USB storage (pendrive) problems

USB storage (pendrive) problems

[Attila BODY]

Hi,

I have some weird problems with my pendrives recently. I just compile a
2.6.9 to check if the problem is still exists there.

current kernel is 2.6.10-rc3 and the situation is the following:

If I copy more than few megabytes to the drive, the activity LED keeps
flashing forever. sync, umount keeps runing forever, normal reboot is
inpossible (alt+sysreq+b seems to work)

Tested with usb 1.1 and 2.0 pendrives, behaviour is the same.

Any ideas?

P.S: Please cc directly to me, I am not subscribed, thanks.

compi

Re: USB storage (pendrive) problems

[Andrew Walrond]

On Monday 20 December 2004 21:54, Attila BODY wrote:
> Hi,
>
> I have some weird problems with my pendrives recently. I just compile a
> 2.6.9 to check if the problem is still exists there.
>
> current kernel is 2.6.10-rc3 and the situation is the following:
>
> If I copy more than few megabytes to the drive, the activity LED keeps
> flashing forever. sync, umount keeps runing forever, normal reboot is
> inpossible (alt+sysreq+b seems to work)
>
> Tested with usb 1.1 and 2.0 pendrives, behaviour is the same.
>

I'm doing exactly that with 2.6.10-rc3. umount does take a very long time (but 
I had just written 600Mb+ over usb 1.1)

Are you sure it doesn't come back if you leave it long enough?

Do the throughput sums; you'll be suprised how long it takes to send more than 
a few Mb over usb 1.1 (1.5Mb/s). Eg 600Mb = 7minutes

Usb 2 should be much faster; Do you have EHCI loaded?

Andrew Walrond

Re: USB storage (pendrive) problems

[Randy.Dunlap]

Andrew Walrond wrote:
> On Monday 20 December 2004 21:54, Attila BODY wrote:
> 
>>Hi,
>>
>>I have some weird problems with my pendrives recently. I just compile a
>>2.6.9 to check if the problem is still exists there.
>>
>>current kernel is 2.6.10-rc3 and the situation is the following:
>>
>>If I copy more than few megabytes to the drive, the activity LED keeps
>>flashing forever. sync, umount keeps runing forever, normal reboot is
>>inpossible (alt+sysreq+b seems to work)
>>
>>Tested with usb 1.1 and 2.0 pendrives, behaviour is the same.
>>
> 
> 
> I'm doing exactly that with 2.6.10-rc3. umount does take a very long time (but 
> I had just written 600Mb+ over usb 1.1)
> 
> Are you sure it doesn't come back if you leave it long enough?
> 
> Do the throughput sums; you'll be suprised how long it takes to send more than 
> a few Mb over usb 1.1 (1.5Mb/s). Eg 600Mb = 7minutes
> 
> Usb 2 should be much faster; Do you have EHCI loaded?

and which usb driver are you using?
ub or usb-storage?  (what's the /dev name that you mount?)

-- 
~Randy

Re: USB storage (pendrive) problems

[Andrew Walrond]

On Monday 20 December 2004 23:21, Randy.Dunlap wrote:
>
> and which usb driver are you using?
> ub or usb-storage?  (what's the /dev name that you mount?)

My pen gets mounted as /dev/sdc1 on the usb1 x86_64 box which I mostly use , 
and /dev/uba1 on a usb2 P4 box. The pen seems to work fine on both.

Andrew

Re: USB storage (pendrive) problems

[Attila BODY]

On Mon, 2004-12-20 at 15:21 -0800, Randy.Dunlap wrote:
> Andrew Walrond wrote:
> > On Monday 20 December 2004 21:54, Attila BODY wrote:
> > 
> >>Hi,
> >>
> >>I have some weird problems with my pendrives recently. I just compile a
> >>2.6.9 to check if the problem is still exists there.
> >>
> >>current kernel is 2.6.10-rc3 and the situation is the following:
> >>
> >>If I copy more than few megabytes to the drive, the activity LED keeps
> >>flashing forever. sync, umount keeps runing forever, normal reboot is
> >>inpossible (alt+sysreq+b seems to work)
> >>
> >>Tested with usb 1.1 and 2.0 pendrives, behaviour is the same.
> >>
> > 
> > 
> > I'm doing exactly that with 2.6.10-rc3. umount does take a very long time (but 
> > I had just written 600Mb+ over usb 1.1)
> > 
> > Are you sure it doesn't come back if you leave it long enough?
> > 
> > Do the throughput sums; you'll be suprised how long it takes to send more than 
> > a few Mb over usb 1.1 (1.5Mb/s). Eg 600Mb = 7minutes
> > 
> > Usb 2 should be much faster; Do you have EHCI loaded?
> 
> and which usb driver are you using?
> ub or usb-storage?  (what's the /dev name that you mount?)
Hi,

