Linux-NFS Archive on lore.kernel.org
 help / color / Atom feed
* [PATCH 00/12] exposing knfsd state to userspace
@ 2019-05-16  1:20 J. Bruce Fields
  2019-05-16  1:20 ` [PATCH 1/2] nfsd: allow fh_want_write to be called twice J. Bruce Fields
                   ` (13 more replies)
  0 siblings, 14 replies; 17+ messages in thread
From: J. Bruce Fields @ 2019-05-16  1:20 UTC (permalink / raw)
  To: linux-nfs; +Cc: linux-fsdevel, J. Bruce Fields

From: "J. Bruce Fields" <bfields@redhat.com>

This is still a little rough, but maybe closer.  Changes since last
time:
	- renamed the "opens" file to "states" and added some (minimal)
	  information about lock, delegation, and layout stateids as
	  well as opens.
	- converted the states file to a YAML-like format.
	- added the ability to remove a client's state by writing
	  "expire\n" into a new nfsd/client/#/ctl file.

Recapping discussion from last time:

The following patches expose information about NFSv4 state held by knfsd
on behalf of NFSv4 clients.  This especially important for opens, which
are currently invisible to userspace on the server, unlike locks
(/proc/locks) and local processes' opens (under /proc/<pid>/).

The approach is to add a new directory /proc/fs/nfsd/clients/ with
subdirectories for each active NFSv4 client.  Each subdirectory has an
"info" file with some basic information to help identify the client and
a "states" directory that lists the open state held by that client.

Currently these pseudofiles look like:

  # find /proc/fs/nfsd/clients -type f|xargs tail 
  ==> /proc/fs/nfsd/clients/3/opens <==
 - 0x020000006a5fdc5c4ad09d9e01000000: { type: open, access: rw, deny: --, superblock: "fd:10:13649", owner: "open id:\x00\x00\x00&\x00\x00\x00\x00\x00\x0046��QH " }
 - 0x010000006a5fdc5c4ad09d9e03000000: { type: open, access: r-, deny: --, superblock: "fd:10:13650", owner: "open id:\x00\x00\x00&\x00\x00\x00\x00\x00\x0046��QH" }
 - 0x010000006a5fdc5c4ad09d9e04000000: { type: deleg, access: r, superblock: "fd:10:13650" }
 - 0x010000006a5fdc5c4ad09d9e06000000: { type: lock, superblock: "fd:10:13649", owner: "lock id:\x00\x00\x00&\x00\x00\x00\x00\x00\x00\x00\x00" }


  ==> /proc/fs/nfsd/clients/3/info <==
  clientid: 6debfb505cc0cd36
  address: 192.168.122.36:907
  name: "Linux NFSv4.2 test2.fieldses.org"
  minor version: 2

I'm conflicted about how I'm representing stateowners and client names,
both opaque byte-streams in the protocol but that often include
human-readable ascii.

Possibly also todo:
	- add some information about krb5 principals to the clients
	  file.
	- add information about the stateids used to represent
	  asynchronous copies.  They're a little different from the
	  other stateids and might end up in a separate "copies" file,
	- this duplicates some functionality of the little-used fault
	  injection code; could we replace it entirely?
	- some of the bits of filesystem code could probably be shared
	  with rpc_pipefs and libfs.

--b.

J. Bruce Fields (10):
  nfsd: persist nfsd filesystem across mounts
  nfsd: rename cl_refcount
  nfsd4: use reference count to free client
  nfsd: add nfsd/clients directory
  nfsd: make client/ directory names small ints
  rpc: replace rpc_filelist by tree_descr
  nfsd4: add a client info file
  nfsd4: add file to display list of client's opens
  nfsd: expose some more information about NFSv4 opens
  nfsd: add more information to client info file

 fs/nfsd/netns.h                |   6 +
 fs/nfsd/nfs4state.c            | 228 ++++++++++++++++++++++++++++++---
 fs/nfsd/nfsctl.c               | 225 +++++++++++++++++++++++++++++++-
 fs/nfsd/nfsd.h                 |  11 ++
 fs/nfsd/state.h                |   9 +-
 fs/seq_file.c                  |  17 +++
 include/linux/seq_file.h       |   2 +
 include/linux/string_helpers.h |   1 +
 lib/string_helpers.c           |   5 +-
 net/sunrpc/rpc_pipe.c          |  37 ++----
 10 files changed, 491 insertions(+), 50 deletions(-)

