DMA-API mapping errors in vmw_pvscsi

DMA-API mapping errors in vmw_pvscsi

[Josh Boyer]

Hi All,

We've had a few reports[1][2] now on the vmw_pvscsi driver throwing
DMA-API mapping errors when the DMA_API_DEBUG option is enabled.  I've
included one from a recent 3.14-rc6 kernel below.  Looking at
vmw_pvscsi.c, I can see pvscsi_map_buffers is missing the calls to
pci_dma_mapping_error, which is what causes the warnings to be thrown.

However, I'm not familiar with this driver and I can't see what the
proper error path should be in this case.  pvscsi_map_buffers is a
void function and doesn't currently have the ability to return an
error to the caller.  Even if it did, I'm not sure what the proper
response to an error should be.  Thoughts?

josh

[    2.962772] ------------[ cut here ]------------
[    2.963764] WARNING: CPU: 1 PID: 6 at lib/dma-debug.c:1140
check_unmap+0x4ee/0x9e0()
[    2.965382] vmw_pvscsi 0000:03:00.0: DMA-API: device driver failed
to check map error[device address=0x0000000078520f80] [size=96 bytes]
[mapped as single]
[    2.968214] Modules linked in:
[    2.968897]  vmwgfx(+) ttm drm ata_generic vmw_pvscsi(+) i2c_core pata_acpi
[    2.970302] CPU: 1 PID: 6 Comm: kworker/u4:0 Not tainted
3.14.0-0.rc6.git4.1.fc21.x86_64 #1
[    2.972028] Hardware name: VMware, Inc. VMware Virtual
Platform/440BX Desktop Reference Platform, BIOS 6.00 07/30/2013
[    2.974230] Workqueue: events_unbound async_run_entry_fn
[    2.975375]  0000000000000000 000000003e328dbc ffff88007fc03bf0
ffffffff817d19a8
[    2.977032]  ffff88007fc03c38 ffff88007fc03c28 ffffffff8109671d
ffff88007bf63540
[    2.978694]  ffff88007bc52e50 ffffffff82d74a90 0000000000000082
ffffffff81a2cce8
[    2.980352] Call Trace:
[    2.980882]  <IRQ>  [<ffffffff817d19a8>] dump_stack+0x4d/0x66
[    2.982147]  [<ffffffff8109671d>] warn_slowpath_common+0x7d/0xa0
[    2.983398]  [<ffffffff8109679c>] warn_slowpath_fmt+0x5c/0x80
[    2.984606]  [<ffffffff813f5bfe>] check_unmap+0x4ee/0x9e0
[    2.985731]  [<ffffffff813f6160>] debug_dma_unmap_page+0x70/0x90
[    2.986981]  [<ffffffffa00196e3>]
pvscsi_unmap_buffers.isra.12+0x123/0x200 [vmw_pvscsi]
[    2.988635]  [<ffffffffa00198c6>]
pvscsi_process_completion_ring+0x106/0x2f0 [vmw_pvscsi]
[    2.990325]  [<ffffffffa0019b64>] pvscsi_isr+0x34/0xa0 [vmw_pvscsi]
[    2.991631]  [<ffffffff8110e26e>] handle_irq_event_percpu+0x3e/0x370
[    2.992952]  [<ffffffff8110e5dd>] handle_irq_event+0x3d/0x60
[    2.994129]  [<ffffffff811112c7>] handle_edge_irq+0x77/0x130
[    2.995320]  [<ffffffff8101dd6f>] handle_irq+0xbf/0x150
[    2.996413]  [<ffffffff8109da62>] ? irq_enter+0x42/0x90
[    2.997514]  [<ffffffff817e7b0f>] do_IRQ+0x4f/0xf0
[    2.998521]  [<ffffffff817db8f2>] common_interrupt+0x72/0x72
[    2.999703]  <EOI>  [<ffffffff817dabab>] ?
_raw_spin_unlock_irqrestore+0x3b/0x70
[    3.001587]  [<ffffffffa001a7c0>] pvscsi_queue+0x130/0x840 [vmw_pvscsi]
[    3.002970]  [<ffffffff8151cbe0>] ?
ftrace_raw_event_scsi_dispatch_cmd_error+0x220/0x220
[    3.004681]  [<ffffffff8151d2f7>] scsi_dispatch_cmd+0xb7/0x4e0
[    3.006095]  [<ffffffff8152575c>] scsi_request_fn+0x33c/0x540
[    3.007299]  [<ffffffff81396003>] __blk_run_queue+0x33/0x40
[    3.008464]  [<ffffffff813a0799>] blk_execute_rq_nowait+0xa9/0x140
[    3.009751]  [<ffffffff813a0993>] blk_execute_rq+0x133/0x1e0
[    3.010954]  [<ffffffff81272639>] ? bio_phys_segments+0x19/0x20
[    3.012195]  [<ffffffff8139bbb2>] ? blk_rq_bio_prep+0x72/0xf0
[    3.013393]  [<ffffffff81524207>] scsi_execute+0xd7/0x160
[    3.014523]  [<ffffffff81525a4c>] scsi_execute_req_flags+0x8c/0x100
[    3.015833]  [<ffffffff81527f95>] scsi_probe_and_add_lun+0x235/0xc50
[    3.036404]  [<ffffffff8150943c>] ? __pm_runtime_resume+0x5c/0x90
[    3.037679]  [<ffffffff81528e70>] __scsi_scan_target+0x110/0x6d0
[    3.038936]  [<ffffffff817daba6>] ? _raw_spin_unlock_irqrestore+0x36/0x70
[    3.040345]  [<ffffffff810f8b45>] ? trace_hardirqs_on_caller+0x105/0x1d0
[    3.041741]  [<ffffffff810f8c1d>] ? trace_hardirqs_on+0xd/0x10
[    3.042957]  [<ffffffff815295a6>] scsi_scan_channel.part.6+0x66/0x90
[    3.044288]  [<ffffffff81529769>] scsi_scan_host_selected+0xf9/0x1c0
[    3.045773]  [<ffffffff815298c1>] do_scsi_scan_host+0x91/0xa0
[    3.046977]  [<ffffffff81529a9c>] do_scan_async+0x1c/0x160
[    3.048125]  [<ffffffff810cc629>] async_run_entry_fn+0x39/0x120
[    3.049358]  [<ffffffff810bba30>] process_one_work+0x220/0x6f0
[    3.050572]  [<ffffffff810bb9c4>] ? process_one_work+0x1b4/0x6f0
[    3.051822]  [<ffffffff810bc01b>] worker_thread+0x11b/0x3a0
[    3.052987]  [<ffffffff810bbf00>] ? process_one_work+0x6f0/0x6f0
[    3.054239]  [<ffffffff810c43bf>] kthread+0xff/0x120
[    3.055285]  [<ffffffff810c42c0>] ? insert_kthread_work+0x80/0x80
[    3.056558]  [<ffffffff817e577c>] ret_from_fork+0x7c/0xb0
[    3.057686]  [<ffffffff810c42c0>] ? insert_kthread_work+0x80/0x80
[    3.058954] ---[ end trace df8f36ebf71b314f ]---
[    3.059922] Mapped at:
[    3.060431]  [<ffffffff813f4821>] debug_dma_map_page+0x91/0x140
[    3.061699]  [<ffffffffa001a911>] pvscsi_queue+0x281/0x840 [vmw_pvscsi]
[    3.063105]  [<ffffffff8151d2f7>] scsi_dispatch_cmd+0xb7/0x4e0
[    3.064351]  [<ffffffff8152575c>] scsi_request_fn+0x33c/0x540
[    3.065607]  [<ffffffff81396003>] __blk_run_queue+0x33/0x40