Here is the log snipet when plug the pendrive in, i thin It should be
clear from this that is uses ehci_hcd, ub, hotplug and udev. For 1.1 it
uses uhci_hcd

Dec 21 08:58:51 smiley kernel: ehci_hcd 0000:00:10.3: port 6 high speed
Dec 21 08:58:51 smiley kernel: ehci_hcd 0000:00:10.3: GetStatus port 6
status 001005 POWER sig=se0  PE CONNECT
Dec 21 08:58:51 smiley kernel: usb 1-6: new device strings: Mfr=2,
Product=3, SerialNumber=4
Dec 21 08:58:51 smiley kernel: usb 1-6: default language 0x0409
Dec 21 08:58:51 smiley kernel: usb 1-6: Product: Cruzer Mini
Dec 21 08:58:51 smiley kernel: usb 1-6: Manufacturer: SanDisk
Corporation
Dec 21 08:58:51 smiley kernel: usb 1-6: SerialNumber: 00567026
Dec 21 08:58:51 smiley kernel: usb 1-6: hotplug
Dec 21 08:58:51 smiley kernel: usb 1-6: adding 1-6:1.0 (config #1,
interface 0)
Dec 21 08:58:51 smiley kernel: usb 1-6:1.0: hotplug
Dec 21 08:58:51 smiley kernel: ub 1-6:1.0: usb_probe_interface
Dec 21 08:58:51 smiley kernel: ub 1-6:1.0: usb_probe_interface - got id
Dec 21 08:58:51 smiley usb.agent[6288]:      ub: already loaded
Dec 21 08:58:52 smiley kernel: uba: device 3 capacity nsec 512000 bsize
512
Dec 21 08:58:52 smiley kernel: uba: device 3 capacity nsec 512000 bsize
512
Dec 21 08:58:52 smiley kernel: /dev/ub/a: p1
Dec 21 08:58:52 smiley udev[6328]: creating device node '/dev/uba1'
Dec 21 08:58:52 smiley udev[6321]: creating device node '/dev/uba'

The problem is reproducable with 2.6.9. Unfortunately I have no time
right now to check 2.6.8.1 if the problem already presists there.
It seems however that the 1.1 device recovers if I wait long enough, but
I left the 2.0 device here all night and it did not. To make it worse,
it seems that the whole USB 2.0 system went down, there was no activity
even if i reconnected the pen, had to reboot to be detected and powered
on again (normal reboot was not an option again, so I had to use
alt-sysrq-B).

Hope this helps to track down this problem.

compi

P.S.: Please cc, I'm not subscribed.

Re: USB storage (pendrive) problems

[Pete Zaitcev]

On Tue, 21 Dec 2004 09:35:03 +0200, Attila BODY <compi@freemail.hu> wrote:

> > >>If I copy more than few megabytes to the drive, the activity LED keeps
> > >>flashing forever. sync, umount keeps runing forever, normal reboot is
> > >>inpossible (alt+sysreq+b seems to work)
> > >>
> > >>Tested with usb 1.1 and 2.0 pendrives, behaviour is the same.

> Dec 21 08:58:52 smiley kernel: /dev/ub/a: p1

> The problem is reproducable with 2.6.9. [...]

> I left the 2.0 device here all night and it did not. To make it worse,
> it seems that the whole USB 2.0 system went down, there was no activity
> even if i reconnected the pen, had to reboot to be detected and powered
> on again (normal reboot was not an option again, so I had to use
> alt-sysrq-B).

Thanks for the note. Unfortunately, 2.6.9 apparently has problems with
its virtual memory and write throttling, so I'm sure a few people suffering
from it will jump on this thread. It's essential to split root causes.
Your case may be different from the original poster's, who apparently
had the same deal with both ub and usb-storage. First things first, did
you try to set CONFIG_BLK_DEV_UB to off? What happens if you do?

-- Pete

Re: USB storage (pendrive) problems

[Attila BODY]

Hi,

On Tue, 2004-12-21 at 13:15 -0800, Pete Zaitcev wrote:

> Thanks for the note. Unfortunately, 2.6.9 apparently has problems with
> its virtual memory and write throttling, so I'm sure a few people suffering
> from it will jump on this thread. It's essential to split root causes.
> Your case may be different from the original poster's, who apparently
> had the same deal with both ub and usb-storage. First things first, did
> you try to set CONFIG_BLK_DEV_UB to off? What happens if you do?

It works then, so the problem seems to be with the ub driver, but it is
still a question why the 1.1 device seems to wprk better with it.

brgds,

compi