On 24.07.2015 14:59, Mathias Nyman wrote:
> On 22.07.2015 17:12, Arkadiusz Miskiewicz wrote:
>>
>> On Tuesday 21 of July 2015, Mathias Nyman wrote:
>>> On 20.07.2015 23:13, Arkadiusz Miskiewicz wrote:
>>>> On Saturday 18 of July 2015, Arkadiusz Miskiewicz wrote:
>>>>> Hi.
>>>>>
>>>>> I'm on 4.2.0-rc2-00077-gf760b87 kernel and while trying to copy some
>>>>> file from usb storage (sata disk behind sata-usb bridge or pendrive;
>>>>> hapens in
>>>>
>>>>> both cases) copying process hangs just early after start with:
>>>> Looks like suspend & resume is enough. Reloading bluetooth firmware done
>>>> by kernel triggers problem:
>>>>
>>>> [ 106.302783] rtc_cmos 00:02: System wakeup disabled by ACPI
>>>> [ 106.313280] PM: resume of devices complete after 3003.032 msecs
>>>> [ 106.314079] Restarting tasks ... done.
>>>> [ 106.326434] Bluetooth: hci0: read Intel version: 370710018002030d00
>>>> [ 106.330422] Bluetooth: hci0: Intel Bluetooth firmware file:
>>>> intel/ibt-hw-37.7.10-fw-1.80.2.3.d.bseq [ 106.398223] xhci_hcd
>>>> 0000:00:14.0: ERROR Transfer event TRB DMA ptr not part of current TD
>>>> ep_index 0 comp_code 1
>>>
>
> Thanks for the logs, They show that the error is related to transfer descriptors that wrap around
> on the endpoint ring buffer by exactly one transfer block.
>
> I don't know yet why this happens, and I might need some help running additional debug
> patches to solve this. I'll take a more in depth look at the code one more time first.
>
I think I found something, The recent ring segment size increase exposed an off by one
error that has been in the driver for a long time. But you need to be unlucky and have
your memory pages allocated in a specific order to trigger it.
small fix, looks like this:
diff --git a/drivers/usb/host/xhci-ring.c b/drivers/usb/host/xhci-ring.c
index 94416ff..77da8fe 100644
--- a/drivers/usb/host/xhci-ring.c
+++ b/drivers/usb/host/xhci-ring.c
@@ -82,7 +82,7 @@ dma_addr_t xhci_trb_virt_to_dma(struct xhci_segment *seg,
return 0;
/* offset in TRBs */
segment_offset = trb - seg->trbs;
- if (segment_offset > TRBS_PER_SEGMENT)
+ if (segment_offset > TRBS_PER_SEGMENT - 1)
return 0;
return seg->dma + (segment_offset * sizeof(*trb));
}
Patch attached, could you try it out?
Thanks
-Mathias
>From 10e909ee20846793e41973941b1367e2303ec313 Mon Sep 17 00:00:00 2001
From: Mathias Nyman <mathias.nyman@xxxxxxxxxxxxxxx>
Date: Fri, 24 Jul 2015 15:56:23 +0300
Subject: [PATCH] xhci: fix off by one error in TRB DMA address boundary check
We need to check that a TRB is part of the current segment in use
before calculating its DMA address.
Previously a ring segment didn't use a full memory page, and every
new ring segment got a new memory page, so the off by one
error in checking the upper bound was never seen.
Now that we use a full memory page, 256 TRBs (4096 bytes) the off by one
caused issues as it doesnt catch the case when a TRB was the first element
of the next segment.
This is triggered if the virtual memory pages for a ring segment are
next to each in increasing order where the ring buffer wraps around and causes
errors like:
[ 106.398223] xhci_hcd 0000:00:14.0: ERROR Transfer event TRB DMA ptr not
part of current TD ep_index 0 comp_code 1
[ 106.398230] xhci_hcd 0000:00:14.0: Looking for event-dma fffd3000
trb-start fffd4fd0 trb-end fffd5000 seg-start fffd4000 seg-end fffd4ff0
the trb-end address is one outside the end-seg address.
Signed-off-by: Mathias Nyman <mathias.nyman@xxxxxxxxxxxxxxx>
---
drivers/usb/host/xhci-ring.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/usb/host/xhci-ring.c b/drivers/usb/host/xhci-ring.c
index 94416ff..77da8fe 100644
--- a/drivers/usb/host/xhci-ring.c
+++ b/drivers/usb/host/xhci-ring.c
@@ -82,7 +82,7 @@ dma_addr_t xhci_trb_virt_to_dma(struct xhci_segment *seg,
return 0;
/* offset in TRBs */
segment_offset = trb - seg->trbs;
- if (segment_offset > TRBS_PER_SEGMENT)
+ if (segment_offset > TRBS_PER_SEGMENT - 1)
return 0;
return seg->dma + (segment_offset * sizeof(*trb));
}
--
1.8.3.2
