| Message ID | 1458215982-13405-1-git-send-email-chris@chris-wilson.co.uk (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Hi, Chris. Comment is below. On Thu, Mar 17, 2016 at 12:59 PM, Chris Wilson <chris@chris-wilson.co.uk> wrote: > vmaps are temporary kernel mappings that may be of long duration. > Reusing a vmap on an object is preferrable for a driver as the cost of > setting up the vmap can otherwise dominate the operation on the object. > However, the vmap address space is rather limited on 32bit systems and > so we add a notification for vmap pressure in order for the driver to > release any cached vmappings. > > The interface is styled after the oom-notifier where the callees are > passed a pointer to an unsigned long counter for them to indicate if they > have freed any space. > > Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> > Cc: Andrew Morton <akpm@linux-foundation.org> > Cc: David Rientjes <rientjes@google.com> > Cc: Roman Pen <r.peniaev@gmail.com> > Cc: Mel Gorman <mgorman@techsingularity.net> > Cc: linux-mm@kvack.org > Cc: linux-kernel@vger.kernel.org > --- > include/linux/vmalloc.h | 4 ++++ > mm/vmalloc.c | 22 ++++++++++++++++++++++ > 2 files changed, 26 insertions(+) > > diff --git a/include/linux/vmalloc.h b/include/linux/vmalloc.h > index d1f1d338af20..edd676b8e112 100644 > --- a/include/linux/vmalloc.h > +++ b/include/linux/vmalloc.h > @@ -187,4 +187,8 @@ pcpu_free_vm_areas(struct vm_struct **vms, int nr_vms) > #define VMALLOC_TOTAL 0UL > #endif > > +struct notitifer_block; > +int register_vmap_purge_notifier(struct notifier_block *nb); > +int unregister_vmap_purge_notifier(struct notifier_block *nb); > + > #endif /* _LINUX_VMALLOC_H */ > diff --git a/mm/vmalloc.c b/mm/vmalloc.c > index fb42a5bffe47..fd2ca94c2732 100644 > --- a/mm/vmalloc.c > +++ b/mm/vmalloc.c > @@ -21,6 +21,7 @@ > #include <linux/debugobjects.h> > #include <linux/kallsyms.h> > #include <linux/list.h> > +#include <linux/notifier.h> > #include <linux/rbtree.h> > #include <linux/radix-tree.h> > #include <linux/rcupdate.h> > @@ -344,6 +345,8 @@ static void __insert_vmap_area(struct vmap_area *va) > > static void purge_vmap_area_lazy(void); > > +static BLOCKING_NOTIFIER_HEAD(vmap_notify_list); > + > /* > * Allocate a region of KVA of the specified size and alignment, within the > * vstart and vend. > @@ -356,6 +359,7 @@ static struct vmap_area *alloc_vmap_area(unsigned long size, > struct vmap_area *va; > struct rb_node *n; > unsigned long addr; > + unsigned long freed; > int purged = 0; > struct vmap_area *first; > > @@ -468,6 +472,12 @@ overflow: > purged = 1; > goto retry; > } > + freed = 0; > + blocking_notifier_call_chain(&vmap_notify_list, 0, &freed); It seems to me that alloc_vmap_area() was designed not to sleep, at least on GFP_NOWAIT path (__GFP_DIRECT_RECLAIM is not set). But blocking_notifier_call_chain() might sleep. Roman.
