Re: [BUG] cannot mount nfs, bisected to 0db10944a76ba ("nfs: Convert to separately allocated bdi")

From: Jan Kara <jack@suse.cz>
To: Corentin Labbe <clabbe.montjoie@gmail.com>
Cc: jack@suse.cz, anna.schumaker@netapp.com, hch@lst.de,
	trond.myklebust@primarydata.com, axboe@fb.com,
	linux-nfs@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [BUG] cannot mount nfs, bisected to 0db10944a76ba ("nfs: Convert to separately allocated bdi")
Date: Tue, 2 May 2017 18:27:04 +0200	[thread overview]
Message-ID: <20170502162704.GM13916@quack2.suse.cz> (raw)
In-Reply-To: <20170428095624.GA22965@Red>

[-- Attachment #1: Type: text/plain, Size: 3666 bytes --]

Hello,

On Fri 28-04-17 11:56:24, Corentin Labbe wrote:
> Since linux next-20170421, mounting nfs give me:
> [  774.994934] ------------[ cut here ]------------
> [  774.994975] WARNING: CPU: 1 PID: 10284 at /linux-next/fs/sysfs/dir.c:31 sysfs_warn_dup+0x64/0x74
> [  774.994985] sysfs: cannot create duplicate filename '/devices/virtual/bdi/0:32'
> [  774.994992] Modules linked in: axp20x_usb_power gpio_axp209 nvmem_sunxi_sid sun4i_dma sun4i_ss virt_dma
> [  774.995047] CPU: 1 PID: 10284 Comm: mount.nfs Not tainted 4.11.0-rc4+ #14
> [  774.995054] Hardware name: Allwinner sun7i (A20) Family
> [  774.995085] [<c010f19c>] (unwind_backtrace) from [<c010bc74>] (show_stack+0x10/0x14)
> [  774.995104] [<c010bc74>] (show_stack) from [<c03c6e24>] (dump_stack+0x78/0x8c)
> [  774.995121] [<c03c6e24>] (dump_stack) from [<c0122200>] (__warn+0xe8/0x100)
> [  774.995135] [<c0122200>] (__warn) from [<c0122250>] (warn_slowpath_fmt+0x38/0x48)
> [  774.995150] [<c0122250>] (warn_slowpath_fmt) from [<c02ac178>] (sysfs_warn_dup+0x64/0x74)
> [  774.995167] [<c02ac178>] (sysfs_warn_dup) from [<c02ac254>] (sysfs_create_dir_ns+0x84/0x94)
> [  774.995184] [<c02ac254>] (sysfs_create_dir_ns) from [<c03c8b8c>] (kobject_add_internal+0x9c/0x2ec)
> [  774.995199] [<c03c8b8c>] (kobject_add_internal) from [<c03c8e24>] (kobject_add+0x48/0x98)
> [  774.995217] [<c03c8e24>] (kobject_add) from [<c048d75c>] (device_add+0xe4/0x5a0)
> [  774.995232] [<c048d75c>] (device_add) from [<c048ddb4>] (device_create_groups_vargs+0xac/0xbc)
> [  774.995247] [<c048ddb4>] (device_create_groups_vargs) from [<c048dde4>] (device_create_vargs+0x20/0x28)
> [  774.995263] [<c048dde4>] (device_create_vargs) from [<c02075c8>] (bdi_register_va+0x44/0xfc)
> [  774.995280] [<c02075c8>] (bdi_register_va) from [<c023d378>] (super_setup_bdi_name+0x48/0xa4)
> [  774.995299] [<c023d378>] (super_setup_bdi_name) from [<c0312ef4>] (nfs_fill_super+0x1a4/0x204)
> [  774.995315] [<c0312ef4>] (nfs_fill_super) from [<c03133f0>] (nfs_fs_mount_common+0x140/0x1e8)
> [  774.995333] [<c03133f0>] (nfs_fs_mount_common) from [<c03335cc>] (nfs4_remote_mount+0x50/0x58)
> [  774.995349] [<c03335cc>] (nfs4_remote_mount) from [<c023ef98>] (mount_fs+0x14/0xa4)
> [  774.995368] [<c023ef98>] (mount_fs) from [<c025cba0>] (vfs_kern_mount+0x54/0x128)
> [  774.995385] [<c025cba0>] (vfs_kern_mount) from [<c033352c>] (nfs_do_root_mount+0x80/0xa0)
> [  774.995400] [<c033352c>] (nfs_do_root_mount) from [<c0333818>] (nfs4_try_mount+0x28/0x3c)
> [  774.995415] [<c0333818>] (nfs4_try_mount) from [<c0313874>] (nfs_fs_mount+0x2cc/0x8c4)
> [  774.995430] [<c0313874>] (nfs_fs_mount) from [<c023ef98>] (mount_fs+0x14/0xa4)
> [  774.995445] [<c023ef98>] (mount_fs) from [<c025cba0>] (vfs_kern_mount+0x54/0x128)
> [  774.995461] [<c025cba0>] (vfs_kern_mount) from [<c02600f0>] (do_mount+0x158/0xc7c)
> [  774.995475] [<c02600f0>] (do_mount) from [<c0260f98>] (SyS_mount+0x8c/0xb4)
> [  774.995491] [<c0260f98>] (SyS_mount) from [<c0107840>] (ret_fast_syscall+0x0/0x3c)
> [  774.995501] ---[ end trace 0665e451f8864ff0 ]---
...
> The mount command is
> mount -t nfs -o tcp,hard,intr,async,rsize=4096,wsize=4096 192.168.1.100:/mnt/local_kernel /usr/src/
> 
> the mount command failling with: "mount.nfs: Cannot allocate memory"

