[PATCH] usb: storage: add quirks for VIA VL817 USB3-SATA bridge

[PATCH] usb: storage: add quirks for VIA VL817 USB3-SATA bridge

[Tobias Jakobi]

The VL817 is used in the RaidSonic Icy Box IB-3740-C31 enclosure. The enclosure
is advertised as having UASP support, but appears to have problems with 4Kn
drives (test was done with two Seagate Exos X, 12TB).

Disable UAS for the VL817 as it behaves highly unstable:

[Aug14 16:31] usb 2-1.2: USB disconnect, device number 4
[  +0.007701] sd 4:0:0:0: [sdb] tag#4 uas_zap_pending 0 uas-tag 1 inflight: CMD
[  +0.000004] sd 4:0:0:0: [sdb] tag#4 CDB: opcode=0x2a 2a 00 00 37 63 da 00 00 80 00
[  +0.000022] sd 4:0:0:0: [sdb] tag#4 UNKNOWN(0x2003) Result: hostbyte=0x01 driverbyte=0x00 cmd_age=19s
[  +0.000001] sd 4:0:0:0: [sdb] tag#4 CDB: opcode=0x2a 2a 00 00 37 63 da 00 00 80 00
[  +0.000001] blk_update_request: I/O error, dev sdb, sector 29040336 op 0x1:(WRITE) flags 0x0 phys_seg 128 prio class 0
[  +0.000028] blk_update_request: I/O error, dev sdb, sector 29041360 op 0x1:(WRITE) flags 0x0 phys_seg 128 prio class 0
[  +0.000000] blk_update_request: I/O error, dev sdb, sector 16 op 0x1:(WRITE) flags 0x800 phys_seg 1 prio class 0
[  +0.000005] md: super_written gets error=-5
[  +0.000002] md/raid1:md126: Disk failure on sdb, disabling device.
              md/raid1:md126: Operation continuing on 1 devices.
[  +0.000024] blk_update_request: I/O error, dev sdb, sector 29042384 op 0x1:(WRITE) flags 0x0 phys_seg 128 prio class 0
[  +0.000222] sd 4:0:0:0: [sdb] Synchronizing SCSI cache
[  +0.078154] blk_update_request: I/O error, dev sdb, sector 29040336 op 0x1:(WRITE) flags 0x800 phys_seg 1 prio class 0
[  +0.000025] blk_update_request: I/O error, dev sdb, sector 29040344 op 0x1:(WRITE) flags 0x800 phys_seg 1 prio class 0
[  +0.007520] blk_update_request: I/O error, dev sdb, sector 29040352 op 0x1:(WRITE) flags 0x800 phys_seg 1 prio class 0
[  +0.000021] blk_update_request: I/O error, dev sdb, sector 29040360 op 0x1:(WRITE) flags 0x800 phys_seg 1 prio class 0
[  +0.000015] blk_update_request: I/O error, dev sdb, sector 29040368 op 0x1:(WRITE) flags 0x800 phys_seg 1 prio class 0
[  +0.000009] blk_update_request: I/O error, dev sdb, sector 29040376 op 0x1:(WRITE) flags 0x800 phys_seg 1 prio class 0
[  +0.023299] sd 4:0:0:0: [sdb] Synchronize Cache(10) failed: Result: hostbyte=0x07 driverbyte=0x00
[  +1.893439] usb 2-1.2: new SuperSpeed Plus Gen 2x1 USB device number 7 using xhci_hcd
[  +0.024064] scsi host7: uas
[ +16.365880] scsi 7:0:0:0: Direct-Access     ST12000N M001G-2MV103     SB2D PQ: 0 ANSI: 6
[  +0.001192] sd 7:0:0:0: Attached scsi generic sg1 type 0
[  +0.000940] sd 7:0:0:0: [sde] 2929721344 4096-byte logical blocks: (12.0 TB/10.9 TiB)
[  +0.000130] sd 7:0:0:0: [sde] Write Protect is off
[  +0.000001] sd 7:0:0:0: [sde] Mode Sense: 2f 00 00 00
[  +0.000265] sd 7:0:0:0: [sde] Write cache: enabled, read cache: enabled, doesn't support DPO or FUA
[  +0.000399] sd 7:0:0:0: [sde] Optimal transfer size 268431360 bytes
[  +0.120240] sd 7:0:0:0: [sde] Attached SCSI disk

Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
---
 drivers/usb/storage/unusual_uas.h | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/drivers/usb/storage/unusual_uas.h b/drivers/usb/storage/unusual_uas.h
index bda0f2cdf093..7d83ecf835c6 100644
--- a/drivers/usb/storage/unusual_uas.h
+++ b/drivers/usb/storage/unusual_uas.h
@@ -125,6 +125,13 @@ UNUSUAL_DEV(0x2109, 0x0711, 0x0000, 0x9999,
 		USB_SC_DEVICE, USB_PR_DEVICE, NULL,
 		US_FL_NO_ATA_1X),
 
+/* Reported-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de> */
+UNUSUAL_DEV(0x2109, 0x0715, 0x0000, 0x9999,
+		"VIA",
+		"VL817",
+		USB_SC_DEVICE, USB_PR_DEVICE, NULL,
+		US_FL_IGNORE_UAS),
+
 /* Reported-by: Icenowy Zheng <icenowy@aosc.io> */
 UNUSUAL_DEV(0x2537, 0x1068, 0x0000, 0x9999,
 		"Norelsys",
-- 
2.32.0

Re: [PATCH] usb: storage: add quirks for VIA VL817 USB3-SATA bridge

[Alan Stern]

On Tue, Sep 21, 2021 at 12:17:52PM +0200, Tobias Jakobi wrote:
> The VL817 is used in the RaidSonic Icy Box IB-3740-C31 enclosure. The enclosure
> is advertised as having UASP support, but appears to have problems with 4Kn
> drives (test was done with two Seagate Exos X, 12TB).
> 
> Disable UAS for the VL817 as it behaves highly unstable:
> 
> [Aug14 16:31] usb 2-1.2: USB disconnect, device number 4

So first the drive was unplugged or disconnected...