-- 
2.20.1


J. Bruce Fields (12):
  nfsd: persist nfsd filesystem across mounts
  nfsd: rename cl_refcount
  nfsd4: use reference count to free client
  nfsd: add nfsd/clients directory
  nfsd: make client/ directory names small ints
  nfsd4: add a client info file
  nfsd: copy client's address including port number to cl_addr
  nfsd: add more information to client info file
  nfsd4: add file to display list of client's opens
  nfsd: show lock and deleg stateids
  nfsd4: show layout stateids
  nfsd: allow forced expiration of NFSv4 clients

 fs/nfsd/netns.h          |   6 +
 fs/nfsd/nfs4state.c      | 418 ++++++++++++++++++++++++++++++++++++---
 fs/nfsd/nfsctl.c         | 225 ++++++++++++++++++++-
 fs/nfsd/nfsd.h           |  11 ++
 fs/nfsd/state.h          |   7 +-
 fs/seq_file.c            |  17 ++
 include/linux/seq_file.h |   2 +
 7 files changed, 661 insertions(+), 25 deletions(-)

-- 
2.21.0


^ permalink raw reply	[flat|nested] 17+ messages in thread
* [PATCH 0/2] nfsd shoudn't call mnt_want_write twice
@ 2019-05-13 15:27 J. Bruce Fields
  2019-05-13 15:27 ` [PATCH 2/2] nfsd: fh_drop_write in nfsd_unlink J. Bruce Fields
  0 siblings, 1 reply; 17+ messages in thread
From: J. Bruce Fields @ 2019-05-13 15:27 UTC (permalink / raw)
  To: linux-nfs; +Cc: J. Bruce Fields

From: "J. Bruce Fields" <bfields@redhat.com>

A fuzzer recently triggered lockdep warnings about potential sb_writers
deadlocks caused by fh_want_write, showing that we haven't been careful
to pair each fh_want_write() with an fh_drop_write().

This isn't normally a problem since fh_put() will call fh_drop_write()
for us.  And that's OK for NFSv3.  But an NFSv4 protocol fuzzer can do
weird things like call unlink twice in a compound.

So we can either make it safe to call fh_want_write() twice, or we can
fix all the callers to call fh_drop_write().

For now I think we have to do the former just to get the bug fixed.

Long term I don't know whether it's best to stay with that or to fix up
the callers.  I fixed nfsd_unlink just because it's easy, but maybe
that's pointless, it's others (like setattr) that are the complicated
ones.

J. Bruce Fields (2):
  nfsd: allow fh_want_write to be called twice
  nfsd: fh_drop_write in nfsd_unlink

 fs/nfsd/vfs.c | 8 +++++---
 fs/nfsd/vfs.h | 5 ++++-
 2 files changed, 9 insertions(+), 4 deletions(-)

-- 
2.21.0


^ permalink raw reply	[flat|nested] 17+ messages in thread

end of thread, back to index

Thread overview: 17+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2019-05-16  1:20 [PATCH 00/12] exposing knfsd state to userspace J. Bruce Fields
2019-05-16  1:20 ` [PATCH 1/2] nfsd: allow fh_want_write to be called twice J. Bruce Fields
2019-05-18 20:04   ` bfields
2019-05-16  1:20 ` [PATCH 01/12] nfsd: persist nfsd filesystem across mounts J. Bruce Fields
2019-05-16  1:20 ` [PATCH 2/2] nfsd: fh_drop_write in nfsd_unlink J. Bruce Fields
2019-05-16  1:20 ` [PATCH 02/12] nfsd: rename cl_refcount J. Bruce Fields
2019-05-16  1:20 ` [PATCH 03/12] nfsd4: use reference count to free client J. Bruce Fields
2019-05-16  1:20 ` [PATCH 04/12] nfsd: add nfsd/clients directory J. Bruce Fields
2019-05-16  1:20 ` [PATCH 05/12] nfsd: make client/ directory names small ints J. Bruce Fields
2019-05-16  1:20 ` [PATCH 06/12] nfsd4: add a client info file J. Bruce Fields
2019-05-16  1:20 ` [PATCH 07/12] nfsd: copy client's address including port number to cl_addr J. Bruce Fields
2019-05-16  1:20 ` [PATCH 08/12] nfsd: add more information to client info file J. Bruce Fields
2019-05-16  1:20 ` [PATCH 09/12] nfsd4: add file to display list of client's opens J. Bruce Fields
2019-05-16  1:20 ` [PATCH 10/12] nfsd: show lock and deleg stateids J. Bruce Fields
2019-05-16  1:20 ` [PATCH 11/12] nfsd4: show layout stateids J. Bruce Fields
2019-05-16  1:20 ` [PATCH 12/12] nfsd: allow forced expiration of NFSv4 clients J. Bruce Fields
  -- strict thread matches above, loose matches on Subject: below --
2019-05-13 15:27 [PATCH 0/2] nfsd shoudn't call mnt_want_write twice J. Bruce Fields
2019-05-13 15:27 ` [PATCH 2/2] nfsd: fh_drop_write in nfsd_unlink J. Bruce Fields

Linux-NFS Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/linux-nfs/0 linux-nfs/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 linux-nfs linux-nfs/ https://lore.kernel.org/linux-nfs \
		linux-nfs@vger.kernel.org linux-nfs@archiver.kernel.org
	public-inbox-index linux-nfs


Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.kernel.vger.linux-nfs


AGPL code for this site: git clone https://public-inbox.org/ public-inbox