int reg, u_char val) { RTC_LOCK; if (rtc_reg != reg) { inb(0x84); outb(IO_RTC, reg); rtc_reg = reg; inb(0x84); } outb(IO_RTC + 1, val); inb(0x84); RTC_UNLOCK; } static __inline int readrtc(int port) { return(bcd2bin(rtcin(port))); } static void atrtc_start(void) { writertc(RTC_STATUSA, rtc_statusa); writertc(RTC_STATUSB, RTCSB_24HR); } static void atrtc_rate(unsigned rate) { rtc_statusa = RTCSA_DIVIDER | rate; writertc(RTC_STATUSA, rtc_statusa); } static void atrtc_enable_intr(void) { rtc_statusb |= RTCSB_PINTR; writertc(RTC_STATUSB, rtc_statusb); rtcin(RTC_INTR); } static void atrtc_disable_intr(void) { rtc_statusb &= ~RTCSB_PINTR; writertc(RTC_STATUSB, rtc_statusb); rtcin(RTC_INTR); } void atrtc_restore(void) { /* Restore all of the RTC's "status" (actually, control) registers. */ rtcin(RTC_STATUSA); /* dummy to get rtc_reg set */ writertc(RTC_STATUSB, RTCSB_24HR); writertc(RTC_STATUSA, rtc_statusa); writertc(RTC_STATUSB, rtc_statusb); rtcin(RTC_INTR); } void atrtc_set(struct timespec *ts) { struct clocktime ct; clock_ts_to_ct(ts, &ct); + mtx_lock(&atrtc_time_lock); + /* Disable RTC updates and interrupts. */ writertc(RTC_STATUSB, RTCSB_HALT | RTCSB_24HR); writertc(RTC_SEC, bin2bcd(ct.sec)); /* Write back Seconds */ writertc(RTC_MIN, bin2bcd(ct.min)); /* Write back Minutes */ writertc(RTC_HRS, bin2bcd(ct.hour)); /* Write back Hours */ writertc(RTC_WDAY, ct.dow + 1); /* Write back Weekday */ writertc(RTC_DAY, bin2bcd(ct.day)); /* Write back Day */ writertc(RTC_MONTH, bin2bcd(ct.mon)); /* Write back Month */ writertc(RTC_YEAR, bin2bcd(ct.year % 100)); /* Write back Year */ #ifdef USE_RTC_CENTURY writertc(RTC_CENTURY, bin2bcd(ct.year / 100)); /* ... and Century */ #endif /* Re-enable RTC updates and interrupts. */ writertc(RTC_STATUSB, rtc_statusb); rtcin(RTC_INTR); + + mtx_unlock(&atrtc_time_lock); } /********************************************************************** * RTC driver for subr_rtc */ struct atrtc_softc { int port_rid, intr_rid; struct resource *port_res; struct resource *intr_res; void *intr_handler; struct eventtimer et; }; static int rtc_start(struct eventtimer *et, sbintime_t first, sbintime_t period) { atrtc_rate(max(fls(period + (period >> 1)) - 17, 1)); atrtc_enable_intr(); return (0); } static int rtc_stop(struct eventtimer *et) { atrtc_disable_intr(); return (0); } /* * This routine receives statistical clock interrupts from the RTC. * As explained above, these occur at 128 interrupts per second. * When profiling, we receive interrupts at a rate of 1024 Hz. * * This does not actually add as much overhead as it sounds, because * when the statistical clock is active, the hardclock driver no longer * needs to keep (inaccurate) statistics on its own. This decouples * statistics gathering from scheduling interrupts. * * The RTC chip requires that we read status register C (RTC_INTR) * to acknowledge an interrupt, before it will generate the next one. * Under high interrupt load, rtcintr() can be indefinitely delayed and * the clock can tick immediately after the read from RTC_INTR. In this * case, the mc146818A interrupt signal will not drop for long enough * to register with the 8259 PIC. If an interrupt is missed, the stat * clock will halt, considerably degrading system performance. This is * why we use 'while' rather than a more straightforward 'if' below. * Stat clock ticks can still be lost, causing minor loss of accuracy * in the statistics, but the stat clock will no longer stop. */ static int rtc_intr(void *arg) { struct atrtc_softc *sc = (struct atrtc_softc *)arg; int flag = 0; while (rtcin(RTC_INTR) & RTCIR_PERIOD) { flag = 1; if (sc->et.et_active) sc->et.et_event_cb(&sc->et, sc->et.et_arg); } return(flag ? FILTER_HANDLED : FILTER_STRAY); } /* * Attach to the ISA PnP descriptors for the timer and realtime clock. */ static struct isa_pnp_id atrtc_ids[] = { { 0x000bd041 /* PNP0B00 */, "AT realtime clock" }, { 0 } }; static int atrtc_probe(device_t dev) { int result; result = ISA_PNP_PROBE(device_get_parent(dev), dev, atrtc_ids); /* ENOENT means no PnP-ID, device is hinted. */ if (result == ENOENT) { device_set_desc(dev, "AT realtime clock"); return (BUS_PROBE_LOW_PRIORITY); } return (result); } static int atrtc_attach(device_t dev) { struct atrtc_softc *sc; rman_res_t s; int i; sc = device_get_softc(dev); sc->port_res = bus_alloc_resource(dev, SYS_RES_IOPORT, &sc->port_rid, IO_RTC, IO_RTC + 1, 2, RF_ACTIVE); if (sc->port_res == NULL) device_printf(dev, "Warning: Couldn't map I/O.