> [  +0.007701] sd 4:0:0:0: [sdb] tag#4 uas_zap_pending 0 uas-tag 1 inflight: CMD
> [  +0.000004] sd 4:0:0:0: [sdb] tag#4 CDB: opcode=0x2a 2a 00 00 37 63 da 00 00 80 00
> [  +0.000022] sd 4:0:0:0: [sdb] tag#4 UNKNOWN(0x2003) Result: hostbyte=0x01 driverbyte=0x00 cmd_age=19s
> [  +0.000001] sd 4:0:0:0: [sdb] tag#4 CDB: opcode=0x2a 2a 00 00 37 63 da 00 00 80 00
> [  +0.000001] blk_update_request: I/O error, dev sdb, sector 29040336 op 0x1:(WRITE) flags 0x0 phys_seg 128 prio class 0
> [  +0.000028] blk_update_request: I/O error, dev sdb, sector 29041360 op 0x1:(WRITE) flags 0x0 phys_seg 128 prio class 0
> [  +0.000000] blk_update_request: I/O error, dev sdb, sector 16 op 0x1:(WRITE) flags 0x800 phys_seg 1 prio class 0
> [  +0.000005] md: super_written gets error=-5
> [  +0.000002] md/raid1:md126: Disk failure on sdb, disabling device.
>               md/raid1:md126: Operation continuing on 1 devices.
> [  +0.000024] blk_update_request: I/O error, dev sdb, sector 29042384 op 0x1:(WRITE) flags 0x0 phys_seg 128 prio class 0
> [  +0.000222] sd 4:0:0:0: [sdb] Synchronizing SCSI cache
> [  +0.078154] blk_update_request: I/O error, dev sdb, sector 29040336 op 0x1:(WRITE) flags 0x800 phys_seg 1 prio class 0
> [  +0.000025] blk_update_request: I/O error, dev sdb, sector 29040344 op 0x1:(WRITE) flags 0x800 phys_seg 1 prio class 0
> [  +0.007520] blk_update_request: I/O error, dev sdb, sector 29040352 op 0x1:(WRITE) flags 0x800 phys_seg 1 prio class 0
> [  +0.000021] blk_update_request: I/O error, dev sdb, sector 29040360 op 0x1:(WRITE) flags 0x800 phys_seg 1 prio class 0
> [  +0.000015] blk_update_request: I/O error, dev sdb, sector 29040368 op 0x1:(WRITE) flags 0x800 phys_seg 1 prio class 0
> [  +0.000009] blk_update_request: I/O error, dev sdb, sector 29040376 op 0x1:(WRITE) flags 0x800 phys_seg 1 prio class 0
> [  +0.023299] sd 4:0:0:0: [sdb] Synchronize Cache(10) failed: Result: hostbyte=0x07 driverbyte=0x00

Then there was a bunch of errors, which is to be expected when a drive 
is suddenly disconnected...

> [  +1.893439] usb 2-1.2: new SuperSpeed Plus Gen 2x1 USB device number 7 using xhci_hcd
> [  +0.024064] scsi host7: uas
> [ +16.365880] scsi 7:0:0:0: Direct-Access     ST12000N M001G-2MV103     SB2D PQ: 0 ANSI: 6
> [  +0.001192] sd 7:0:0:0: Attached scsi generic sg1 type 0
> [  +0.000940] sd 7:0:0:0: [sde] 2929721344 4096-byte logical blocks: (12.0 TB/10.9 TiB)
> [  +0.000130] sd 7:0:0:0: [sde] Write Protect is off
> [  +0.000001] sd 7:0:0:0: [sde] Mode Sense: 2f 00 00 00
> [  +0.000265] sd 7:0:0:0: [sde] Write cache: enabled, read cache: enabled, doesn't support DPO or FUA
> [  +0.000399] sd 7:0:0:0: [sde] Optimal transfer size 268431360 bytes
> [  +0.120240] sd 7:0:0:0: [sde] Attached SCSI disk

And then the drive reconnected, this time successfully.  How does this 
show that UAS was the reason for the problem?  Indeed, how does this 
show there was any problem at all?

> Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
> ---
>  drivers/usb/storage/unusual_uas.h | 7 +++++++
>  1 file changed, 7 insertions(+)
> 
> diff --git a/drivers/usb/storage/unusual_uas.h b/drivers/usb/storage/unusual_uas.h
> index bda0f2cdf093..7d83ecf835c6 100644
> --- a/drivers/usb/storage/unusual_uas.h
> +++ b/drivers/usb/storage/unusual_uas.h
> @@ -125,6 +125,13 @@ UNUSUAL_DEV(0x2109, 0x0711, 0x0000, 0x9999,
>  		USB_SC_DEVICE, USB_PR_DEVICE, NULL,
>  		US_FL_NO_ATA_1X),
>  
> +/* Reported-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de> */
> +UNUSUAL_DEV(0x2109, 0x0715, 0x0000, 0x9999,
> +		"VIA",
> +		"VL817",
> +		USB_SC_DEVICE, USB_PR_DEVICE, NULL,
> +		US_FL_IGNORE_UAS),
> +
>  /* Reported-by: Icenowy Zheng <icenowy@aosc.io> */
>  UNUSUAL_DEV(0x2537, 0x1068, 0x0000, 0x9999,
>  		"Norelsys",

Instead of IGNORE_UAS, have you tried the NO_ATA_1X flag, which seems to 
help in the preceding entry (a different device from the same vendor)?