On Thu, Mar 17, 2016 at 01:37:06PM +0100, Roman Peniaev wrote: > > + freed = 0; > > + blocking_notifier_call_chain(&vmap_notify_list, 0, &freed); > > It seems to me that alloc_vmap_area() was designed not to sleep, > at least on GFP_NOWAIT path (__GFP_DIRECT_RECLAIM is not set). > > But blocking_notifier_call_chain() might sleep. Indeed, I had not anticipated anybody using GFP_ATOMIC or equivalently restrictive gfp_t for vmap and yes there are such callers. Would guarding the notifier with gfp & __GFP_DIRECT_RECLAIM and !(gfp & __GFP_NORETRY) == be sufficient? Is that enough for GFP_NOFS? -Chris
On Thu, Mar 17, 2016 at 1:57 PM, Chris Wilson <chris@chris-wilson.co.uk> wrote: > On Thu, Mar 17, 2016 at 01:37:06PM +0100, Roman Peniaev wrote: >> > + freed = 0; >> > + blocking_notifier_call_chain(&vmap_notify_list, 0, &freed); >> >> It seems to me that alloc_vmap_area() was designed not to sleep, >> at least on GFP_NOWAIT path (__GFP_DIRECT_RECLAIM is not set). >> >> But blocking_notifier_call_chain() might sleep. > > Indeed, I had not anticipated anybody using GFP_ATOMIC or equivalently > restrictive gfp_t for vmap and yes there are such callers. > > Would guarding the notifier with gfp & __GFP_DIRECT_RECLAIM and > !(gfp & __GFP_NORETRY) == be sufficient? Is that enough for GFP_NOFS? I would use gfpflags_allow_blocking() for that purpose. Roman
On Thu, Mar 17, 2016 at 02:21:40PM +0100, Roman Peniaev wrote: > On Thu, Mar 17, 2016 at 1:57 PM, Chris Wilson <chris@chris-wilson.co.uk> wrote: > > On Thu, Mar 17, 2016 at 01:37:06PM +0100, Roman Peniaev wrote: > >> > + freed = 0; > >> > + blocking_notifier_call_chain(&vmap_notify_list, 0, &freed); > >> > >> It seems to me that alloc_vmap_area() was designed not to sleep, > >> at least on GFP_NOWAIT path (__GFP_DIRECT_RECLAIM is not set). > >> > >> But blocking_notifier_call_chain() might sleep. > > > > Indeed, I had not anticipated anybody using GFP_ATOMIC or equivalently > > restrictive gfp_t for vmap and yes there are such callers. > > > > Would guarding the notifier with gfp & __GFP_DIRECT_RECLAIM and > > !(gfp & __GFP_NORETRY) == be sufficient? Is that enough for GFP_NOFS? > > I would use gfpflags_allow_blocking() for that purpose. Thanks, -Chris
We're not catching it, but this gives deadlock trace when running
kms_pipe_crc_basic@suspend-read-crc-pipe-A (happens on BSW):
[ 132.555497] kms_pipe_crc_basic: starting subtest suspend-read-crc-pipe-A
[ 132.734041] PM: Syncing filesystems ... done.
[ 132.751624] Freezing user space processes ... (elapsed 0.003 seconds) done.
[ 132.755240] Freezing remaining freezable tasks ... (elapsed 0.002 seconds) done.
[ 132.758372] Suspending console(s) (use no_console_suspend to debug)
[ 132.768157] sd 0:0:0:0: [sda] Synchronizing SCSI cache
[ 132.780482] sd 0:0:0:0: [sda] Stopping disk
[ 132.889902] PM: suspend of devices complete after 129.133 msecs
[ 132.924359] PM: late suspend of devices complete after 34.440 msecs
[ 132.932433] r8169 0000:03:00.0: System wakeup enabled by ACPI
[ 132.938105] xhci_hcd 0000:00:14.0: System wakeup enabled by ACPI
[ 132.948029] PM: noirq suspend of devices complete after 23.660 msecs
[ 132.948073] ACPI: Preparing to enter system sleep state S3
[ 132.960567] PM: Saving platform NVS memory
[ 132.960803] Disabling non-boot CPUs ...
[ 132.999863] Broke affinity for irq 116
[ 133.002229] smpboot: CPU 1 is now offline
[ 133.022915] ======================================================
[ 133.022916] [ INFO: possible circular locking dependency detected ]
[ 133.022921] 4.5.0-gfxbench-Patchwork_315+ #1 Tainted: G U
[ 133.022922] -------------------------------------------------------
[ 133.022925] rtcwake/5998 is trying to acquire lock:
[ 133.022942] (s_active#6){++++.+}, at: [<ffffffff81252b10>] kernfs_remove_by_name_ns+0x40/0x90
[ 133.022943]
but task is already holding lock:
[ 133.022953] (cpu_hotplug.lock){+.+.+.}, at: [<ffffffff8107908d>] cpu_hotplug_begin+0x6d/0xc0
[ 133.022954]
which lock already depends on the new lock.
