[Devel] [PATCH RHEL9 COMMIT] DMA-API: exceeded 7 overlapping mappings of cacheline share
Konstantin Khorenko
khorenko at virtuozzo.com
Wed Oct 20 11:40:42 MSK 2021
The commit is pushed to "branch-rh9-5.14.vz9.1.x-ovz" and will appear at https://src.openvz.org/scm/ovz/vzkernel.git
after rh9-5.14.0-4.vz9.10.12
------>
commit af0a409211edf348ffcdc44fd04c915274271966
Author: Michael S. Tsirkin <mst at redhat.com>
Date: Wed Oct 20 11:40:42 2021 +0300
DMA-API: exceeded 7 overlapping mappings of cacheline share
OK, so I enabled CONFIG_DMA_API_DEBUG: and now I get:
[ 178.789451] ------------[ cut here ]------------
[ 178.789558] DMA-API: exceeded 7 overlapping mappings of cacheline 0x000000001a161a80
[ 178.789578] WARNING: CPU: 7 PID: 1223 at kernel/dma/debug.c:523 add_dma_entry+0x1f6/0x200
[ 178.789580] Modules linked in: kvm_intel nouveau kvm iwlmvm iwlwifi
[ 178.789592] CPU: 7 PID: 1223 Comm: Xorg Tainted: G W 4.20.0-rc3-guest-debug+ #183
[ 178.789595] Hardware name: LENOVO 20MDCTO1WW/20MDCTO1WW, BIOS N2EET31W (1.13 ) 10/19/2018
[ 178.789599] RIP: 0010:add_dma_entry+0x1f6/0x200
[ 178.789603] Code: eb 9e 7e ed 80 3d 17 17 18 03 00 75 e4 48 8d 54 24 28 be 07 00 00 00 48 c7 c7 40 38 e9 82 c6 05 fd 16 18 03 01 e8 74 87 f2 ff <0f> 0b eb c3 e8 81 83 f2 ff 90 0f 1f 44 00 00 80 3d 31 20 1e 03 00
[ 178.789606] RSP: 0018:ffff888733227348 EFLAGS: 00010086
[ 178.789609] RAX: 0000000000000000 RBX: 00000000ffffffef RCX: ffffffff8129f98e
[ 178.789612] RDX: 0000000000000000 RSI: 0000000000000008 RDI: ffff8888413de840
[ 178.789615] RBP: 000000001a161a80 R08: ffffed110827bd09 R09: ffffed110827bd08
[ 178.789618] R10: ffffed110827bd08 R11: ffff8888413de847 R12: 0000000000000202
[ 178.789621] R13: 1ffff110e6644e6a R14: ffffffff85a9b6c8 R15: ffff88877ca53528
[ 178.789624] FS: 00007efd13337ac0(0000) GS:ffff8888413c0000(0000) knlGS:0000000000000000
[ 178.789627] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 178.789630] CR2: 00007f6fb654b024 CR3: 000000073bef6003 CR4: 00000000003606e0
[ 178.789632] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[ 178.789635] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[ 178.789638] Call Trace:
[ 178.789646] ? active_cacheline_set_overlap.part.21+0x50/0x50
[ 178.789655] ? i915_gem_do_execbuffer+0x768/0x1c00
[ 178.789665] debug_dma_map_sg+0xb2/0x450
[ 178.789677] i915_gem_gtt_prepare_pages+0xdf/0x160
[ 178.789688] __i915_gem_userptr_alloc_pages+0x83/0x190
[ 178.789696] i915_gem_userptr_get_pages+0x1e7/0x2d0
[ 178.789704] __i915_gem_object_get_pages+0x71/0x80
[ 178.789710] __i915_vma_do_pin+0x65d/0xa30
[ 178.789724] ? i915_gem_valid_gtt_space+0xd0/0xd0
[ 178.789737] eb_lookup_vmas+0xd2c/0x1570
[ 178.789753] ? eb_relocate_vma+0x2c0/0x2c0
[ 178.789759] ? __i915_gem_object_unset_pages+0x3a0/0x3a0
[ 178.789764] ? _raw_spin_unlock_irqrestore+0x43/0x50
[ 178.789771] ? lockdep_hardirqs_on+0x17a/0x250
[ 178.789778] ? __pm_runtime_resume+0x59/0xb0
[ 178.789787] i915_gem_do_execbuffer+0x768/0x1c00
[ 178.789793] ? __kmalloc_node+0x1b1/0x440
[ 178.789797] ? i915_gem_execbuffer2_ioctl+0x12b/0x550
[ 178.789801] ? drm_ioctl_kernel+0x14a/0x1a0
[ 178.789804] ? drm_ioctl+0x311/0x550
[ 178.789815] ? eb_relocate_slow+0x610/0x610
[ 178.789824] ? i915_gem_execbuffer2_ioctl+0x12b/0x550
[ 178.789840] ? match_held_lock+0x1b/0x230
[ 178.789847] ? find_held_lock+0xac/0xd0
[ 178.789856] ? __might_fault+0x7d/0xe0
[ 178.789876] i915_gem_execbuffer2_ioctl+0x316/0x550
[ 178.789887] ? i915_gem_execbuffer_ioctl+0x4e0/0x4e0
[ 178.789900] ? i915_gem_execbuffer_ioctl+0x4e0/0x4e0
[ 178.789904] drm_ioctl_kernel+0x14a/0x1a0
[ 178.789911] ? drm_setversion+0x330/0x330
[ 178.789924] drm_ioctl+0x311/0x550
