Re: NFS client broken in 4.1.0-rc2

From: Russell King - ARM Linux
Date: Fri May 15 2015 - 10:50:36 EST

On Fri, May 15, 2015 at 10:26:48AM -0400, Chuck Lever wrote:
>
> On May 15, 2015, at 10:24 AM, Russell King - ARM Linux <linux@xxxxxxxxxxxxxxxx> wrote:
>
> > While trying to update a kernel and modules on one of my test systems,
> > I was greeted by these errors:
> >
> > tar: lib/modules/4.1.0-rc2+/kernel/drivers/media/platform/coda/coda.ko: Cannot utime
> > tar: lib/modules/4.1.0-rc2+/kernel/drivers/media/dvb-frontends/drx39xyj/drx39xyj.ko: Cannot utime
> > tar: lib/modules/4.1.0-rc2+/kernel/drivers/media/usb/em28xx/em28xx.ko: Cannot utime
> > tar: lib/modules/4.1.0-rc2+/kernel/drivers/usb/serial/option.ko: Cannot utime
> > tar: lib/modules/4.1.0-rc2+/kernel/drivers/usb/serial/ftdi_sio.ko: Cannot utime
> > tar: lib/modules/4.1.0-rc2+/kernel/drivers/net/wireless/brcm80211/brcmfmac/brcmfmac.ko: Cannot utime
> > tar: lib/modules/4.1.0-rc2+/kernel/drivers/input/mouse/psmouse.ko: Cannot utime
> > tar: lib/modules/4.1.0-rc2+/kernel/fs/udf/udf.ko: Cannot utime
> > tar: lib/modules/4.1.0-rc2+/kernel/fs/fuse/fuse.ko: Cannot utime
> > tar: lib/modules/4.1.0-rc2+/kernel/fs/nfsd/nfsd.ko: Cannot utime
> > tar: lib/modules/4.1.0-rc2+/kernel/sound/soc/codecs/snd-soc-wm8962.ko: Cannot utime
> > tar: lib/modules/4.1.0-rc2+/kernel/net/bluetooth/bluetooth.ko: Cannot utime
> > tar: lib/modules/4.1.0-rc2+/modules.alias.bin: Cannot utime
> > tar: lib/modules/4.1.0-rc2+/modules.alias: Cannot utime
> > tar: Exiting with failure status due to previous errors
> >
> > Searching google wasn't helpful, as all the "Cannot utime" errors that
> > google could find are followed by an errno string.
> >
> > stracing at first sight didn't seem to be helpful, as no syscalls (apart
> > from openat() with a pre-existing file) were failing.
> >
> > Having recently updated to fc21 tar generating the archive, I thought
> > maybe it was a tar format bug between fc21 tar and the target's tar.
> > That was until I tried to "apt-get source tar" on the target, and was
> > greeted by the same error.
> >
> > So I then tried untaring the tar source archive onto a ramfs, which
> > worked without complaint. The difference being that it's a root NFS
> > box, and so I was untaring onto NFS.
> >
> > Here's the entry from /proc/mounts:
> >
> > x.y.z.221:/var/boot/ci on / type nfs (rw,nolock,vers=4,addr=x.y.z.221,clientaddr=a.b.c.55)
> >
> > Looking closer at the strace reveals this:
> >
> > openat(AT_FDCWD, "lib/modules/4.1.0-rc2+/kernel/drivers/media/platform/coda/coda.ko", O_WRONLY|O_CREAT|O_EXCL|O_NOCTTY|O_NONBLOCK|O_LARGEFILE|O_CLOEXEC, 0600) = -1 EEXIST (File exists)
> > unlinkat(AT_FDCWD, "lib/modules/4.1.0-rc2+/kernel/drivers/media/platform/coda/coda.ko", 0) = 0
> > openat(AT_FDCWD, "lib/modules/4.1.0-rc2+/kernel/drivers/media/platform/coda/coda.ko", O_WRONLY|O_CREAT|O_EXCL|O_NOCTTY|O_NONBLOCK|O_LARGEFILE|O_CLOEXEC, 0600) = 4
> > write(4, "\177ELF\1\1\1\0\0\0\0\0\0\0\0\0\1\0(\0\1\0\0\0\0\0\0\0\0\0\0\0"..., 4096) = 4096
> > ...
> > write(4, "\300H\0\0\34\345\1\0\314H\0\0\34\345\1\0\330H\0\0\34\345\1\0<I\0\0\34\370\1\0"..., 7312) = 7312
> > dup2(4, 4) = 4
> > fstat64(4, {st_mode=0757221, st_size=13181880119170311768, ...}) = 21
> > write(2, "tar: ", 5) = 5
> > write(2, "lib/modules/4.1.0-rc2+/kernel/dr"..., 79) = 79
> > write(2, "\n", 1) = 1
> > fchown32(4, 0, 0) = 0
> > fchmod(4, 0664) = 0
> > close(4) = 0
> >
> > Look closely at that fstat64, and you'll notice that it's returning crap.
>
> This is likely fixed by:
>
> http://marc.info/?l=linux-nfs&m=143095122604344&w=2

Yes, it appears to be fixed by that change. I'll add it to my tree so
my test systems work.

Many thanks.

--
FTTC broadband for 0.8mile line: currently at 10.5Mbps down 400kbps up
according to speedtest.net.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/