radeon 0000:01:00.0: swiotlb buffer is full (sz: 2097152 bytes)

radeon 0000:01:00.0: swiotlb buffer is full (sz: 2097152 bytes)

[Borislav Petkov]

Hi guys,

seen this already?

I see it has happened during resume. Kernel is rc7+tip/master.

Box is stable otherwise while I'm working on it...

[66861.818432] usb 10-1: USB disconnect, device number 2
[75380.827447] perf: interrupt took too long (2527 > 2500), lowering kernel.perf_event_max_sample_rate to 79000
[94022.728431] radeon 0000:01:00.0: swiotlb buffer is full (sz: 2097152 bytes)
[94022.735717] swiotlb: coherent allocation failed for device 0000:01:00.0 size=2097152
[94022.743525] CPU: 2 PID: 3069 Comm: Xorg Not tainted 4.15.0-rc7+ #4
[94022.749711] Hardware name: To be filled by O.E.M. To be filled by O.E.M./M5A97 EVO R2.0, BIOS 1503 01/16/2013
[94022.749711] Call Trace:
[94022.749717]  dump_stack+0x67/0x8f
[94022.749720]  swiotlb_alloc_coherent+0x169/0x170
[94022.749726]  ttm_dma_pool_get_pages+0x1ea/0x450 [ttm]
[94022.749731]  ttm_dma_populate+0x248/0x330 [ttm]
[94022.749734]  ttm_tt_bind+0x23/0x50 [ttm]
[94022.749737]  ttm_bo_handle_move_mem+0x3a1/0x3e0 [ttm]
[94022.749741]  ? ttm_bo_mem_space+0x3bc/0x4a0 [ttm]
[94022.749744]  ttm_bo_validate+0x139/0x150 [ttm]
[94022.749746]  ? _raw_write_unlock+0x12/0x30
[94022.749748]  ? drm_vma_offset_add+0x6a/0x90
[94022.749751]  ttm_bo_init_reserved+0x3a5/0x470 [ttm]
[94022.749754]  ttm_bo_init+0x4d/0xb0 [ttm]
[94022.749778]  ? radeon_update_memory_usage.isra.0+0x60/0x60 [radeon]
[94022.749784]  ? drm_gem_object_init+0x31/0x50
[94022.749796]  radeon_bo_create+0x1bf/0x290 [radeon]
[94022.749809]  ? radeon_update_memory_usage.isra.0+0x60/0x60 [radeon]
[94022.749822]  radeon_gem_object_create+0xa9/0x1b0 [radeon]
[94022.749835]  ? radeon_gem_pwrite_ioctl+0x30/0x30 [radeon]
[94022.749848]  radeon_gem_create_ioctl+0x6a/0xf0 [radeon]
[94022.749862]  ? radeon_gem_pwrite_ioctl+0x30/0x30 [radeon]
[94022.749863]  drm_ioctl_kernel+0x6e/0xd0
[94022.749865]  ? unix_state_double_unlock+0x30/0x30
[94022.749866]  drm_ioctl+0x33b/0x3f0
[94022.749879]  ? radeon_gem_pwrite_ioctl+0x30/0x30 [radeon]
[94022.749881]  ? preempt_count_sub+0xa8/0x100
[94022.749882]  ? _raw_spin_unlock_irqrestore+0x25/0x50
[94022.749883]  ? preempt_count_sub+0xa8/0x100
[94022.749894]  radeon_drm_ioctl+0x5d/0xa0 [radeon]
[94022.749896]  do_vfs_ioctl+0xa2/0x600
[94022.749898]  ? __fget+0x67/0xb0
[94022.749899]  SyS_ioctl+0x4c/0x90
[94022.749901]  entry_SYSCALL_64_fastpath+0x22/0x8a
[94022.749902] RIP: 0033:0x7f28b58145e7
[94022.749903] RSP: 002b:00007ffdac6c6948 EFLAGS: 00000246
[94022.770322] radeon 0000:01:00.0: swiotlb buffer is full (sz: 2097152 bytes)
[94022.770323] swiotlb: coherent allocation failed for device 0000:01:00.0 size=2097152

-- 
Regards/Gruss,
    Boris.

Good mailing practices for 400: avoid top-posting and trim the reply.