I've tried reproducing this (both with NFSv3 and NFSv4) and failed. Also I
have looked through the code and I fail to see how this could happen. Is
this the only NFS mount that you have on your system? Didn't also the
WARN_ON in super_setup_bdi_name() trigger?

Can you run with the attached debug patch and post full dmesg after the
failure?

								Honza
-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR

[-- Attachment #2: 0001-nfs-Debug-bdi-registration-failure.patch --]
[-- Type: text/x-patch, Size: 1888 bytes --]

>From 90f66349ce91d8dd4811eca5efffb605cf084193 Mon Sep 17 00:00:00 2001
From: Jan Kara <jack@suse.cz>
Date: Tue, 2 May 2017 18:24:57 +0200
Subject: [PATCH] nfs: Debug bdi registration failure

Debug failure and fixup error code propagation.

Signed-off-by: Jan Kara <jack@suse.cz>
---
 fs/nfs/super.c   | 5 ++++-
 mm/backing-dev.c | 2 ++
 2 files changed, 6 insertions(+), 1 deletion(-)

diff --git a/fs/nfs/super.c b/fs/nfs/super.c
index dc69314d455e..f60a387105ad 100644
--- a/fs/nfs/super.c
+++ b/fs/nfs/super.c
@@ -2344,6 +2344,8 @@ int nfs_fill_super(struct super_block *sb, struct nfs_mount_info *mount_info)
 
  	nfs_initialise_sb(sb);
 
+	printk("Registering bdi %u:%u for server %p\n", MAJOR(server->s_dev), MINOR(server->s_dev), server);
+	dump_stack();
 	ret = super_setup_bdi_name(sb, "%u:%u", MAJOR(server->s_dev),
 				   MINOR(server->s_dev));
 	if (ret)
@@ -2607,7 +2609,7 @@ struct dentry *nfs_fs_mount_common(struct nfs_server *server,
 		/* initial superblock/root creation */
 		error = mount_info->fill_super(s, mount_info);
 		if (error)
-			goto error_splat_super;
+			goto error_splat_super_err;
 		nfs_get_cache_cookie(s, mount_info->parsed, mount_info->cloned);
 	}
 
@@ -2630,6 +2632,7 @@ struct dentry *nfs_fs_mount_common(struct nfs_server *server,
 
 error_splat_root:
 	dput(mntroot);
+error_splat_super_err:
 	mntroot = ERR_PTR(error);
 error_splat_super:
 	deactivate_locked_super(s);
diff --git a/mm/backing-dev.c b/mm/backing-dev.c
index f028a9a472fd..ac173d06834a 100644
--- a/mm/backing-dev.c
+++ b/mm/backing-dev.c
@@ -923,6 +923,8 @@ static void bdi_remove_from_list(struct backing_dev_info *bdi)
 
 void bdi_unregister(struct backing_dev_info *bdi)
 {
+	printk("Unregistering bdi %s\n", dev_name(bdi->dev));
+	dump_stack();
 	/* make sure nobody finds us on the bdi_list anymore */
 	bdi_remove_from_list(bdi);
 	wb_shutdown(&bdi->wb);
-- 
2.12.0