[1] https://bugzilla.redhat.com/show_bug.cgi?id=926917
[2] https://bugzilla.redhat.com/show_bug.cgi?id=1077118

Re: DMA-API mapping errors in vmw_pvscsi

[James Bottomley]

On Mon, 2014-03-17 at 12:52 -0400, Josh Boyer wrote:
> Hi All,
> 
> We've had a few reports[1][2] now on the vmw_pvscsi driver throwing
> DMA-API mapping errors when the DMA_API_DEBUG option is enabled.  I've
> included one from a recent 3.14-rc6 kernel below.  Looking at
> vmw_pvscsi.c, I can see pvscsi_map_buffers is missing the calls to
> pci_dma_mapping_error, which is what causes the warnings to be thrown.
> 
> However, I'm not familiar with this driver and I can't see what the
> proper error path should be in this case.  pvscsi_map_buffers is a
> void function and doesn't currently have the ability to return an
> error to the caller.  Even if it did, I'm not sure what the proper
> response to an error should be.  Thoughts?

It's missing checking in two places: from pci_map_single() and on
scsi_dma_map().  In the latter case, pvscsi_map_buffers needs to return
an error code and then pvscsi_queue_ring can return -1.

James

Re: DMA-API mapping errors in vmw_pvscsi

[Arvind Kumar]

>It's missing checking in two places: from pci_map_single() and on
>scsi_dma_map().  In the latter case, pvscsi_map_buffers needs to return
>an error code and then pvscsi_queue_ring can return -1.

Agreed! And in addition to that pvscsi_queue_ring() also calls pci_map_single()
to map the sense buffer, which should also have the check. And if 
pvscsi_map_buffers() fails then we need to unmap the sense buffer too.

We already have an internal bug filed and we are working on this.

Thanks!
Arvind

----- Original Message -----
From: "James Bottomley" Bottomley@HansenPartnership.com>
To: "Josh Boyer" <jwboyer@fedoraproject.org>
Cc: "Arvind Kumar" <arvindkumar@vmware.com>, "VMware PV-Drivers" , linux-scsi@vger.kernel.org, "Linux-Kernel@Vger. Kernel. Org" vger.kernel.org>
Sent: Monday, March 17, 2014 10:08:10 AM
Subject: Re: DMA-API mapping errors in vmw_pvscsi

On Mon, 2014-03-17 at 12:52 -0400, Josh Boyer wrote:
> Hi All,
> 
> We've had a few reports[1][2] now on the vmw_pvscsi driver throwing
> DMA-API mapping errors when the DMA_API_DEBUG option is enabled.  I've
> included one from a recent 3.14-rc6 kernel below.  Looking at
> vmw_pvscsi.c, I can see pvscsi_map_buffers is missing the calls to
> pci_dma_mapping_error, which is what causes the warnings to be thrown.
> 
> However, I'm not familiar with this driver and I can't see what the
> proper error path should be in this case.  pvscsi_map_buffers is a
> void function and doesn't currently have the ability to return an
> error to the caller.  Even if it did, I'm not sure what the proper
> response to an error should be.  Thoughts?

It's missing checking in two places: from pci_map_single() and on
scsi_dma_map().  In the latter case, pvscsi_map_buffers needs to return
an error code and then pvscsi_queue_ring can return -1.

James