Alan Stern

Re: [PATCH] usb: storage: add quirks for VIA VL817 USB3-SATA bridge

[Tobias Jakobi]

Hi Alan,

sorry but your analysis of the log is wrong. Nothing was disconnected or 
unplugged when the device behaves this way. The enclosure is connected 
to the power the entire time, and the same applies to the physical USB 
connection to my system.

To make things very clear: This happens in under five minutes after 
having powered up the enclosure and starting a file transfer to the 
installed RAID. After blacklisting UASP the enclosure works perfectly 
fine for hours. I hope this clears things up.

With best wishes,
Tobias

On 9/21/21 5:13 PM, Alan Stern wrote:
> On Tue, Sep 21, 2021 at 12:17:52PM +0200, Tobias Jakobi wrote:
>> The VL817 is used in the RaidSonic Icy Box IB-3740-C31 enclosure. The enclosure
>> is advertised as having UASP support, but appears to have problems with 4Kn
>> drives (test was done with two Seagate Exos X, 12TB).
>>
>> Disable UAS for the VL817 as it behaves highly unstable:
>>
>> [Aug14 16:31] usb 2-1.2: USB disconnect, device number 4
> So first the drive was unplugged or disconnected...
>
>> [  +0.007701] sd 4:0:0:0: [sdb] tag#4 uas_zap_pending 0 uas-tag 1 inflight: CMD
>> [  +0.000004] sd 4:0:0:0: [sdb] tag#4 CDB: opcode=0x2a 2a 00 00 37 63 da 00 00 80 00
>> [  +0.000022] sd 4:0:0:0: [sdb] tag#4 UNKNOWN(0x2003) Result: hostbyte=0x01 driverbyte=0x00 cmd_age=19s
>> [  +0.000001] sd 4:0:0:0: [sdb] tag#4 CDB: opcode=0x2a 2a 00 00 37 63 da 00 00 80 00
>> [  +0.000001] blk_update_request: I/O error, dev sdb, sector 29040336 op 0x1:(WRITE) flags 0x0 phys_seg 128 prio class 0
>> [  +0.000028] blk_update_request: I/O error, dev sdb, sector 29041360 op 0x1:(WRITE) flags 0x0 phys_seg 128 prio class 0
>> [  +0.000000] blk_update_request: I/O error, dev sdb, sector 16 op 0x1:(WRITE) flags 0x800 phys_seg 1 prio class 0
>> [  +0.000005] md: super_written gets error=-5
>> [  +0.000002] md/raid1:md126: Disk failure on sdb, disabling device.
>>                md/raid1:md126: Operation continuing on 1 devices.
>> [  +0.000024] blk_update_request: I/O error, dev sdb, sector 29042384 op 0x1:(WRITE) flags 0x0 phys_seg 128 prio class 0
>> [  +0.000222] sd 4:0:0:0: [sdb] Synchronizing SCSI cache
>> [  +0.078154] blk_update_request: I/O error, dev sdb, sector 29040336 op 0x1:(WRITE) flags 0x800 phys_seg 1 prio class 0
>> [  +0.000025] blk_update_request: I/O error, dev sdb, sector 29040344 op 0x1:(WRITE) flags 0x800 phys_seg 1 prio class 0
>> [  +0.007520] blk_update_request: I/O error, dev sdb, sector 29040352 op 0x1:(WRITE) flags 0x800 phys_seg 1 prio class 0
>> [  +0.000021] blk_update_request: I/O error, dev sdb, sector 29040360 op 0x1:(WRITE) flags 0x800 phys_seg 1 prio class 0
>> [  +0.000015] blk_update_request: I/O error, dev sdb, sector 29040368 op 0x1:(WRITE) flags 0x800 phys_seg 1 prio class 0
>> [  +0.000009] blk_update_request: I/O error, dev sdb, sector 29040376 op 0x1:(WRITE) flags 0x800 phys_seg 1 prio class 0
>> [  +0.023299] sd 4:0:0:0: [sdb] Synchronize Cache(10) failed: Result: hostbyte=0x07 driverbyte=0x00
> Then there was a bunch of errors, which is to be expected when a drive
> is suddenly disconnected...
>
>> [  +1.893439] usb 2-1.2: new SuperSpeed Plus Gen 2x1 USB device number 7 using xhci_hcd
>> [  +0.024064] scsi host7: uas
>> [ +16.365880] scsi 7:0:0:0: Direct-Access     ST12000N M001G-2MV103     SB2D PQ: 0 ANSI: 6
>> [  +0.001192] sd 7:0:0:0: Attached scsi generic sg1 type 0
>> [  +0.000940] sd 7:0:0:0: [sde] 2929721344 4096-byte logical blocks: (12.0 TB/10.9 TiB)
>> [  +0.000130] sd 7:0:0:0: [sde] Write Protect is off
>> [  +0.000001] sd 7:0:0:0: [sde] Mode Sense: 2f 00 00 00
>> [  +0.000265] sd 7:0:0:0: [sde] Write cache: enabled, read cache: enabled, doesn't support DPO or FUA
>> [  +0.000399] sd 7:0:0:0: [sde] Optimal transfer size 268431360 bytes
>> [  +0.120240] sd 7:0:0:0: [sde] Attached SCSI disk
> And then the drive reconnected, this time successfully.  How does this
> show that UAS was the reason for the problem?  Indeed, how does this
> show there was any problem at all?
>
>> Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
>> ---
>>   drivers/usb/storage/unusual_uas.h | 7 +++++++
>>   1 file changed, 7 insertions(+)
>>
>> diff --git a/drivers/usb/storage/unusual_uas.h b/drivers/usb/storage/unusual_uas.h
>> index bda0f2cdf093..7d83ecf835c6 100644
>> --- a/drivers/usb/storage/unusual_uas.h
>> +++ b/drivers/usb/storage/unusual_uas.h
>> @@ -125,6 +125,13 @@ UNUSUAL_DEV(0x2109, 0x0711, 0x0000, 0x9999,
>>   		USB_SC_DEVICE, USB_PR_DEVICE, NULL,
>>   		US_FL_NO_ATA_1X),
>>   
>> +/* Reported-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de> */
>> +UNUSUAL_DEV(0x2109, 0x0715, 0x0000, 0x9999,
>> +		"VIA",
>> +		"VL817",
>> +		USB_SC_DEVICE, USB_PR_DEVICE, NULL,
>> +		US_FL_IGNORE_UAS),
>> +
>>   /* Reported-by: Icenowy Zheng <icenowy@aosc.io> */
>>   UNUSUAL_DEV(0x2537, 0x1068, 0x0000, 0x9999,
>>   		"Norelsys",
> Instead of IGNORE_UAS, have you tried the NO_ATA_1X flag, which seems to
> help in the preceding entry (a different device from the same vendor)?
>
> Alan Stern

