From mboxrd@z Thu Jan 1 00:00:00 1970 From: Chuck Lever Subject: Re: NFS client broken in 4.1.0-rc2 Date: Wed, 20 May 2015 21:16:18 -0400 Message-ID: <4DD061F4-B1A8-4CBE-A817-ACC921661A89@oracle.com> References: <20150515142403.GL2067@n2100.arm.linux.org.uk> <20150515145013.GM2067@n2100.arm.linux.org.uk> <20150520232228.GC2067@n2100.arm.linux.org.uk> Mime-Version: 1.0 (Mac OS X Mail 7.3 \(1878.6\)) Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 8BIT Cc: Linux NFS Mailing List , Trond Myklebust , LKML Kernel , linux-fsdevel , Anna Schumaker , linux-arm-kernel-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r@public.gmane.org To: Russell King - ARM Linux Return-path: In-Reply-To: <20150520232228.GC2067-l+eeeJia6m9vn6HldHNs0ANdhmdF6hFW@public.gmane.org> Sender: linux-nfs-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org List-Id: linux-fsdevel.vger.kernel.org On May 20, 2015, at 7:22 PM, Russell King - ARM Linux 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 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>>> 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-nfs" in the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org More majordomo info at http://vger.kernel.org/majordomo-info.html