KDL when working many files

[X512]

This is hrev42696.

When I try to copy my BFS partition(about 130000 inodes) to removable HDD I get KDL. As I understand there are something wrong with the block cache. Checkfs says that all is OK.

I tried to copy 3 times with Tracker and Terminal and get same result.

[X512]

KDL

[anevilyak]

That actually looks like your removable disk is timing out a request.

[mmlr]

When you're in KDL can you run "syslog | tail 150" (or just syslog and page through to the end) and get the last few lines. It should show what's going on, i.e. why it failed to read the blocks. The timeout might be preceeded by some errors that'd need special recovery. It's also possible that a burst write is simply getting so large that it actually does take this much time. The usb_disk driver (still) doesn't do a flexible timeout depending on transfer size, neither does it chop up large chunks, so it can absolutely happen that a huge burst write takes too long for the timeout to kick in.

[X512]

Replying to mmlr:

When you're in KDL can you run "syslog | tail 150" (or just syslog and page through to the end) and get the last few lines.

OK.

[X512]

syslog, other pages look like similar

[X512]

syslog, with diffrent errors

[mmlr]

It looks like the physical memory pool of USB gets exhausted. There might be a leak causing this, but it might also just be too much data being pushed through.

[mmlr]

I've made the PMA more robust in hrev43125. If it is just too many transfers then it should work now. If there's a transfer leak somewhere, then the problem should remain, though there should only be a single PMA message and it'd wait there forever.

[X512]

I tried to backup BFS partition to USB HDD(13GB). Coping ended correctly.

[mmlr]

Replying to X512:

I tried to backup BFS partition to USB HDD(13GB). Coping ended correctly.

Great, thanks for the feedback! Closing as fixed in hrev43125.