UBIFS: unlink failure

UBIFS: unlink failure

[Richard Weinberger]

Hi!

While testing UBI fastmap I encountered a problem.
After multiple writes unlink fails.

This is my test case:
---cut---
function do_mount_write1()
{
        ubiattach -m 0 /dev/ubi_ctrl
        mount /dev/ubi0_0 /mnt/ -t ubifs
        dd if=/dev/urandom of=/mnt/x bs=1M count=1
        dd if=/dev/urandom of=/mnt/x bs=1M count=2
        dd if=/dev/urandom of=/mnt/x bs=1M count=10
        dd if=/dev/urandom of=/mnt/x bs=1M count=30
        rm x
        umount /mnt
        ubidetach -m 0 /dev/ubi_ctrl
}

flash_eraseall /dev/mtd0

ubiattach -m 0 /dev/ubi_ctrl
ubimkvol -N test -s 100MiB /dev/ubi0
mkfs.ubifs /dev/ubi0_0
ubidetach -m 0 /dev/ubi_ctrl

do_mount_write1
---cut---

rm does newfstatat() and unlinkat(), both fail with ENOENT.
The issue is 100% reproducible on a current kernel (3.4)
Adding a "sleep 10" before the rm does _not_ help.

Thanks,
//richard

Re: UBIFS: unlink failure

[Artem Bityutskiy]

[-- Attachment #1: Type: text/plain, Size: 382 bytes --]

On Tue, 2012-06-12 at 23:56 +0200, Richard Weinberger wrote:
>         dd if=/dev/urandom of=/mnt/x bs=1M count=1
>         dd if=/dev/urandom of=/mnt/x bs=1M count=2
>         dd if=/dev/urandom of=/mnt/x bs=1M count=10
>         dd if=/dev/urandom of=/mnt/x bs=1M count=30
>         rm x

rm /mnt/x, not rm x. Then it works.

:-)

-- 
Best Regards,
Artem Bityutskiy

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 836 bytes --]

Re: UBIFS: unlink failure

[Richard Weinberger]

Am Tue, 26 Jun 2012 17:26:00 +0300
schrieb Artem Bityutskiy <dedekind1@gmail.com>:

> On Tue, 2012-06-12 at 23:56 +0200, Richard Weinberger wrote:
> >         dd if=/dev/urandom of=/mnt/x bs=1M count=1
> >         dd if=/dev/urandom of=/mnt/x bs=1M count=2
> >         dd if=/dev/urandom of=/mnt/x bs=1M count=10
> >         dd if=/dev/urandom of=/mnt/x bs=1M count=30
> >         rm x
> 
> rm /mnt/x, not rm x. Then it works.

Hmmmmmmmmmm, I quite sure this is just a typo in the mail.
Anyway, I'll retest :-D

Thanks,
//richard