[ 133.022955]
the existing dependency chain (in reverse order) is:
[ 133.022962]
-> #3 (cpu_hotplug.lock){+.+.+.}:
[ 133.022968] [<ffffffff810ce2a1>] lock_acquire+0xb1/0x200
[ 133.022974] [<ffffffff817c3a72>] mutex_lock_nested+0x62/0x3c0
[ 133.022978] [<ffffffff81078d41>] get_online_cpus+0x61/0x80
[ 133.022983] [<ffffffff8111a2cb>] stop_machine+0x1b/0xe0
[ 133.023046] [<ffffffffa012a3ad>] gen8_ggtt_insert_entries__BKL+0x2d/0x30 [i915]
[ 133.023096] [<ffffffffa012dc06>] ggtt_bind_vma+0x46/0x70 [i915]
[ 133.023146] [<ffffffffa012f55d>] i915_vma_bind+0xed/0x260 [i915]
[ 133.023197] [<ffffffffa0137153>] i915_gem_object_do_pin+0x873/0xb20 [i915]
[ 133.023248] [<ffffffffa0137428>] i915_gem_object_pin+0x28/0x30 [i915]
[ 133.023301] [<ffffffffa014b5f5>] intel_init_pipe_control+0xb5/0x200 [i915]
[ 133.023354] [<ffffffffa01483be>] intel_logical_rings_init+0x14e/0x1080 [i915]
[ 133.023405] [<ffffffffa0137ee3>] i915_gem_init+0xf3/0x130 [i915]
[ 133.023462] [<ffffffffa01bcfeb>] i915_driver_load+0xbeb/0x1950 [i915]
[ 133.023470] [<ffffffff8151b7f4>] drm_dev_register+0xa4/0xb0
[ 133.023474] [<ffffffff8151d9de>] drm_get_pci_dev+0xce/0x1d0
[ 133.023519] [<ffffffffa00f72ff>] i915_pci_probe+0x2f/0x50 [i915]
[ 133.023525] [<ffffffff814490a5>] pci_device_probe+0x85/0xf0
[ 133.023530] [<ffffffff8153fd67>] driver_probe_device+0x227/0x440
[ 133.023534] [<ffffffff81540003>] __driver_attach+0x83/0x90
[ 133.023538] [<ffffffff8153d8b1>] bus_for_each_dev+0x61/0xa0
[ 133.023542] [<ffffffff8153f4c9>] driver_attach+0x19/0x20
[ 133.023546] [<ffffffff8153efb9>] bus_add_driver+0x1e9/0x280
[ 133.023550] [<ffffffff81540bab>] driver_register+0x5b/0xd0
[ 133.023555] [<ffffffff81447ffb>] __pci_register_driver+0x5b/0x60
[ 133.023559] [<ffffffff8151dbb6>] drm_pci_init+0xd6/0x100
[ 133.023563] [<ffffffffa0230092>] 0xffffffffa0230092
[ 133.023571] [<ffffffff810003d6>] do_one_initcall+0xa6/0x1d0
[ 133.023577] [<ffffffff8115c832>] do_init_module+0x5a/0x1c8
[ 133.023585] [<ffffffff81108d9d>] load_module+0x1efd/0x25a0
[ 133.023591] [<ffffffff81109658>] SyS_finit_module+0x98/0xc0
[ 133.023598] [<ffffffff817c82db>] entry_SYSCALL_64_fastpath+0x16/0x6f
[ 133.023603]
-> #2 (&dev->struct_mutex){+.+.+.}:
[ 133.023608] [<ffffffff810ce2a1>] lock_acquire+0xb1/0x200
[ 133.023616] [<ffffffff81516d0f>] drm_gem_mmap+0x19f/0x2a0
[ 133.023622] [<ffffffff8119adb9>] mmap_region+0x389/0x5f0
[ 133.023626] [<ffffffff8119b38a>] do_mmap+0x36a/0x420
[ 133.023632] [<ffffffff8117f03d>] vm_mmap_pgoff+0x6d/0xa0
[ 133.023638] [<ffffffff811992c3>] SyS_mmap_pgoff+0x183/0x220
[ 133.023645] [<ffffffff8100a246>] SyS_mmap+0x16/0x20
[ 133.023649] [<ffffffff817c82db>] entry_SYSCALL_64_fastpath+0x16/0x6f
[ 133.023656]
-> #1 (&mm->mmap_sem){++++++}:
[ 133.023661] [<ffffffff810ce2a1>] lock_acquire+0xb1/0x200
[ 133.023667] [<ffffffff8118fda5>] __might_fault+0x75/0xa0
[ 133.023673] [<ffffffff8125350a>] kernfs_fop_write+0x8a/0x180