RE: radeon 0000:01:00.0: swiotlb buffer is full (sz: 2097152 bytes)

[Deucher, Alexander]

> -----Original Message-----
> From: Borislav Petkov [mailto:bp@alien8.de]
> Sent: Wednesday, January 17, 2018 4:04 PM
> To: Deucher, Alexander <Alexander.Deucher@amd.com>; Koenig, Christian
> <Christian.Koenig@amd.com>
> Cc: lkml <linux-kernel@vger.kernel.org>
> Subject: radeon 0000:01:00.0: swiotlb buffer is full (sz: 2097152 bytes)
> 
> Hi guys,
> 
> seen this already?
> 
> I see it has happened during resume. Kernel is rc7+tip/master.
> 
> Box is stable otherwise while I'm working on it...

Being investigated here:
https://bugs.freedesktop.org/show_bug.cgi?id=104082

Alex

> 
> [66861.818432] usb 10-1: USB disconnect, device number 2 [75380.827447]
> perf: interrupt took too long (2527 > 2500), lowering
> kernel.perf_event_max_sample_rate to 79000 [94022.728431] radeon
> 0000:01:00.0: swiotlb buffer is full (sz: 2097152 bytes) [94022.735717] swiotlb:
> coherent allocation failed for device 0000:01:00.0 size=2097152
> [94022.743525] CPU: 2 PID: 3069 Comm: Xorg Not tainted 4.15.0-rc7+ #4
> [94022.749711] Hardware name: To be filled by O.E.M. To be filled by
> O.E.M./M5A97 EVO R2.0, BIOS 1503 01/16/2013 [94022.749711] Call Trace:
> [94022.749717]  dump_stack+0x67/0x8f
> [94022.749720]  swiotlb_alloc_coherent+0x169/0x170
> [94022.749726]  ttm_dma_pool_get_pages+0x1ea/0x450 [ttm]
> [94022.749731]  ttm_dma_populate+0x248/0x330 [ttm] [94022.749734]
> ttm_tt_bind+0x23/0x50 [ttm] [94022.749737]
> ttm_bo_handle_move_mem+0x3a1/0x3e0 [ttm] [94022.749741]  ?
> ttm_bo_mem_space+0x3bc/0x4a0 [ttm] [94022.749744]
> ttm_bo_validate+0x139/0x150 [ttm] [94022.749746]  ?
> _raw_write_unlock+0x12/0x30 [94022.749748]  ?
> drm_vma_offset_add+0x6a/0x90 [94022.749751]
> ttm_bo_init_reserved+0x3a5/0x470 [ttm] [94022.749754]
> ttm_bo_init+0x4d/0xb0 [ttm] [94022.749778]  ?
> radeon_update_memory_usage.isra.0+0x60/0x60 [radeon] [94022.749784]  ?
> drm_gem_object_init+0x31/0x50 [94022.749796]
> radeon_bo_create+0x1bf/0x290 [radeon] [94022.749809]  ?
> radeon_update_memory_usage.isra.0+0x60/0x60 [radeon] [94022.749822]
> radeon_gem_object_create+0xa9/0x1b0 [radeon] [94022.749835]  ?
> radeon_gem_pwrite_ioctl+0x30/0x30 [radeon] [94022.749848]
> radeon_gem_create_ioctl+0x6a/0xf0 [radeon] [94022.749862]  ?
> radeon_gem_pwrite_ioctl+0x30/0x30 [radeon] [94022.749863]
> drm_ioctl_kernel+0x6e/0xd0 [94022.749865]  ?
> unix_state_double_unlock+0x30/0x30
> [94022.749866]  drm_ioctl+0x33b/0x3f0
> [94022.749879]  ? radeon_gem_pwrite_ioctl+0x30/0x30 [radeon]
> [94022.749881]  ? preempt_count_sub+0xa8/0x100 [94022.749882]  ?
> _raw_spin_unlock_irqrestore+0x25/0x50
> [94022.749883]  ? preempt_count_sub+0xa8/0x100 [94022.749894]
> radeon_drm_ioctl+0x5d/0xa0 [radeon] [94022.749896]
> do_vfs_ioctl+0xa2/0x600 [94022.749898]  ? __fget+0x67/0xb0 [94022.749899]
> SyS_ioctl+0x4c/0x90 [94022.749901]  entry_SYSCALL_64_fastpath+0x22/0x8a
> [94022.749902] RIP: 0033:0x7f28b58145e7
> [94022.749903] RSP: 002b:00007ffdac6c6948 EFLAGS: 00000246 [94022.770322]
> radeon 0000:01:00.0: swiotlb buffer is full (sz: 2097152 bytes) [94022.770323]
> swiotlb: coherent allocation failed for device 0000:01:00.0 size=2097152
> 
> --
> Regards/Gruss,
>     Boris.
> 
> Good mailing practices for 400: avoid top-posting and trim the reply.