NFS client broken in 4.1.0-rc2

[Russell King - ARM Linux]

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 at arm.linux.org.uk> 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.