[ 133.023679] [<ffffffff811d5d03>] __vfs_write+0x23/0xe0
[ 133.023684] [<ffffffff811d6ac2>] vfs_write+0xa2/0x190
[ 133.023687] [<ffffffff811d7914>] SyS_write+0x44/0xb0
[ 133.023691] [<ffffffff817c82db>] entry_SYSCALL_64_fastpath+0x16/0x6f
[ 133.023697]
-> #0 (s_active#6){++++.+}:
[ 133.023701] [<ffffffff810cd961>] __lock_acquire+0x1e81/0x1ef0
[ 133.023705] [<ffffffff810ce2a1>] lock_acquire+0xb1/0x200
[ 133.023709] [<ffffffff81251d81>] __kernfs_remove+0x241/0x320
[ 133.023713] [<ffffffff81252b10>] kernfs_remove_by_name_ns+0x40/0x90
[ 133.023717] [<ffffffff812544a0>] sysfs_remove_file_ns+0x10/0x20
[ 133.023722] [<ffffffff8153b504>] device_del+0x124/0x240
[ 133.023726] [<ffffffff8153b639>] device_unregister+0x19/0x60
[ 133.023731] [<ffffffff81545d42>] cpu_cache_sysfs_exit+0x52/0xb0
[ 133.023735] [<ffffffff81546318>] cacheinfo_cpu_callback+0x38/0x70
[ 133.023739] [<ffffffff8109bc59>] notifier_call_chain+0x39/0xa0
[ 133.023743] [<ffffffff8109bcc9>] __raw_notifier_call_chain+0x9/0x10
[ 133.023748] [<ffffffff8107900e>] cpu_notify_nofail+0x1e/0x30
[ 133.023751] [<ffffffff81079320>] _cpu_down+0x200/0x330
[ 133.023756] [<ffffffff8107987a>] disable_nonboot_cpus+0xaa/0x3b0
[ 133.023761] [<ffffffff810d47f8>] suspend_devices_and_enter+0x478/0xc30
[ 133.023765] [<ffffffff810d54c5>] pm_suspend+0x515/0x9e0
[ 133.023771] [<ffffffff810d3617>] state_store+0x77/0xe0
[ 133.023777] [<ffffffff81403eaf>] kobj_attr_store+0xf/0x20
[ 133.023781] [<ffffffff81254200>] sysfs_kf_write+0x40/0x50
[ 133.023785] [<ffffffff812535bc>] kernfs_fop_write+0x13c/0x180
[ 133.023790] [<ffffffff811d5d03>] __vfs_write+0x23/0xe0
[ 133.023794] [<ffffffff811d6ac2>] vfs_write+0xa2/0x190
[ 133.023798] [<ffffffff811d7914>] SyS_write+0x44/0xb0
[ 133.023802] [<ffffffff817c82db>] entry_SYSCALL_64_fastpath+0x16/0x6f
[ 133.023805]
other info that might help us debug this:
[ 133.023813] Chain exists of:
s_active#6 --> &dev->struct_mutex --> cpu_hotplug.lock
[ 133.023814] Possible unsafe locking scenario:
[ 133.023815] CPU0 CPU1
[ 133.023816] ---- ----
[ 133.023819] lock(cpu_hotplug.lock);
[ 133.023823] lock(&dev->struct_mutex);
[ 133.023826] lock(cpu_hotplug.lock);
[ 133.023830] lock(s_active#6);
[ 133.023831]
*** DEADLOCK ***
[ 133.023835] 8 locks held by rtcwake/5998:
[ 133.023847] #0: (sb_writers#6){.+.+.+}, at: [<ffffffff811da0f2>] __sb_start_write+0xb2/0xf0
[ 133.023855] #1: (&of->mutex){+.+.+.}, at: [<ffffffff812534e1>] kernfs_fop_write+0x61/0x180
[ 133.023863] #2: (s_active#105){.+.+.+}, at: [<ffffffff812534e9>] kernfs_fop_write+0x69/0x180
[ 133.023871] #3: (pm_mutex){+.+...}, at: [<ffffffff810d501f>] pm_suspend+0x6f/0x9e0
[ 133.023882] #4: (acpi_scan_lock){+.+.+.}, at: [<ffffffff8147c1ef>] acpi_scan_lock_acquire+0x12/0x14
[ 133.023891] #5: (cpu_add_remove_lock){+.+.+.}, at: [<ffffffff810797f4>] disable_nonboot_cpus+0x24/0x3b0
