Re: 2.6.2-rc2 nfsd+xfs spins in i_size_read()

From: Miquel van Smoorenburg
Date: Fri Jan 30 2004 - 20:26:29 EST


On Sat, 31 Jan 2004 00:13:16, Andrew Morton wrote:
> Christoph Hellwig <hch@xxxxxxxxxxxxx> wrote:
> >
> > On Fri, Jan 30, 2004 at 02:34:59PM -0800, Andrew Morton wrote:
> > > If two CPUs hit i_size_write() at the same time we have a bug. That
> > > function requires that the caller provide external serialisation, via i_sem.
> >
> > O_APPEND|O_DIRECT writes could do that under XFS..
>
> Sigh.
>
> diff -puN mm/filemap.c~i_size_write-check mm/filemap.c
> --- 25/mm/filemap.c~i_size_write-check Fri Jan 30 15:10:23 2004
> +++ 25-akpm/mm/filemap.c Fri Jan 30 15:11:41 2004
> @@ -2010,3 +2010,18 @@ out:
> }
>
> EXPORT_SYMBOL_GPL(generic_file_direct_IO);
> +
> +void i_size_write_check(struct inode *inode)
> +{
> + static int count = 0;
> +
> + if (down_trylock(&inode->i_sem) == 0) {
> + if (count < 10) {

You want to set this to 100 at least, since at boot time the message
happens _often_ even without XFS.

It's caused by sysfs:

Jan 31 00:48:33 meghan kernel: i_size_write() called without i_sem
Jan 31 00:48:33 meghan kernel: Call Trace:
Jan 31 00:48:33 meghan kernel: [i_size_write_check+95/97] i_size_write_check+0x5f/0x61
Jan 31 00:48:33 meghan kernel: [simple_commit_write+74/148] simple_commit_write+0x4a/0x94
Jan 31 00:48:33 meghan kernel: [page_symlink+210/414] page_symlink+0xd2/0x19e
Jan 31 00:48:33 meghan kernel: [sysfs_symlink+107/126] sysfs_symlink+0x6b/0x7e
Jan 31 00:48:33 meghan kernel: [sysfs_create_link+313/317] sysfs_create_link+0x139/0x13d
Jan 31 00:48:33 meghan kernel: [bus_add_device+143/161] bus_add_device+0x8f/0xa1
[etc etc]

Also, bd_set_size runs unlocked:

Jan 31 02:02:39 meghan kernel: I am buggy
Jan 31 02:02:39 meghan kernel: Call Trace:
Jan 31 02:02:39 meghan kernel: [bd_set_size+201/218] bd_set_size+0xc9/0xda
Jan 31 02:02:39 meghan kernel: [do_open+282/1014] do_open+0x11a/0x3f6
Jan 31 02:02:39 meghan kernel: [blkdev_get+114/128] blkdev_get+0x72/0x80
Jan 31 02:02:39 meghan kernel: [do_open+603/1014] do_open+0x25b/0x3f6
Jan 31 02:02:39 meghan kernel: [blkdev_get+114/128] blkdev_get+0x72/0x80
Jan 31 02:02:39 meghan kernel: [open_bdev_excl+87/172] open_bdev_excl+0x57/0xacJan 31 02:02:39 meghan kernel: [get_sb_bdev+55/344] get_sb_bdev+0x37/0x158
Jan 31 02:02:39 meghan kernel: [ext3_get_sb+47/51] ext3_get_sb+0x2f/0x33
Jan 31 02:02:39 meghan kernel: [ext3_fill_super+0/2949] ext3_fill_super+0x0/0xb85
Jan 31 02:02:39 meghan kernel: [do_kern_mount+95/213] do_kern_mount+0x5f/0xd5
Jan 31 02:02:39 meghan kernel: [do_add_mount+120/334] do_add_mount+0x78/0x14e
Jan 31 02:02:39 meghan kernel: [do_mount+289/359] do_mount+0x121/0x167
Jan 31 02:02:39 meghan kernel: [__copy_from_user_ll+104/108] __copy_from_user_ll+0x68/0x6c
Jan 31 02:02:39 meghan kernel: [copy_mount_options+128/234] copy_mount_options+0x80/0xea
Jan 31 02:02:39 meghan kernel: [sys_mount+203/309] sys_mount+0xcb/0x135
Jan 31 02:02:39 meghan kernel: [do_mount_root+47/156] do_mount_root+0x2f/0x9c
Jan 31 02:02:39 meghan kernel: [mount_block_root+84/285] mount_block_root+0x54/0x11d
Jan 31 02:02:39 meghan kernel: [mount_root+94/102] mount_root+0x5e/0x66
Jan 31 02:02:39 meghan kernel: [prepare_namespace+38/195] prepare_namespace+0x26/0xc3
Jan 31 02:02:39 meghan kernel: [init+80/334] init+0x50/0x14e
Jan 31 02:02:39 meghan kernel: [init+0/334] init+0x0/0x14e
Jan 31 02:02:39 meghan kernel: [kernel_thread_helper+5/11] kernel_thread_helper+0x5/0xb

But the XFS problem appears to be vn_revalidate which calls i_size_write()
without holding i_sem:

Jan 31 02:04:00 meghan kernel: I am buggy
Jan 31 02:04:00 meghan kernel: Call Trace:
Jan 31 02:04:00 meghan kernel: [__crc_pm_idle+1813107/4617778] vn_revalidate+0x18e/0x1a9 [xfs]
Jan 31 02:04:00 meghan kernel: [nfsd_commit+205/219] nfsd_commit+0xcd/0xdb
Jan 31 02:04:00 meghan kernel: [__crc_pm_idle+1798087/4617778] linvfs_getattr+0x39/0x3f [xfs]
Jan 31 02:04:00 meghan kernel: [vfs_getattr+57/154] vfs_getattr+0x39/0x9a
Jan 31 02:04:00 meghan kernel: [encode_post_op_attr+104/577] encode_post_op_attr+0x68/0x241
Jan 31 02:04:00 meghan kernel: [nfs3svc_encode_commitres+42/115] nfs3svc_encode_commitres+0x2a/0x73
Jan 31 02:04:00 meghan kernel: [nfsd_dispatch+297/485] nfsd_dispatch+0x129/0x1e5
Jan 31 02:04:00 meghan kernel: [svc_process+1170/1621] svc_process+0x492/0x655
Jan 31 02:04:00 meghan kernel: [apic_timer_interrupt+26/32] apic_timer_interrupt+0x1a/0x20
Jan 31 02:04:00 meghan kernel: [nfsd+490/915] nfsd+0x1ea/0x393
Jan 31 02:04:00 meghan kernel: [nfsd+0/915] nfsd+0x0/0x393
Jan 31 02:04:00 meghan kernel: [kernel_thread_helper+5/11] kernel_thread_helper+0x5/0xb

Mike.
-
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/