Re: [PATCH] usb: storage: add quirks for VIA VL817 USB3-SATA bridge

[Alan Stern]

On Tue, Sep 21, 2021 at 06:06:45PM +0200, Tobias Jakobi wrote:
> Hi Alan,
> 
> sorry but your analysis of the log is wrong. Nothing was disconnected or
> unplugged when the device behaves this way. The enclosure is connected to
> the power the entire time, and the same applies to the physical USB
> connection to my system.

That may be so, but if it is then the log extract you included with the 
patch is very misleading.  For instance, you didn't include any part of 
the log before and leading up to the line saying "usb 2-1.2: USB 
disconnect, device number 4".  Thus there is no way for the reader to 
tell what caused this event, whether it was a physical unplug or not.

> To make things very clear: This happens in under five minutes after having
> powered up the enclosure and starting a file transfer to the installed RAID.
> After blacklisting UASP the enclosure works perfectly fine for hours. I hope
> this clears things up.

You didn't answer my question about using NO_ATA_1X instead of 
IGNORE_UAS.  This is a perfect example of one of the dangers of 
top-posting -- it makes it far too easy for people to miss important 
points in the email they are replying to.  (Hint: Don't top-post!)

Alan Stern

> With best wishes,
> Tobias
> 
> On 9/21/21 5:13 PM, Alan Stern wrote:
> > On Tue, Sep 21, 2021 at 12:17:52PM +0200, Tobias Jakobi wrote:
> > > The VL817 is used in the RaidSonic Icy Box IB-3740-C31 enclosure. The enclosure
> > > is advertised as having UASP support, but appears to have problems with 4Kn
> > > drives (test was done with two Seagate Exos X, 12TB).
> > > 
> > > Disable UAS for the VL817 as it behaves highly unstable:
> > > 
> > > [Aug14 16:31] usb 2-1.2: USB disconnect, device number 4
> > So first the drive was unplugged or disconnected...
> > 
> > > [  +0.007701] sd 4:0:0:0: [sdb] tag#4 uas_zap_pending 0 uas-tag 1 inflight: CMD
> > > [  +0.000004] sd 4:0:0:0: [sdb] tag#4 CDB: opcode=0x2a 2a 00 00 37 63 da 00 00 80 00
> > > [  +0.000022] sd 4:0:0:0: [sdb] tag#4 UNKNOWN(0x2003) Result: hostbyte=0x01 driverbyte=0x00 cmd_age=19s
> > > [  +0.000001] sd 4:0:0:0: [sdb] tag#4 CDB: opcode=0x2a 2a 00 00 37 63 da 00 00 80 00
> > > [  +0.000001] blk_update_request: I/O error, dev sdb, sector 29040336 op 0x1:(WRITE) flags 0x0 phys_seg 128 prio class 0
> > > [  +0.000028] blk_update_request: I/O error, dev sdb, sector 29041360 op 0x1:(WRITE) flags 0x0 phys_seg 128 prio class 0
> > > [  +0.000000] blk_update_request: I/O error, dev sdb, sector 16 op 0x1:(WRITE) flags 0x800 phys_seg 1 prio class 0
> > > [  +0.000005] md: super_written gets error=-5
> > > [  +0.000002] md/raid1:md126: Disk failure on sdb, disabling device.
> > >                md/raid1:md126: Operation continuing on 1 devices.
> > > [  +0.000024] blk_update_request: I/O error, dev sdb, sector 29042384 op 0x1:(WRITE) flags 0x0 phys_seg 128 prio class 0
> > > [  +0.000222] sd 4:0:0:0: [sdb] Synchronizing SCSI cache
> > > [  +0.078154] blk_update_request: I/O error, dev sdb, sector 29040336 op 0x1:(WRITE) flags 0x800 phys_seg 1 prio class 0
> > > [  +0.000025] blk_update_request: I/O error, dev sdb, sector 29040344 op 0x1:(WRITE) flags 0x800 phys_seg 1 prio class 0
> > > [  +0.007520] blk_update_request: I/O error, dev sdb, sector 29040352 op 0x1:(WRITE) flags 0x800 phys_seg 1 prio class 0
> > > [  +0.000021] blk_update_request: I/O error, dev sdb, sector 29040360 op 0x1:(WRITE) flags 0x800 phys_seg 1 prio class 0
> > > [  +0.000015] blk_update_request: I/O error, dev sdb, sector 29040368 op 0x1:(WRITE) flags 0x800 phys_seg 1 prio class 0
> > > [  +0.000009] blk_update_request: I/O error, dev sdb, sector 29040376 op 0x1:(WRITE) flags 0x800 phys_seg 1 prio class 0
> > > [  +0.023299] sd 4:0:0:0: [sdb] Synchronize Cache(10) failed: Result: hostbyte=0x07 driverbyte=0x00
> > Then there was a bunch of errors, which is to be expected when a drive
> > is suddenly disconnected...
> > 
> > > [  +1.893439] usb 2-1.2: new SuperSpeed Plus Gen 2x1 USB device number 7 using xhci_hcd
> > > [  +0.024064] scsi host7: uas
> > > [ +16.365880] scsi 7:0:0:0: Direct-Access     ST12000N M001G-2MV103     SB2D PQ: 0 ANSI: 6
> > > [  +0.001192] sd 7:0:0:0: Attached scsi generic sg1 type 0
> > > [  +0.000940] sd 7:0:0:0: [sde] 2929721344 4096-byte logical blocks: (12.0 TB/10.9 TiB)
> > > [  +0.000130] sd 7:0:0:0: [sde] Write Protect is off
> > > [  +0.000001] sd 7:0:0:0: [sde] Mode Sense: 2f 00 00 00
> > > [  +0.000265] sd 7:0:0:0: [sde] Write cache: enabled, read cache: enabled, doesn't support DPO or FUA
> > > [  +0.000399] sd 7:0:0:0: [sde] Optimal transfer size 268431360 bytes
> > > [  +0.120240] sd 7:0:0:0: [sde] Attached SCSI disk
> > And then the drive reconnected, this time successfully.  How does this
> > show that UAS was the reason for the problem?  Indeed, how does this
> > show there was any problem at all?
> > 
> > > Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
> > > ---
> > >   drivers/usb/storage/unusual_uas.h | 7 +++++++
> > >   1 file changed, 7 insertions(+)
> > > 
> > > diff --git a/drivers/usb/storage/unusual_uas.h b/drivers/usb/storage/unusual_uas.h
> > > index bda0f2cdf093..7d83ecf835c6 100644
> > > --- a/drivers/usb/storage/unusual_uas.h
> > > +++ b/drivers/usb/storage/unusual_uas.h
> > > @@ -125,6 +125,13 @@ UNUSUAL_DEV(0x2109, 0x0711, 0x0000, 0x9999,
> > >   		USB_SC_DEVICE, USB_PR_DEVICE, NULL,
> > >   		US_FL_NO_ATA_1X),
> > > +/* Reported-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de> */
> > > +UNUSUAL_DEV(0x2109, 0x0715, 0x0000, 0x9999,
> > > +		"VIA",
> > > +		"VL817",
> > > +		USB_SC_DEVICE, USB_PR_DEVICE, NULL,
> > > +		US_FL_IGNORE_UAS),
> > > +
> > >   /* Reported-by: Icenowy Zheng <icenowy@aosc.io> */
> > >   UNUSUAL_DEV(0x2537, 0x1068, 0x0000, 0x9999,
> > >   		"Norelsys",
> > Instead of IGNORE_UAS, have you tried the NO_ATA_1X flag, which seems to
> > help in the preceding entry (a different device from the same vendor)?
> > 
> > Alan Stern
> 

