data loss reading from SD card with 4.0.1 (and 4.0.0) kernels?

[Matthew Miller]

Okay, not to panic everyone too much, but: if you have an SD card with
a bunch of JPEG files on it, can you try copying files off in a loop to
see if you can recreate
https://bugzilla.redhat.com/show_bug.cgi?id=1220096?

(In my situation, with a Lenovo laptop with built-in reader.)

The big caveats are: 1. It's late at night and 2. I haven't tested with
a second card, _but_, it seems I can go back and forth between 3.19 and
4.0 and reliably read -- or not.

[Joerg Lechner]

Hi,
I always like to play a little bit with Fedora. My system is an Acer Laptop Aspire E15, F22 (Kernel 4.0.1-300.fc22.x86_64) on an external disk, internal card reader, High quality SD card with 1133 jpg files (about 4GB jpgs). So far I downloaded successfully with rapid-foto-downloader 0.4.10-2.fc21.noarch. Can You tell me how You code Your loop to download Your jpegs? I can try to test, but I think I should use the same code (application) as You use.
Kind Regards

-----Ursprüngliche Mitteilung----- 
Von: Matthew Miller 
An: test 
Verschickt: So, 10 Mai 2015 5:29 am
Betreff: data loss reading from SD card with 4.0.1 (and 4.0.0) kernels?

Okay, not to panic everyone too much, but: if you have an SD card with
a bunch
of JPEG files on it, can you try copying files off in a loop to
see if you can
recreate
https://bugzilla.redhat.com/show_bug.cgi?id=1220096?

(In my situation,
with a Lenovo laptop with built-in reader.)

The big caveats are: 1. It's late
at night and 2. I haven't tested with
a second card, _but_, it seems I can go
back and forth between 3.19 and
4.0 and reliably read -- or not.

[Joerg Lechner]

Hi
copied now with command string
[root@linux 100NIKON]# cp DSCN* /home/joerg/Bilder/jpg-test
200 jpgs successfully, this means without error message. Is that, what You wanted?
Done on an Acer Aspire E15, fast SD card, Kernel 4.0.1-300.fc22.x86_64.
F22 on an external disk. Can not reproduce this error desribed in bug1220096.
Kind Regards

-----Ursprüngliche Mitteilung----- 
Von: Matthew Miller 
An: test 
Verschickt: So, 10 Mai 2015 5:29 am
Betreff: data loss reading from SD card with 4.0.1 (and 4.0.0) kernels?

Okay, not to panic everyone too much, but: if you have an SD card with
a bunch
of JPEG files on it, can you try copying files off in a loop to
see if you can
recreate
https://bugzilla.redhat.com/show_bug.cgi?id=1220096?

(In my situation,
with a Lenovo laptop with built-in reader.)

The big caveats are: 1. It's late
at night and 2. I haven't tested with
a second card, _but_, it seems I can go
back and forth between 3.19 and
4.0 and reliably read -- or not.

[Matthew Miller]

On Sun, May 10, 2015 at 07:04:51AM -0400, Joerg Lechner wrote:
> copied now with command string
> [root@linux 100NIKON]# cp DSCN* /home/joerg/Bilder/jpg-test
> 200 jpgs successfully, this means without error message. Is that, what You wanted?

Yep, basically that. However, do that over and over again:

while sleep 1; do cp -v /home/joerg/Bilder/jpg-test/; done

while running `journalctl -f` in another window. (As root if you're not
in the wheel group.)

[Joerg Lechner]

Hi,
did this:
[root@linux DCIM]# dir
100NIKON  101NIKON  102NIKON  103NIKON	104NIKON  105NIKON
[root@linux DCIM]# while sleep 1
> do
> cp -v *NIKON/DSCN* /home/joerg/Bilder/jpg-test
> done

„100NIKON/DSCN8479.JPG“ -> „/home/joerg/Bilder/jpg-test/DSCN8479.JPG“
„100NIKON/DSCN8480.JPG“ -> „/home/joerg/Bilder/jpg-test/DSCN8480.JPG“
„100NIKON/DSCN8481.JPG“ -> „/home/joerg/Bilder/jpg-test/DSCN8481.JPG“
„100NIKON/DSCN8482.JPG“ -> „/home/joerg/Bilder/jpg-test/DSCN8482.JPG“
etc.