[ 133.023899] #6: (cpu_hotplug.dep_map){++++++}, at: [<ffffffff81079020>] cpu_hotplug_begin+0x0/0xc0
[ 133.023906] #7: (cpu_hotplug.lock){+.+.+.}, at: [<ffffffff8107908d>] cpu_hotplug_begin+0x6d/0xc0
[ 133.023907]
stack backtrace:
[ 133.023913] CPU: 0 PID: 5998 Comm: rtcwake Tainted: G U 4.5.0-gfxbench-Patchwork_315+ #1
[ 133.023915] Hardware name: /NUC5CPYB, BIOS PYBSWCEL.86A.0043.2015.0904.1904 09/04/2015
[ 133.023922] 0000000000000000 ffff880274967870 ffffffff81401d15 ffffffff825c52d0
[ 133.023927] ffffffff82586bd0 ffff8802749678b0 ffffffff810ca1b0 ffff880274967900
[ 133.023932] ffff880273845328 ffff880273844b00 ffff880273845440 0000000000000008
[ 133.023935] Call Trace:
[ 133.023941] [<ffffffff81401d15>] dump_stack+0x67/0x92
[ 133.023946] [<ffffffff810ca1b0>] print_circular_bug+0x1e0/0x2e0
[ 133.023949] [<ffffffff810cd961>] __lock_acquire+0x1e81/0x1ef0
[ 133.023953] [<ffffffff810ce2a1>] lock_acquire+0xb1/0x200
[ 133.023959] [<ffffffff81252b10>] ? kernfs_remove_by_name_ns+0x40/0x90
[ 133.023962] [<ffffffff81251d81>] __kernfs_remove+0x241/0x320
[ 133.023965] [<ffffffff81252b10>] ? kernfs_remove_by_name_ns+0x40/0x90
[ 133.023969] [<ffffffff812518a7>] ? kernfs_find_ns+0x97/0x140
[ 133.023972] [<ffffffff81252b10>] kernfs_remove_by_name_ns+0x40/0x90
[ 133.023975] [<ffffffff812544a0>] sysfs_remove_file_ns+0x10/0x20
[ 133.023979] [<ffffffff8153b504>] device_del+0x124/0x240
[ 133.023982] [<ffffffff810cb82d>] ? trace_hardirqs_on+0xd/0x10
[ 133.023986] [<ffffffff8153b639>] device_unregister+0x19/0x60
[ 133.023989] [<ffffffff81545d42>] cpu_cache_sysfs_exit+0x52/0xb0
[ 133.023992] [<ffffffff81546318>] cacheinfo_cpu_callback+0x38/0x70
[ 133.023995] [<ffffffff8109bc59>] notifier_call_chain+0x39/0xa0
[ 133.023999] [<ffffffff8109bcc9>] __raw_notifier_call_chain+0x9/0x10
[ 133.024002] [<ffffffff8107900e>] cpu_notify_nofail+0x1e/0x30
[ 133.024005] [<ffffffff81079320>] _cpu_down+0x200/0x330
[ 133.024011] [<ffffffff810e5d70>] ? __call_rcu.constprop.58+0x2f0/0x2f0
[ 133.024014] [<ffffffff810e5dd0>] ? call_rcu_bh+0x20/0x20
[ 133.024019] [<ffffffff810e1830>] ? trace_raw_output_rcu_utilization+0x60/0x60
[ 133.024023] [<ffffffff810e1830>] ? trace_raw_output_rcu_utilization+0x60/0x60
[ 133.024027] [<ffffffff8107987a>] disable_nonboot_cpus+0xaa/0x3b0
[ 133.024031] [<ffffffff810d47f8>] suspend_devices_and_enter+0x478/0xc30
[ 133.024035] [<ffffffff810d54c5>] pm_suspend+0x515/0x9e0
[ 133.024038] [<ffffffff810d3617>] state_store+0x77/0xe0
[ 133.024043] [<ffffffff81403eaf>] kobj_attr_store+0xf/0x20
[ 133.024046] [<ffffffff81254200>] sysfs_kf_write+0x40/0x50
[ 133.024049] [<ffffffff812535bc>] kernfs_fop_write+0x13c/0x180
[ 133.024054] [<ffffffff811d5d03>] __vfs_write+0x23/0xe0
[ 133.024059] [<ffffffff810c7882>] ? percpu_down_read+0x52/0x90
[ 133.024062] [<ffffffff811da0f2>] ? __sb_start_write+0xb2/0xf0
[ 133.024065] [<ffffffff811da0f2>] ? __sb_start_write+0xb2/0xf0