Re: [PATCH] usb: storage: add quirks for VIA VL817 USB3-SATA bridge

[Tobias Jakobi]

Hello Alan,

On 9/21/21 6:42 PM, Alan Stern wrote:
> On Tue, Sep 21, 2021 at 06:06:45PM +0200, Tobias Jakobi wrote:
>> Hi Alan,
>>
>> sorry but your analysis of the log is wrong. Nothing was disconnected or
>> unplugged when the device behaves this way. The enclosure is connected to
>> the power the entire time, and the same applies to the physical USB
>> connection to my system.
> That may be so, but if it is then the log extract you included with the
> patch is very misleading.  For instance, you didn't include any part of
> the log before and leading up to the line saying "usb 2-1.2: USB
> disconnect, device number 4".  Thus there is no way for the reader to
> tell what caused this event, whether it was a physical unplug or not.

I included the part of the kernel log which shows how the issue 
manifests itself. Do you think I'm so stupid as to believe I could 
prevent a physical unplug of the enclosure by blacklisting UAS? Really, 
this is getting ridiculous...

>> To make things very clear: This happens in under five minutes after having
>> powered up the enclosure and starting a file transfer to the installed RAID.
>> After blacklisting UASP the enclosure works perfectly fine for hours. I hope
>> this clears things up.
> You didn't answer my question about using NO_ATA_1X instead of
> IGNORE_UAS.  This is a perfect example of one of the dangers of
> top-posting -- it makes it far too easy for people to miss important
> points in the email they are replying to.  (Hint: Don't top-post!)

I did not answer this question, because I didn't have the answer to it 
yet. I have tested your suggestion today, but sadly I'm running into the 
same type of problem with NO_ATA_1X. You can find the complete kernel 
log here:
https://www.math.uni-bielefeld.de/~tjakobi/archive/dmesg_VL817.log

The RAID1 is broken after such an event.