These are 1137 jpegs in 1 run, did about 5 runs, cannot reproduce this error.

journalctl -in my opinion?- doesn't show anything relevant.:

Mai 10 21:44:04 linux.fritz.box org.gtk.vfs.Daemon[1852]: ** (process:2719): WARNING **: send_infos_cb: Keine derartige Schnittstelle »org.gtk.vfs.Enumerator« des Objekts im Pfad /org/gtk/vfs/client/enumerator/3202 (g-dbus-error-quark, 19)
Mai 10 21:44:04 linux.fritz.box org.gtk.vfs.Daemon[1852]: ** (process:2719): WARNING **: send_infos_cb: Keine derartige Schnittstelle »org.gtk.vfs.Enumerator« des Objekts im Pfad /org/gtk/vfs/client/enumerator/3202 (g-dbus-error-quark, 19)
Mai 10 21:44:04 linux.fritz.box org.gtk.vfs.Daemon[1852]: ** (process:2719): WARNING **: send_done_cb: Keine derartige Schnittstelle »org.gtk.vfs.Enumerator« des Objekts im Pfad /org/gtk/vfs/client/enumerator/3202 (g-dbus-error-quark, 19)
Mai 10 21:44:04 linux.fritz.box org.gtk.vfs.Daemon[1852]: ** (process:2719): WARNING **: send_infos_cb: Keine derartige Schnittstelle »org.gtk.vfs.Enumerator« des Objekts im Pfad /org/gtk/vfs/client/enumerator/3203 (g-dbus-error-quark, 19)
Mai 10 21:44:04 linux.fritz.box org.gtk.vfs.Daemon[1852]: ** (process:2719): WARNING **: send_infos_cb: Keine derartige Schnittstelle »org.gtk.vfs.Enumerator« des Objekts im Pfad /org/gtk/vfs/client/enumerator/3203 (g-dbus-error-quark, 19)
Mai 10 21:48:51 linux.fritz.box systemd-journal[540]: Suppressed 66682 messages from /user.slice/user-1000.slice
Mai 10 21:48:51 linux.fritz.box org.gnome.Nautilus[1852]: Gtk-Message: GtkDialog mapped without a transient parent. This is discouraged.

Any further idea?
Kind Regards

-----Ursprüngliche Mitteilung----- 
Von: Matthew Miller 
An: For testing and quality assurance of Fedora releases 
Verschickt: So, 10 Mai 2015 9:00 pm
Betreff: Re: data loss reading from SD card with 4.0.1 (and 4.0.0) kernels?

On Sun, May 10, 2015 at 07:04:51AM -0400, Joerg Lechner wrote:
> copied now with

command string
> [root@linux 100NIKON]# cp DSCN* /home/joerg/Bilder/jpg-test

200 jpgs successfully, this means without error message. Is that, what You
wanted?

Yep, basically that. However, do that over and over again:

while
sleep 1; do cp -v /home/joerg/Bilder/jpg-test/; done

while running `journalctl
-f` in another window. (As root if you're not
in the wheel group.)

[Matthew Miller]

On Sun, May 10, 2015 at 03:55:33PM -0400, Joerg Lechner wrote:
> These are 1137 jpegs in 1 run, did about 5 runs, cannot reproduce this error.
> journalctl -in my opinion?- doesn't show anything relevant.:

Okay, thanks. I'll see if I can make it happen with another card, and
with another card reader. Both of my test systems are Lenovo laptops,
and it's possible there is something hardware-specific.

[Chris Murphy]

On Sat, May 9, 2015 at 9:28 PM, Matthew Miller  wrote:
> Okay, not to panic everyone too much, but: if you have an SD card with
> a bunch of JPEG files on it, can you try copying files off in a loop to
> see if you can recreate
> https://bugzilla.redhat.com/show_bug.cgi?id=1220096?