\n"); atrtc_start(); clock_register(dev, 1000000); bzero(&sc->et, sizeof(struct eventtimer)); if (!atrtcclock_disable && (resource_int_value(device_get_name(dev), device_get_unit(dev), "clock", &i) != 0 || i != 0)) { sc->intr_rid = 0; while (bus_get_resource(dev, SYS_RES_IRQ, sc->intr_rid, &s, NULL) == 0 && s != 8) sc->intr_rid++; sc->intr_res = bus_alloc_resource(dev, SYS_RES_IRQ, &sc->intr_rid, 8, 8, 1, RF_ACTIVE); if (sc->intr_res == NULL) { device_printf(dev, "Can't map interrupt.\n"); return (0); } else if ((bus_setup_intr(dev, sc->intr_res, INTR_TYPE_CLK, rtc_intr, NULL, sc, &sc->intr_handler))) { device_printf(dev, "Can't setup interrupt.\n"); return (0); } else { /* Bind IRQ to BSP to avoid live migration. */ bus_bind_intr(dev, sc->intr_res, 0); } sc->et.et_name = "RTC"; sc->et.et_flags = ET_FLAGS_PERIODIC | ET_FLAGS_POW2DIV; sc->et.et_quality = 0; sc->et.et_frequency = 32768; sc->et.et_min_period = 0x00080000; sc->et.et_max_period = 0x80000000; sc->et.et_start = rtc_start; sc->et.et_stop = rtc_stop; sc->et.et_priv = dev; et_register(&sc->et); } return(0); } static int atrtc_resume(device_t dev) { atrtc_restore(); return(0); } static int atrtc_settime(device_t dev __unused, struct timespec *ts) { atrtc_set(ts); return (0); } static int atrtc_gettime(device_t dev, struct timespec *ts) { struct clocktime ct; /* Look if we have a RTC present and the time is valid */ if (!(rtcin(RTC_STATUSD) & RTCSD_PWR)) { device_printf(dev, "WARNING: Battery failure indication\n"); return (EINVAL); } /* * wait for time update to complete * If RTCSA_TUP is zero, we have at least 244us before next update. * This is fast enough on most hardware, but a refinement would be * to make sure that no more than 240us pass after we start reading, * and try again if so. */ + mtx_lock(&atrtc_time_lock); while (rtcin(RTC_STATUSA) & RTCSA_TUP) continue; critical_enter(); ct.nsec = 0; ct.sec = readrtc(RTC_SEC); ct.min = readrtc(RTC_MIN); ct.hour = readrtc(RTC_HRS); ct.day = readrtc(RTC_DAY); ct.dow = readrtc(RTC_WDAY) - 1; ct.mon = readrtc(RTC_MONTH); ct.year = readrtc(RTC_YEAR); #ifdef USE_RTC_CENTURY ct.year += readrtc(RTC_CENTURY) * 100; #else ct.year += (ct.year < 80 ? 2000 : 1900); #endif critical_exit(); + mtx_unlock(&atrtc_time_lock); /* Set dow = -1 because some clocks don't set it correctly. */ ct.dow = -1; return (clock_ct_to_ts(&ct, ts)); } static device_method_t atrtc_methods[] = { /* Device interface */ DEVMETHOD(device_probe, atrtc_probe), DEVMETHOD(device_attach, atrtc_attach), DEVMETHOD(device_detach, bus_generic_detach), DEVMETHOD(device_shutdown, bus_generic_shutdown), DEVMETHOD(device_suspend, bus_generic_suspend), /* XXX stop statclock? */ DEVMETHOD(device_resume, atrtc_resume), /* clock interface */ DEVMETHOD(clock_gettime, atrtc_gettime), DEVMETHOD(clock_settime, atrtc_settime), { 0, 0 } }; static driver_t atrtc_driver = { "atrtc", atrtc_methods, sizeof(struct atrtc_softc), }; static devclass_t atrtc_devclass; DRIVER_MODULE(atrtc, isa, atrtc_driver, atrtc_devclass, 0, 0); DRIVER_MODULE(atrtc, acpi, atrtc_driver, atrtc_devclass, 0, 0); #include "opt_ddb.h" #ifdef DDB #include DB_SHOW_COMMAND(rtc, rtc) { printf("%02x/%02x/%02x %02x:%02x:%02x, A = %02x, B = %02x, C = %02x\n", rtcin(RTC_YEAR), rtcin(RTC_MONTH), rtcin(RTC_DAY), rtcin(RTC_HRS), rtcin(RTC_MIN), rtcin(RTC_SEC), rtcin(RTC_STATUSA), rtcin(RTC_STATUSB), rtcin(RTC_INTR)); } #endif /* DDB */ Index: stable/11 =================================================================== --- stable/11 (revision 323446) +++ stable/11 (revision 323447) Property changes on: stable/11 ___________________________________________________________________ Modified: svn:mergeinfo ## -0,0 +0,1 ## Merged /head:r320901-320902,320996-320997,321002,321048,321400,321743,321745