With best wishes,
Tobias


>
> Alan Stern
>
>> With best wishes,
>> Tobias
>>
>> On 9/21/21 5:13 PM, Alan Stern wrote:
>>> On Tue, Sep 21, 2021 at 12:17:52PM +0200, Tobias Jakobi wrote:
>>>> The VL817 is used in the RaidSonic Icy Box IB-3740-C31 enclosure. The enclosure
>>>> is advertised as having UASP support, but appears to have problems with 4Kn
>>>> drives (test was done with two Seagate Exos X, 12TB).
>>>>
>>>> Disable UAS for the VL817 as it behaves highly unstable:
>>>>
>>>> [Aug14 16:31] usb 2-1.2: USB disconnect, device number 4
>>> So first the drive was unplugged or disconnected...
>>>
>>>> [  +0.007701] sd 4:0:0:0: [sdb] tag#4 uas_zap_pending 0 uas-tag 1 inflight: CMD
>>>> [  +0.000004] sd 4:0:0:0: [sdb] tag#4 CDB: opcode=0x2a 2a 00 00 37 63 da 00 00 80 00
>>>> [  +0.000022] sd 4:0:0:0: [sdb] tag#4 UNKNOWN(0x2003) Result: hostbyte=0x01 driverbyte=0x00 cmd_age=19s
>>>> [  +0.000001] sd 4:0:0:0: [sdb] tag#4 CDB: opcode=0x2a 2a 00 00 37 63 da 00 00 80 00
>>>> [  +0.000001] blk_update_request: I/O error, dev sdb, sector 29040336 op 0x1:(WRITE) flags 0x0 phys_seg 128 prio class 0
>>>> [  +0.000028] blk_update_request: I/O error, dev sdb, sector 29041360 op 0x1:(WRITE) flags 0x0 phys_seg 128 prio class 0
>>>> [  +0.000000] blk_update_request: I/O error, dev sdb, sector 16 op 0x1:(WRITE) flags 0x800 phys_seg 1 prio class 0
>>>> [  +0.000005] md: super_written gets error=-5
>>>> [  +0.000002] md/raid1:md126: Disk failure on sdb, disabling device.
>>>>                 md/raid1:md126: Operation continuing on 1 devices.
>>>> [  +0.000024] blk_update_request: I/O error, dev sdb, sector 29042384 op 0x1:(WRITE) flags 0x0 phys_seg 128 prio class 0
>>>> [  +0.000222] sd 4:0:0:0: [sdb] Synchronizing SCSI cache
>>>> [  +0.078154] blk_update_request: I/O error, dev sdb, sector 29040336 op 0x1:(WRITE) flags 0x800 phys_seg 1 prio class 0
>>>> [  +0.000025] blk_update_request: I/O error, dev sdb, sector 29040344 op 0x1:(WRITE) flags 0x800 phys_seg 1 prio class 0
>>>> [  +0.007520] blk_update_request: I/O error, dev sdb, sector 29040352 op 0x1:(WRITE) flags 0x800 phys_seg 1 prio class 0
>>>> [  +0.000021] blk_update_request: I/O error, dev sdb, sector 29040360 op 0x1:(WRITE) flags 0x800 phys_seg 1 prio class 0
>>>> [  +0.000015] blk_update_request: I/O error, dev sdb, sector 29040368 op 0x1:(WRITE) flags 0x800 phys_seg 1 prio class 0
>>>> [  +0.000009] blk_update_request: I/O error, dev sdb, sector 29040376 op 0x1:(WRITE) flags 0x800 phys_seg 1 prio class 0
>>>> [  +0.023299] sd 4:0:0:0: [sdb] Synchronize Cache(10) failed: Result: hostbyte=0x07 driverbyte=0x00
>>> Then there was a bunch of errors, which is to be expected when a drive
>>> is suddenly disconnected...
>>>
>>>> [  +1.893439] usb 2-1.2: new SuperSpeed Plus Gen 2x1 USB device number 7 using xhci_hcd
>>>> [  +0.024064] scsi host7: uas
>>>> [ +16.365880] scsi 7:0:0:0: Direct-Access     ST12000N M001G-2MV103     SB2D PQ: 0 ANSI: 6
>>>> [  +0.001192] sd 7:0:0:0: Attached scsi generic sg1 type 0
>>>> [  +0.000940] sd 7:0:0:0: [sde] 2929721344 4096-byte logical blocks: (12.0 TB/10.9 TiB)
>>>> [  +0.000130] sd 7:0:0:0: [sde] Write Protect is off
>>>> [  +0.000001] sd 7:0:0:0: [sde] Mode Sense: 2f 00 00 00
>>>> [  +0.000265] sd 7:0:0:0: [sde] Write cache: enabled, read cache: enabled, doesn't support DPO or FUA
>>>> [  +0.000399] sd 7:0:0:0: [sde] Optimal transfer size 268431360 bytes
>>>> [  +0.120240] sd 7:0:0:0: [sde] Attached SCSI disk
>>> And then the drive reconnected, this time successfully.  How does this
>>> show that UAS was the reason for the problem?  Indeed, how does this
>>> show there was any problem at all?
>>>
>>>> Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
>>>> ---
>>>>    drivers/usb/storage/unusual_uas.h | 7 +++++++
>>>>    1 file changed, 7 insertions(+)
>>>>
>>>> diff --git a/drivers/usb/storage/unusual_uas.h b/drivers/usb/storage/unusual_uas.h
>>>> index bda0f2cdf093..7d83ecf835c6 100644
>>>> --- a/drivers/usb/storage/unusual_uas.h
>>>> +++ b/drivers/usb/storage/unusual_uas.h
>>>> @@ -125,6 +125,13 @@ UNUSUAL_DEV(0x2109, 0x0711, 0x0000, 0x9999,
>>>>    		USB_SC_DEVICE, USB_PR_DEVICE, NULL,
>>>>    		US_FL_NO_ATA_1X),
>>>> +/* Reported-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de> */
>>>> +UNUSUAL_DEV(0x2109, 0x0715, 0x0000, 0x9999,
>>>> +		"VIA",
>>>> +		"VL817",
>>>> +		USB_SC_DEVICE, USB_PR_DEVICE, NULL,
>>>> +		US_FL_IGNORE_UAS),
>>>> +
>>>>    /* Reported-by: Icenowy Zheng <icenowy@aosc.io> */
>>>>    UNUSUAL_DEV(0x2537, 0x1068, 0x0000, 0x9999,
>>>>    		"Norelsys",
>>> Instead of IGNORE_UAS, have you tried the NO_ATA_1X flag, which seems to
>>> help in the preceding entry (a different device from the same vendor)?
>>>
>>> Alan Stern