[ 178.789931] ? i915_gem_execbuffer_ioctl+0x4e0/0x4e0
[ 178.789936] ? drm_ioctl_kernel+0x1a0/0x1a0
[ 178.789942] ? mark_held_locks+0xb0/0xb0
[ 178.789960] ? __fget+0x208/0x310
[ 178.789963] ? match_held_lock+0x1b/0x230
[ 178.789967] ? __lock_is_held+0x38/0xc0
[ 178.789978] do_vfs_ioctl+0x137/0x9d0
[ 178.789985] ? ioctl_preallocate+0x140/0x140
[ 178.789990] ? selinux_file_ioctl+0x2a2/0x350
[ 178.789998] ? selinux_file_mprotect+0x240/0x240
[ 178.790004] ? __fget+0x208/0x310
[ 178.790012] ? __ia32_sys_dup2+0x200/0x200
[ 178.790029] ksys_ioctl+0x60/0x90
[ 178.790032] ? lockdep_hardirqs_off+0xb5/0x100
[ 178.790039] __x64_sys_ioctl+0x3d/0x50
[ 178.790045] do_syscall_64+0x68/0x210
[ 178.790051] entry_SYSCALL_64_after_hwframe+0x49/0xbe
[ 178.790055] RIP: 0033:0x7efd1385cfeb
[ 178.790059] Code: 0f 1e fa 48 8b 05 9d be 0c 00 64 c7 00 26 00 00 00 48 c7 c0 ff ff ff ff c3 66 0f 1f 44 00 00 f3 0f 1e fa b8 10 00 00 00 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d 6d be 0c 00 f7 d8 64 89 01 48
[ 178.790062] RSP: 002b:00007ffd768ff718 EFLAGS: 00000246 ORIG_RAX: 0000000000000010
[ 178.790066] RAX: ffffffffffffffda RBX: 000000000000000d RCX: 00007efd1385cfeb
[ 178.790068] RDX: 00007ffd768ff750 RSI: 0000000040406469 RDI: 000000000000000d
[ 178.790071] RBP: 00005651b6f38500 R08: 0000000000000002 R09: 0000000000000001
[ 178.790074] R10: 00007efd051ef000 R11: 0000000000000246 R12: 0000000000001000
[ 178.790077] R13: 00007efd10527000 R14: 00007ffd768ff750 R15: 00007efd105274a0
[ 178.790108] irq event stamp: 4678142
[ 178.790113] hardirqs last enabled at (4678141): [<ffffffff82899123>] _raw_spin_unlock_irqrestore+0x43/0x50
[ 178.790116] hardirqs last disabled at (4678142): [<ffffffff82898ee2>] _raw_spin_lock_irqsave+0x12/0x50
[ 178.790120] softirqs last enabled at (4678074): [<ffffffff81bd84cf>] i915_request_add+0x53f/0x8a0
[ 178.790124] softirqs last disabled at (4678068): [<ffffffff81bd840a>] i915_request_add+0x47a/0x8a0
[ 178.790126] ---[ end trace 1d0f05369f2675fe ]---
Using a warning here makes all kind of userspace tools jumpy
and IIUC all this is trying to tell us is that there are not
enough bits to track mappings reliably.
This was also reported here:
https://bugs.freedesktop.org/show_bug.cgi?id=100455
So can we convert it to a printk?
Signed-off-by: Michael S. Tsirkin <mst at redhat.com>
https://lkml.org/lkml/2018/11/27/449
Mainstream did not accept the patch, but we need it as
WARN makes our autotests unhappy.
Faced this while tring to reproduce
https://jira.sw.ru/browse/PSBM-93919
Signed-off-by: Konstantin Khorenko <khorenko at virtuozzo.com>
Rebase to vz8:
- code moved to kernel/dma/debug.c;
- decrease code offsets
https://jira.sw.ru/browse/PSBM-127846
(cherry-picked from vz7 commit eb9c64c3e00d ("DMA-API: exceeded 7
overlapping mappings of cacheline share"))
Signed-off-by: Pavel Tikhomirov <ptikhomirov at virtuozzo.com>
(cherry picked from vz8 commit a60ed8dff6132704c3fdbbf3d97b4de9be58c0eb)
Signed-off-by: Andrey Zhadchenko <andrey.zhadchenko at virtuozzo.com>
---
kernel/dma/debug.c | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/kernel/dma/debug.c b/kernel/dma/debug.c
index dadae6255d05..d8fceb1e85fe 100644
--- a/kernel/dma/debug.c
+++ b/kernel/dma/debug.c
@@ -495,7 +495,10 @@ static void active_cacheline_inc_overlap(phys_addr_t cln)
/* If we overflowed the overlap counter then we're potentially
* leaking dma-mappings.
*/
- WARN_ONCE(overlap > ACTIVE_CACHELINE_MAX_OVERLAP,
+ if (overlap <= ACTIVE_CACHELINE_MAX_OVERLAP)
+ return;
+
+ printk_once(KERN_WARNING
pr_fmt("exceeded %d overlapping mappings of cacheline %pa\n"),
ACTIVE_CACHELINE_MAX_OVERLAP, &cln);
}
More information about the Devel
mailing list