I haven't seen this. But I've had severe performance problems with
kernel 4.0.0 and 4.0.1. Normally I get ~38MB/s off this SD card (Btrfs
on LUKS file on HFS+ on sdcard, yes I'm trying to break shit), but
with 4.0.0 and 4.0.1 I non-deterministically get almost complete
stalls with less than 1MB/s transfers, including direct to the card so
the other layers aren't a factor. But Btrfs reports no corruptions
throughout, including doing full volume scrubs. But the performance
was so bad, and I never got any sort of call trace or errors, so I
just stopped using it rather than punish it and myself more.

Today with 4.1.0-0.rc2.git0.1.fc23.x86_64 the problem isn't happening,
I'm getting 38MB/s transfers. For now.

[Peter Robinson]

On Sun, May 10, 2015 at 4:28 AM, Matthew Miller
 wrote:
> Okay, not to panic everyone too much, but: if you have an SD card with
> a bunch of JPEG files on it, can you try copying files off in a loop to
> see if you can recreate
> https://bugzilla.redhat.com/show_bug.cgi?id=1220096?
>
> (In my situation, with a Lenovo laptop with built-in reader.)
>
> The big caveats are: 1. It's late at night and 2. I haven't tested with
> a second card, _but_, it seems I can go back and forth between 3.19 and
> 4.0 and reliably read -- or not.

I've not seen it with my x220 pushing gigs over my SD slot flashing
out ARM images with 4.0.1 (laptop currently up 10 days). It might be a
different use case, the writing out of ARM images is primarily dd and
resizing/moving images.

Going back to around 3.16 I use to get a limited amount of uptime
before the SD card would cease to work and you'd need to reboot to
make it work again.

[Matthew Miller]

On Sun, May 10, 2015 at 09:03:14PM +0100, Peter Robinson wrote:
> I've not seen it with my x220 pushing gigs over my SD slot flashing
> out ARM images with 4.0.1 (laptop currently up 10 days). It might be a
> different use case, the writing out of ARM images is primarily dd and
> resizing/moving images.

Out of curiousity, does that device show up as generic storage or
mmcblk?

I was *not* able to repoduce using a separate transcend USB reader
reader, for whatever it's worth -- but I *could* make it happen with
another card, and both cards seem fine if I go back to 3.19.6.

[Peter Robinson]

On Mon, May 11, 2015 at 5:28 PM, Matthew Miller
 wrote:
> On Sun, May 10, 2015 at 09:03:14PM +0100, Peter Robinson wrote:
> > I've not seen it with my x220 pushing gigs over my SD slot flashing
> > out ARM images with 4.0.1 (laptop currently up 10 days). It might be a
> > different use case, the writing out of ARM images is primarily dd and
> > resizing/moving images.
>
> Out of curiousity, does that device show up as generic storage or
> mmcblk?

mmcblk

0d:00.0 System peripheral: Ricoh Co Ltd MMC/SD Host Controller (rev 07)

[Joerg Lechner]

Hi,
if it's of interest to show the configuration, where I could not reproduce this error:
Laptop Acer Aspire E15
Internal Card Reader Realtek PCIE
SD Card Sandisk Extreme (80MB/sec) 8GB.
Kind Regards

-----Ursprüngliche Mitteilung----- 
Von: Matthew Miller 
An: For testing and quality assurance of Fedora releases 
Verschickt: Mo, 11 Mai 2015 6:28 pm
Betreff: Re: data loss reading from SD card with 4.0.1 (and 4.0.0) kernels?

On Sun, May 10, 2015 at 09:03:14PM +0100, Peter Robinson wrote:
> I've not seen

it with my x220 pushing gigs over my SD slot flashing
> out ARM images with

4.0.1 (laptop currently up 10 days). It might be a
> different use case, the

writing out of ARM images is primarily dd and
> resizing/moving images.

Out of
curiousity, does that device show up as generic storage or
mmcblk?

I was *not*
able to repoduce using a separate transcend USB reader
reader, for whatever it's
worth -- but I *could* make it happen with
another card, and both cards seem
fine if I go back to 3.19.6.