[ 133.024069] [<ffffffff811d6ac2>] vfs_write+0xa2/0x190
[ 133.024073] [<ffffffff811f549a>] ? __fget_light+0x6a/0x90
[ 133.024076] [<ffffffff811d7914>] SyS_write+0x44/0xb0
[ 133.024080] [<ffffffff817c82db>] entry_SYSCALL_64_fastpath+0x16/0x6f
[ 133.027055] ACPI: Low-level resume complete
[ 133.027294] PM: Restoring platform NVS memory
On Tue, Mar 29, 2016 at 08:34:36AM +0000, Patchwork wrote:
> == Series Details ==
>
> Series: series starting with [v3] mm/vmap: Add a notifier for when we run out of vmap address space (rev3)
> URL : https://patchwork.freedesktop.org/series/4569/
> State : failure
>
> == Summary ==
>
> Series 4569v3 Series without cover letter
> http://patchwork.freedesktop.org/api/1.0/series/4569/revisions/3/mbox/
>
> Test kms_flip:
> Subgroup basic-flip-vs-wf_vblank:
> pass -> FAIL (snb-x220t)
> Test pm_rpm:
> Subgroup basic-pci-d3-state:
> pass -> DMESG-WARN (bsw-nuc-2)
> Subgroup basic-rte:
> dmesg-warn -> PASS (byt-nuc) UNSTABLE
>
> bdw-nuci7 total:192 pass:179 dwarn:0 dfail:0 fail:1 skip:12
> bdw-ultra total:192 pass:170 dwarn:0 dfail:0 fail:1 skip:21
> bsw-nuc-2 total:192 pass:154 dwarn:1 dfail:0 fail:0 skip:37
> byt-nuc total:192 pass:157 dwarn:0 dfail:0 fail:0 skip:35
> hsw-brixbox total:192 pass:170 dwarn:0 dfail:0 fail:0 skip:22
> hsw-gt2 total:192 pass:175 dwarn:0 dfail:0 fail:0 skip:17
> ivb-t430s total:192 pass:167 dwarn:0 dfail:0 fail:0 skip:25
> skl-i7k-2 total:192 pass:169 dwarn:0 dfail:0 fail:0 skip:23
> skl-nuci5 total:192 pass:181 dwarn:0 dfail:0 fail:0 skip:11
> snb-dellxps total:192 pass:158 dwarn:0 dfail:0 fail:0 skip:34
> snb-x220t total:192 pass:157 dwarn:0 dfail:0 fail:2 skip:33
>
> Results at /archive/results/CI_IGT_test/Patchwork_1725/
>
> f5d413cccefa1f93d64c34f357151d42add63a84 drm-intel-nightly: 2016y-03m-24d-14h-34m-29s UTC integration manifest
> d9f9cda1b4e8a64ad1ac9bef0392e2c701b0d9f7 drm/i915/shrinker: Hook up vmap allocation failure notifier
> 93698e6141bccbabfc898a7d2cad5577f5af893b mm/vmap: Add a notifier for when we run out of vmap address space
>
> _______________________________________________
> Intel-gfx mailing list
> Intel-gfx@lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/intel-gfx
On Tue, Mar 29, 2016 at 03:51:12PM +0300, Marius Vlad wrote: > We're not catching it, but this gives deadlock trace when running > kms_pipe_crc_basic@suspend-read-crc-pipe-A (happens on BSW): > [ 133.022915] ====================================================== > [ 133.022916] [ INFO: possible circular locking dependency detected ] > [ 133.022921] 4.5.0-gfxbench-Patchwork_315+ #1 Tainted: G U > [ 133.022922] ------------------------------------------------------- That's the warning (just boring impossible warning) that the patch to move the kernfs locking around to avoid nesting mmap_sem inside it resolves. -Chris