Re: [PATCH] usb: storage: add quirks for VIA VL817 USB3-SATA bridge

[Alan Stern]

On Sun, Sep 26, 2021 at 08:14:46PM +0200, Tobias Jakobi wrote:
> Hello Alan,
> 
> On 9/21/21 6:42 PM, Alan Stern wrote:
> > On Tue, Sep 21, 2021 at 06:06:45PM +0200, Tobias Jakobi wrote:
> > > Hi Alan,
> > > 
> > > sorry but your analysis of the log is wrong. Nothing was disconnected or
> > > unplugged when the device behaves this way. The enclosure is connected to
> > > the power the entire time, and the same applies to the physical USB
> > > connection to my system.
> > That may be so, but if it is then the log extract you included with the
> > patch is very misleading.  For instance, you didn't include any part of
> > the log before and leading up to the line saying "usb 2-1.2: USB
> > disconnect, device number 4".  Thus there is no way for the reader to
> > tell what caused this event, whether it was a physical unplug or not.
> 
> I included the part of the kernel log which shows how the issue manifests
> itself. Do you think I'm so stupid as to believe I could prevent a physical
> unplug of the enclosure by blacklisting UAS? Really, this is getting
> ridiculous...

No, I don't believe that.  However, I do believe you are so stupid as 
to post a patch with a description that does not justify it and 
includes a lot of irrelevant details.  (Hey, you brought up the issue 
of how stupid you are!  :-)

> > > To make things very clear: This happens in under five minutes after having
> > > powered up the enclosure and starting a file transfer to the installed RAID.
> > > After blacklisting UASP the enclosure works perfectly fine for hours. I hope
> > > this clears things up.
> > You didn't answer my question about using NO_ATA_1X instead of
> > IGNORE_UAS.  This is a perfect example of one of the dangers of
> > top-posting -- it makes it far too easy for people to miss important
> > points in the email they are replying to.  (Hint: Don't top-post!)
> 
> I did not answer this question, because I didn't have the answer to it yet.
> I have tested your suggestion today, but sadly I'm running into the same
> type of problem with NO_ATA_1X. You can find the complete kernel log here:
> https://www.math.uni-bielefeld.de/~tjakobi/archive/dmesg_VL817.log

Good, thank you.  Unfortunately the log doesn't include any smoking 
guns pointing to an underlying cause.

> The RAID1 is broken after such an event.
> 
> With best wishes,
> Tobias
> 
> 
> > 
> > Alan Stern
> > 
> > > With best wishes,
> > > Tobias
> > > 
> > > On 9/21/21 5:13 PM, Alan Stern wrote:
> > > > On Tue, Sep 21, 2021 at 12:17:52PM +0200, Tobias Jakobi wrote:
> > > > > The VL817 is used in the RaidSonic Icy Box IB-3740-C31 enclosure. The enclosure
> > > > > is advertised as having UASP support, but appears to have problems with 4Kn
> > > > > drives (test was done with two Seagate Exos X, 12TB).
> > > > > 
> > > > > Disable UAS for the VL817 as it behaves highly unstable:

I still have to wonder if the enclosure works okay with other types of 
disk drive.  And if it doesn't, why don't these errors show up on 
Windows systems?  Or on other VIA enclosures?

That's why I'm cautious about accepting this patch.  I don't want to 
slow down unnecessarily a bunch of USB disks that could work just fine 
at the higher UAS transfer rates.

By the way, does the enclosure have its own power source, or does it 
rely entirely on power provided over the USB cable?  Note that UAS can 
use more power than the older mass-storage protocols, because it queues 
more operations in rapid succession (which is also why it runs faster).

Alan Stern

Re: [PATCH] usb: storage: add quirks for VIA VL817 USB3-SATA bridge

[Tobias Jakobi]

On 9/27/21 5:04 PM, Alan Stern wrote:

> On Sun, Sep 26, 2021 at 08:14:46PM +0200, Tobias Jakobi wrote:
>> Hello Alan,
>>
>> On 9/21/21 6:42 PM, Alan Stern wrote:
>>> On Tue, Sep 21, 2021 at 06:06:45PM +0200, Tobias Jakobi wrote:
>>>> Hi Alan,
>>>>
>>>> sorry but your analysis of the log is wrong. Nothing was disconnected or
>>>> unplugged when the device behaves this way. The enclosure is connected to
>>>> the power the entire time, and the same applies to the physical USB
>>>> connection to my system.
>>> That may be so, but if it is then the log extract you included with the
>>> patch is very misleading.  For instance, you didn't include any part of
>>> the log before and leading up to the line saying "usb 2-1.2: USB
>>> disconnect, device number 4".  Thus there is no way for the reader to
>>> tell what caused this event, whether it was a physical unplug or not.
>> I included the part of the kernel log which shows how the issue manifests
>> itself. Do you think I'm so stupid as to believe I could prevent a physical
>> unplug of the enclosure by blacklisting UAS? Really, this is getting
>> ridiculous...
> No, I don't believe that.  However, I do believe you are so stupid as
> to post a patch with a description that does not justify it and
> includes a lot of irrelevant details.  (Hey, you brought up the issue
> of how stupid you are!  :-)
>
>>>> To make things very clear: This happens in under five minutes after having
>>>> powered up the enclosure and starting a file transfer to the installed RAID.
>>>> After blacklisting UASP the enclosure works perfectly fine for hours. I hope
>>>> this clears things up.
>>> You didn't answer my question about using NO_ATA_1X instead of
>>> IGNORE_UAS.  This is a perfect example of one of the dangers of
>>> top-posting -- it makes it far too easy for people to miss important
>>> points in the email they are replying to.  (Hint: Don't top-post!)
>> I did not answer this question, because I didn't have the answer to it yet.
>> I have tested your suggestion today, but sadly I'm running into the same
>> type of problem with NO_ATA_1X. You can find the complete kernel log here:
>> https://www.math.uni-bielefeld.de/~tjakobi/archive/dmesg_VL817.log
> Good, thank you.  Unfortunately the log doesn't include any smoking
> guns pointing to an underlying cause.

