Calibre keeps disconnecting my kindle.

[Pigi_102]

Hi all,
I'm trying to figure out what's the problem.

I have a Kindle ( super basic edition ) that has started to give me some problems while sending books to it.

What happens is this.
I run calibre, which starts normally. I then connect my kindle to usb port and it get detected from the system ( linux ) and calibre.
I can see what's inside it but if I try to send a new book to it I see the kindle screen as it has been disconnected, and calibre gives me this error :
QStandardPaths: XDG_RUNTIME_DIR not set, defaulting to '/tmp/runtime-pigi'
[Errno 13] Permission denied: '/mnt/sdb1/documents'

Traceback (most recent call last):
File "site-packages/calibre/gui2/device.py", line 90, in run
File "site-packages/calibre/gui2/device.py", line 594, in _upload_books
File "site-packages/calibre/devices/usbms/driver.py", line 324, in upload_books
File "site-packages/calibre/devices/usbms/device.py", line 998, in create_upload_path
File "os.py", line 150, in makedirs
File "os.py", line 157, in makedirs
OSError: [Errno 13] Permission denied: '/mnt/sdb1/documents'

and in the dmesg I get:
[ 1220.207223] sd 6:0:0:0: [sdb] tag#0 access beyond end of device
[ 1220.207228] blk_update_request: I/O error, dev sdb, sector 17 op 0x1WRITE) flags 0x0 phys_seg 1 prio class 0
[ 1220.207231] Buffer I/O error on dev sdb1, logical block 1, lost async page write
[ 1220.207250] sd 6:0:0:0: [sdb] tag#0 access beyond end of device
[ 1220.207252] blk_update_request: I/O error, dev sdb, sector 8380 op 0x1WRITE) flags 0x0 phys_seg 2 prio class 0
[ 1220.207254] Buffer I/O error on dev sdb1, logical block 8364, lost async page write
[ 1220.207255] Buffer I/O error on dev sdb1, logical block 8365, lost async page write

I've tracked down this one as it happens because the mount point get somehow disconnected.


At this point the only chance I get is to disconnect the usb port and connect again, but as soon as I try to write to the kindle I get the same error.

Initially I thought that the problm could have been on kindle filesystem or flash, thus I did an fsck to it and also tried on an other computer but with the same results.

If, instead, I do close calibre, mount the kindle by hand ( mount /dev/sdb1 /mnt/sdb1 ) I can write to it the books or whatever I want without any problem.
I can the read those book without any problem.
This mean ( to me ) that kindle is perfectly working, and also that the usb side of my linux is working too.

Any hint on how to track down the problem ?

Thanks in advance

Pigi_102

[jhowell]

This thread may be related: [Kindle] calibre losing sight of device after one transfer

[ilvez]

Originally Posted by jhowell

This thread may be related: [Kindle] calibre losing sight of device after one transfer

This thread mentions this bug https://bugs.launchpad.net/calibre/+bug/1834641, but I'm currently a bit confused whether it's actually taken care of or not. It seems like fix has been released, but I have the exact same issue as Pigi_102. I don't want to cause noise in that bug report if that's not connected.

I can reproduce this every single time with Calibre (latest) + latest Kindle update (5.13.1) + Linux 5.7.6. And it has been through different version combinations quite some time now.

Should I open new bug report?

dmesg output: https://pastebin.com/8Dx4vJHM

[kovidgoyal]

the linux kernel bug was never fixed. A work around for it was added to calibre a while ago. If you are experiencing disconnects it will be because either you are running a distro calibre and hence not the latest or some other unrelated issue.

[ilvez]

Thank you for your fast response. I reproduced this with non-distro Calibre, just downloaded latest pre-compiled binary, extracted and run.

Normal mount + copy works without disconnects.

So I guess this is a new bug then? Should I open new bug ticket?

[ilvez]

Hm. Actually I was bit hasty with my response. I had Debian calibre installed. I remove it completely and follow installation instructions and see the result. I was suspecting that system libraries are actually being read or something like this.

EDIT: I followed official installer and confirmed. It behaves the same.

[kovidgoyal]

Opening a new bug is not going to help. I cannot reproduce this behavior with my kindle, as such there is nothing I can do to help you. I am fairly certain you are still using an outdated calibre, since no one else has reported this.

[ilvez]

So I took these steps to do clean install yesterday:

sudo apt purge calibre
sudo apt autoremove
sudo -v && wget -nv -O- https://download.calibre-ebook.com/linux-installer.sh | sudo sh /dev/stdin


ilvez@merkuur ~ % ls -l `which calibre`
lrwxrwxrwx 1 root root 20 Jul 19 09:23 /usr/bin/calibre -> /opt/calibre/calibre

Then I just ran calibre.

To be extra sure, I then ran:

sudo /usr/bin/calibre-uninstall
sudo rm -rf /opt/calibre
sudo -v && wget -nv -O- https://download.calibre-ebook.com/linux-installer.sh | sudo sh /dev/stdin

Then I ran calibre again (version 4.21) and reproduced it.

If you have any ideas what do debug here, I'm willing to do any extra step. I can recompile kernel or try some older Linux. I'm just don't know how to provide more useful information at this point.

I have been years Calibre user and really appriciate your job done on the software. This error is currently major headache for me to use the software. Do throw you bone here, I'm actually thinking of becoming patreon as well, since I really am grateful for Calibre. But I understand if this kind of errors are just party-poopers and stopping you for doing some fun development work.

[kovidgoyal]

Without being able to reproduce, I'm afraid there is not anything I can do to fix it, since this can only be a kernel bug, I need to be able to experiment to find a workaround. My kindle paperwhite works perfectly with up-to-date firmware, calibre 4.21 and Linux kernel 5.7.9-arch1-1

If you want to try older kernels. go back to the version referenced in the bug report, before they broke the VFAT driver. If it is indeed the same bug then that should confirm it.

[ilvez]

Thank you.

I skipped currently running older kernels. Instead I went and read about the lengthy discussion about underlying kernel bug https://bugzilla.kernel.org/show_bug.cgi?id=203973

The last comment I found workaround. I don't know the implications for changing write_cache to "write through" with:

   echo "write through" > /sys/block/<dev>/queue/write_cache

... but it seems to work.

I'm currently trying to figure out how to make this into a udev rule.