diff --git a/include/linux/vmalloc.h b/include/linux/vmalloc.h index d1f1d338af20..edd676b8e112 100644 --- a/include/linux/vmalloc.h +++ b/include/linux/vmalloc.h @@ -187,4 +187,8 @@ pcpu_free_vm_areas(struct vm_struct **vms, int nr_vms) #define VMALLOC_TOTAL 0UL #endif +struct notitifer_block; +int register_vmap_purge_notifier(struct notifier_block *nb); +int unregister_vmap_purge_notifier(struct notifier_block *nb); + #endif /* _LINUX_VMALLOC_H */ diff --git a/mm/vmalloc.c b/mm/vmalloc.c index fb42a5bffe47..fd2ca94c2732 100644 --- a/mm/vmalloc.c +++ b/mm/vmalloc.c @@ -21,6 +21,7 @@ #include <linux/debugobjects.h> #include <linux/kallsyms.h> #include <linux/list.h> +#include <linux/notifier.h> #include <linux/rbtree.h> #include <linux/radix-tree.h> #include <linux/rcupdate.h> @@ -344,6 +345,8 @@ static void __insert_vmap_area(struct vmap_area *va) static void purge_vmap_area_lazy(void); +static BLOCKING_NOTIFIER_HEAD(vmap_notify_list); + /* * Allocate a region of KVA of the specified size and alignment, within the * vstart and vend. @@ -356,6 +359,7 @@ static struct vmap_area *alloc_vmap_area(unsigned long size, struct vmap_area *va; struct rb_node *n; unsigned long addr; + unsigned long freed; int purged = 0; struct vmap_area *first; @@ -468,6 +472,12 @@ overflow: purged = 1; goto retry; } + freed = 0; + blocking_notifier_call_chain(&vmap_notify_list, 0, &freed); + if (freed > 0) { + purged = 0; + goto retry; + } if (printk_ratelimit()) pr_warn("vmap allocation for size %lu failed: " "use vmalloc=<size> to increase size.\n", size); @@ -475,6 +485,18 @@ overflow: return ERR_PTR(-EBUSY); } +int register_vmap_purge_notifier(struct notifier_block *nb) +{ + return blocking_notifier_chain_register(&vmap_notify_list, nb); +} +EXPORT_SYMBOL_GPL(register_vmap_purge_notifier); + +int unregister_vmap_purge_notifier(struct notifier_block *nb) +{ + return blocking_notifier_chain_unregister(&vmap_notify_list, nb); +} +EXPORT_SYMBOL_GPL(unregister_vmap_purge_notifier); + static void __free_vmap_area(struct vmap_area *va) { BUG_ON(RB_EMPTY_NODE(&va->rb_node));
vmaps are temporary kernel mappings that may be of long duration. Reusing a vmap on an object is preferrable for a driver as the cost of setting up the vmap can otherwise dominate the operation on the object. However, the vmap address space is rather limited on 32bit systems and so we add a notification for vmap pressure in order for the driver to release any cached vmappings. The interface is styled after the oom-notifier where the callees are passed a pointer to an unsigned long counter for them to indicate if they have freed any space. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: David Rientjes <rientjes@google.com> Cc: Roman Pen <r.peniaev@gmail.com> Cc: Mel Gorman <mgorman@techsingularity.net> Cc: linux-mm@kvack.org Cc: linux-kernel@vger.kernel.org --- include/linux/vmalloc.h | 4 ++++ mm/vmalloc.c | 22 ++++++++++++++++++++++ 2 files changed, 26 insertions(+)