I'm open to suggestions regarding identifying the cause. As you might 
guess, I'm also not happy that I had to disable UAS for the enclosure -- 
in particular since I selected this particular product because it was 
advertised with having support.


>
>> The RAID1 is broken after such an event.
>>
>> With best wishes,
>> Tobias
>>
>>
>>> Alan Stern
>>>
>>>> With best wishes,
>>>> Tobias
>>>>
>>>> On 9/21/21 5:13 PM, Alan Stern wrote:
>>>>> On Tue, Sep 21, 2021 at 12:17:52PM +0200, Tobias Jakobi wrote:
>>>>>> The VL817 is used in the RaidSonic Icy Box IB-3740-C31 enclosure. The enclosure
>>>>>> is advertised as having UASP support, but appears to have problems with 4Kn
>>>>>> drives (test was done with two Seagate Exos X, 12TB).
>>>>>>
>>>>>> Disable UAS for the VL817 as it behaves highly unstable:
> I still have to wonder if the enclosure works okay with other types of
> disk drive.  And if it doesn't, why don't these errors show up on
> Windows systems?  Or on other VIA enclosures?

I only experienced this after installing the two 4Kn drives, never with 
the two 512e drives that were installed first. My guess is that 4Kn 
drives are still rare and if they're used, then natively attached to a 
SATA port. I don't have any Windows system here to test this, and even 
if, I wouldn't know how to assemble the RAID1 there anyway.


> That's why I'm cautious about accepting this patch.  I don't want to
> slow down unnecessarily a bunch of USB disks that could work just fine
> at the higher UAS transfer rates.

I understand. If that's the case, I'm just going to continue to keep the 
patch in my local kernel tree.


>
> By the way, does the enclosure have its own power source, or does it
> rely entirely on power provided over the USB cable?  Note that UAS can
> use more power than the older mass-storage protocols, because it queues
> more operations in rapid succession (which is also why it runs faster).

This is the enclosure:
https://icybox.de/product.php?id=155

It has a external power supply (quite a bulky one) and it does not work 
without it. So it doesn't draw anything (significant) from the USB 
cable. I first also suspected this to be a power supply related problem, 
but I discarded that idea since the whole thing works as MSC. I can't 
imagine the power draw to be so much different for UAS, but maybe I'm 
just naive there.

With best wishes,
Tobias


>
> Alan Stern

Re: [PATCH] usb: storage: add quirks for VIA VL817 USB3-SATA bridge

[Alan Stern]

On Fri, Oct 01, 2021 at 10:36:59PM +0200, Tobias Jakobi wrote:
> On 9/27/21 5:04 PM, Alan Stern wrote:
> 
> > Good, thank you.  Unfortunately the log doesn't include any smoking
> > guns pointing to an underlying cause.
> 
> I'm open to suggestions regarding identifying the cause. As you might
> guess, I'm also not happy that I had to disable UAS for the enclosure -- in
> particular since I selected this particular product because it was
> advertised with having support.

One thing you might try is to capture a usbmon trace showing what 
happens, starting from when the USB cable is attached up to the point 
when the failure occurs.  However, I'm doubtful that the trace will 
show anything useful, since the kernel log indicated that the first 
failed command was an ordinary write.

> > I still have to wonder if the enclosure works okay with other types of
> > disk drive.  And if it doesn't, why don't these errors show up on
> > Windows systems?  Or on other VIA enclosures?
> 
> I only experienced this after installing the two 4Kn drives, never with the
> two 512e drives that were installed first. My guess is that 4Kn drives are
> still rare and if they're used, then natively attached to a SATA port. I
> don't have any Windows system here to test this, and even if, I wouldn't
> know how to assemble the RAID1 there anyway.

It's entirely possible that this failure has nothing to do with the 
use of RAID.

> > That's why I'm cautious about accepting this patch.  I don't want to
> > slow down unnecessarily a bunch of USB disks that could work just fine
> > at the higher UAS transfer rates.
> 
> I understand. If that's the case, I'm just going to continue to keep the
> patch in my local kernel tree.

You don't even need a patch; you can accomplish the same effect 
simply by specifying a module parameter for usb-storage:

	quirks=2109:0715:u

(You might have to rebuild your initramfs image, if you need the 
quirk to take effect during the early stages of boot-up.)

> > By the way, does the enclosure have its own power source, or does it
> > rely entirely on power provided over the USB cable?  Note that UAS can
> > use more power than the older mass-storage protocols, because it queues
> > more operations in rapid succession (which is also why it runs faster).
> 
> This is the enclosure:
> https://icybox.de/product.php?id=155
> 
> It has a external power supply (quite a bulky one) and it does not work
> without it. So it doesn't draw anything (significant) from the USB cable. I
> first also suspected this to be a power supply related problem, but I
> discarded that idea since the whole thing works as MSC. I can't imagine the
> power draw to be so much different for UAS, but maybe I'm just naive there.

Yeah, you might be surprised.  There have been other instances of 
people submitting patches to disable UAS for their drives, when it 
seemed quite likely that lack of power was the underlying cause.  
Although to be fair, these occurred in very power-constrained 
situations, such as running a bus-powered drive attached to a 
Raspberry Pi.  Not like your case.  Still, you see why I had to ask.

Alan Stern