Return-Path: Received: from pandora.arm.linux.org.uk ([78.32.30.218]:60922 "EHLO pandora.arm.linux.org.uk" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752759AbbETXWm (ORCPT ); Wed, 20 May 2015 19:22:42 -0400 Date: Thu, 21 May 2015 00:22:28 +0100 From: Russell King - ARM Linux To: Chuck Lever Cc: Linux NFS Mailing List , Trond Myklebust , linux-kernel@vger.kernel.org, linux-fsdevel@vger.kernel.org, Anna Schumaker , linux-arm-kernel@lists.infradead.org Subject: Re: NFS client broken in 4.1.0-rc2 Message-ID: <20150520232228.GC2067@n2100.arm.linux.org.uk> References: <20150515142403.GL2067@n2100.arm.linux.org.uk> <20150515145013.GM2067@n2100.arm.linux.org.uk> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii In-Reply-To: <20150515145013.GM2067@n2100.arm.linux.org.uk> Sender: linux-nfs-owner@vger.kernel.org List-ID: 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? -- FTTC broadband for 0.8mile line: currently at 10.5Mbps down 400kbps up according to speedtest.net.