Re: NFS client broken in 4.1.0-rc2

[Chuck Lever]

On May 20, 2015, at 7:22 PM, Russell King - ARM Linux <linux@xxxxxxxxxxxxxxxx> wrote:

> On Fri, May 15, 2015 at 03:50:14PM +0100, Russell King - ARM Linux wrote:
>> 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.
> 
> I notice NFS patches were merged by Linus recently, but this patch was not
> in the set.  Is this patch scheduled to go in for 4.1?

Trond sent Linus a pull request yesterday for 2 patches, including
the one you are asking about. See:

http://marc.info/?l=linux-nfs&m=143205093309624&w=2

--
Chuck Lever
chuck[dot]lever[at]oracle[dot]com



--
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/