Re: high speed data to usb disk makes the kernel think that is has been unmounted

From: Patrik Nilsson <nipatriknilsson@gmail.com>
To: Greg KH <gregkh@linuxfoundation.org>
Cc: Alan Stern <stern@rowland.harvard.edu>, linux-usb@vger.kernel.org
Subject: Re: high speed data to usb disk makes the kernel think that is has been unmounted
Date: Tue, 18 Aug 2020 20:22:55 +0200	[thread overview]
Message-ID: <83a5be3e-9f93-e601-5161-cbb5d1957c5a@gmail.com> (raw)
In-Reply-To: <20200818085502.GB28036@kroah.com>

Hi Greg,

Thank you for debugging the issue!

Is there a way to set a speed limit to the device? I mean like setting 
the usb-quirks. (i.e. usb-storage.quirks=174c:55aa:u)

Instead of buying a new enclosing for the drive... :(

Best regards,
Patrik

On 18/08/2020 10:55, Greg KH wrote:
> On Sun, Aug 16, 2020 at 05:31:09PM +0200, Patrik Nilsson wrote:
>> [ 7981.203457] perf: interrupt took too long (3133 > 3132), lowering kernel.perf_event_max_sample_rate to 63750
>> [ 8263.612587] usb 2-2: USB disconnect, device number 3
> Your device disconnected itself from the bus, that is an electrical
> thing that the kernel can not control/prevent/change.
>
>> [ 8263.628508] blk_update_request: I/O error, dev sdb, sector 2678161400 op 0x1:(WRITE) flags 0x0 phys_seg 1674 prio class 0
>> [ 8263.628515] Buffer I/O error on dev sdb, logical block 2678161400, lost async page write
>> [ 8263.628519] Buffer I/O error on dev sdb, logical block 2678161401, lost async page write
>> [ 8263.628521] Buffer I/O error on dev sdb, logical block 2678161402, lost async page write
>> [ 8263.628524] Buffer I/O error on dev sdb, logical block 2678161403, lost async page write
>> [ 8263.628526] Buffer I/O error on dev sdb, logical block 2678161404, lost async page write
>> [ 8263.628528] Buffer I/O error on dev sdb, logical block 2678161405, lost async page write
>> [ 8263.628530] Buffer I/O error on dev sdb, logical block 2678161406, lost async page write
>> [ 8263.628533] Buffer I/O error on dev sdb, logical block 2678161407, lost async page write
>> [ 8263.628538] Buffer I/O error on dev sdb, logical block 2678161408, lost async page write
>> [ 8263.628541] Buffer I/O error on dev sdb, logical block 2678161409, lost async page write
>> [ 8263.629271] blk_update_request: I/O error, dev sdb, sector 2678163074 op 0x1:(WRITE) flags 0x4000 phys_seg 2048 prio class 0
>> [ 8263.630073] blk_update_request: I/O error, dev sdb, sector 2678165122 op 0x1:(WRITE) flags 0x0 phys_seg 6 prio class 0
>> [ 8263.630146] blk_update_request: I/O error, dev sdb, sector 2678165128 op 0x1:(WRITE) flags 0x800 phys_seg 112 prio class 0
>> [ 8263.630879] sd 4:0:0:0: [sdb] Synchronizing SCSI cache
>> [ 8263.630912] sd 4:0:0:0: [sdb] Synchronize Cache(10) failed: Result: hostbyte=DID_NO_CONNECT driverbyte=DRIVER_OK
>> [ 8263.630974] scsi 4:0:0:0: rejecting I/O to dead device
>> [ 8263.630980] blk_update_request: I/O error, dev sdb, sector 2678165240 op 0x1:(WRITE) flags 0x0 phys_seg 104 prio class 0
>> [ 8275.232507] usb 2-2: new SuperSpeed Gen 1 USB device number 10 using xhci_hcd
> And now it came back.
>
> So I really think this device is just sucking up too much power and
> eventually the hub disconnects it (or it disconnects itself by dropping
> the voltage too low) and then adds itself again.
>
> If you write data "slower" to it, it's fine, but by writing 0's to your
> device, it's constantly hammered and that's the hardest thing for a
> flash device to handle (flash starts out at all bits set and then drops
> them as needed to 0, so you are asking for the most work possible to be
> done by the flash chips.)
>
> It might be better to just buy a different disk drive :)
>
> sorry,
>
> greg k-h

-- 
PGP-key fingerprint: 1B30 7F61 AF9E 538A FCD6  2BE7 CED7 B0E4 3BF9 8D6C