Skip to contentSkip to conversation

4.18-rc1 i915 dma mapping warning

DISCUSSION Joonas Lahtinen 2 participants 2 messages Jun 19, 2018 1 message not in the archive

lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

#1 lore raw

+ Chris,

Somehow this message managed to dodge the mailing list?

Regards, Joonas

Quoting Dave Jones (2018-06-19 05:52:23)

Quoted 41 lines

The new DMA mapping debug option in 4.18-rc1 (CONFIG_DMA_API_DEBUG_SG) seems to dislike something about i915..

[    1.203923] i915 0000:00:02.0: DMA-API: mapping sg segment longer than device claims to support [len=69632] [max=65536]
[    1.203968] WARNING: CPU: 1 PID: 1 at lib/dma-debug.c:1301 debug_dma_map_sg+0x26f/0x3f0
[    1.203989] CPU: 1 PID: 1 Comm: swapper/0 Not tainted 4.18.0-rc1 #1
[    1.204008] Hardware name: Shuttle Inc. SH87R/FH87, BIOS 2.03 06/19/2014
[    1.204025] RIP: 0010:debug_dma_map_sg+0x26f/0x3f0
[    1.204038] Code: 08 48 8b 54 24 10 4c 8b 5c 24 18 4c 8b 4c 24 20 8b 4c 24 2c 48 c7 c7 a8 80 01 99 4c 89 4c 24 10 4c 89 5c 24 08 e8 51 58 ca ff <0f> 0b 4c 8b 4c 24 10 4c 8b 5c 24 08 8b 35 e7 27 f6 00 85 f6 0f 85 
[    1.204175] RSP: 0000:ffffa11a00023b10 EFLAGS: 00010246
[    1.204190] RAX: 0000000000000000 RBX: ffff9782d0922008 RCX: 0000000000000000
[    1.204207] RDX: 0000000000000001 RSI: ffffffff980b94cf RDI: ffffffff980b94cf
[    1.204225] RBP: ffff9782d2b7c5e8 R08: 0000000000000001 R09: 0000000000000001
[    1.204242] R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000000
[    1.204259] R13: 0000000000000001 R14: 0000000087654321 R15: ffff9782d57f58c8
[    1.204277] FS:  0000000000000000(0000) GS:ffff9782d6400000(0000) knlGS:0000000000000000
[    1.204296] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[    1.204311] CR2: 0000000000000000 CR3: 000000013f224001 CR4: 00000000000606e0
[    1.204328] Call Trace:
[    1.204340]  i915_gem_gtt_prepare_pages+0x78/0xd0
[    1.204354]  i915_gem_object_get_pages_gtt+0x3e9/0x790
[    1.204368]  __i915_gem_object_get_pages+0x58/0x70
[    1.204382]  i915_gem_object_set_to_gtt_domain+0x51/0x100
[    1.204398]  intel_ring_context_pin+0xe4/0x230
[    1.204411]  i915_request_alloc+0x51/0x400
[    1.204423]  i915_gem_init+0x2c9/0x480
[    1.204436]  i915_driver_load+0x9ad/0xec0
[    1.204449]  ? _raw_spin_unlock_irqrestore+0x3f/0x70
[    1.204463]  pci_device_probe+0xa8/0x130
[    1.204477]  driver_probe_device+0x29d/0x330
[    1.204489]  __driver_attach+0xc0/0xd0
[    1.204501]  ? driver_probe_device+0x330/0x330
[    1.204514]  bus_for_each_dev+0x6b/0xc0
[    1.204525]  bus_add_driver+0x11d/0x210
[    1.204538]  ? mipi_dsi_bus_init+0x11/0x11
[    1.204551]  driver_register+0x5b/0xe0
[    1.204563]  do_one_initcall+0x10b/0x345
[    1.204575]  ? do_early_param+0x8e/0x8e
[    1.204588]  ? rcu_read_lock_sched_held+0x66/0x70
[    1.204602]  ? do_early_param+0x8e/0x8e
[    1.204613]  kernel_init_freeable+0x1c6/0x24c
[    1.204627]  ? rest_init+0xc0/0xc0
Quoted 9 lines
[    1.204637]  kernel_init+0xa/0x100
[    1.204648]  ret_from_fork+0x3a/0x50
[    1.204659] irq event stamp: 1452970
[    1.204672] hardirqs last  enabled at (1452969): [<ffffffff980b8c47>] console_unlock+0x4f7/0x680
[    1.204693] hardirqs last disabled at (1452970): [<ffffffff98a01036>] error_entry+0x86/0x100
[    1.204715] softirqs last  enabled at (1448282): [<ffffffff98c00409>] __do_softirq+0x409/0x551
[    1.204737] softirqs last disabled at (1448143): [<ffffffff9805e9a5>] irq_exit+0xc5/0xd0
[    1.204804] ---[ end trace e7fa4b5bd5f1190a ]---

_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

Reviewed-by Sumit Semwal Acked-by Marek Szyprowski
#2 lore raw

Quoting Joonas Lahtinen (2018-06-19 14:16:08)

Quoted 8 lines

+ Chris,

Somehow this message managed to dodge the mailing list?

Regards, Joonas

Quoting Dave Jones (2018-06-19 05:52:23)

Quoted 1 line

The new DMA mapping debug option in 4.18-rc1 (CONFIG_DMA_API_DEBUG_SG) seems to dislike something about i915..

See

commit 002edb6f6f2a79bea50de11260ddc9572e6db731

Author: Robin Murphy <[email protected]>
Date:   Fri Nov 6 16:32:51 2015 -0800

dma-mapping: tidy up dma_parms default handling

Many DMA controllers and other devices set max_segment_size to indicate their scatter-gather capability, but have no interest in segment_boundary_mask. However, the existence of a dma_parms structure precludes the use of any default value, leaving them as zeros (assuming a properly kzalloc'ed structure). If a well-behaved IOMMU (or SWIOTLB) then tries to respect this by ensuring a mapped segment does not cross a zero-byte boundary, hilarity ensues.

Since zero is a nonsensical value for either parameter, treat it as an indicator for "default", as might be expected. In the process, clean up a bit by replacing the bare constants with slightly more meaningful macros and removing the superfluous "else" statements.

for no explanation whatever for the magical value. -Chris
_______________________________________________
Intel-gfx mailing list [email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx