openwrt/staging/blogic.git
12 years agoSUNRPC: parametrize rpc_uaddr2sockaddr() by network context
Stanislav Kinsbursky [Fri, 13 Jan 2012 09:09:27 +0000 (13:09 +0400)]
SUNRPC: parametrize rpc_uaddr2sockaddr() by network context

Parametrize rpc_uaddr2sockaddr() by network context and thus force it's callers to pass
in network context instead of using hard-coded "init_net".

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoSUNRPC: parametrize rpc_pton() by network context
Stanislav Kinsbursky [Fri, 13 Jan 2012 09:09:19 +0000 (13:09 +0400)]
SUNRPC: parametrize rpc_pton() by network context

Parametrize rpc_pton() by network context and thus force it's callers to pass
in network context instead of using hard-coded "init_net".

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoSUNRPC: parametrize rpc_pton6() by network context
Stanislav Kinsbursky [Fri, 13 Jan 2012 09:09:11 +0000 (13:09 +0400)]
SUNRPC: parametrize rpc_pton6() by network context

Parametrize rpc_pton6() by network context and thus force it's caller
to pass in network context instead of using hard-coded "init_net".

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoSUNRPC: parametrize rpc_parse_scope_id() by network context
Stanislav Kinsbursky [Fri, 13 Jan 2012 09:09:03 +0000 (13:09 +0400)]
SUNRPC: parametrize rpc_parse_scope_id() by network context

Parametrize rpc_parse_scope_id() by network context and thus force it's caller
to pass in network context instead of using hard-coded "init_net".

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoSUNRPC: parametrize local rpcbind clients creation with net ns
Stanislav Kinsbursky [Fri, 13 Jan 2012 08:52:51 +0000 (12:52 +0400)]
SUNRPC: parametrize local rpcbind clients creation with net ns

These client are per network namespace and thus can be created for different
network namespaces.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoSUNRPC: register rpcbind programs in passed network namespase context
Stanislav Kinsbursky [Fri, 13 Jan 2012 08:52:43 +0000 (12:52 +0400)]
SUNRPC: register rpcbind programs in passed network namespase context

Registering rpcbind program requires rpcbind clients, which are per network
namespace context.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoSUNRPC: create rpcbind client in passed network namespace context
Stanislav Kinsbursky [Fri, 13 Jan 2012 08:52:35 +0000 (12:52 +0400)]
SUNRPC: create rpcbind client in passed network namespace context

Rpcbind clients are per network namespace.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoSUNRPC: optimize net_ns dereferencing in rpcbind registering calls
Stanislav Kinsbursky [Fri, 13 Jan 2012 08:52:26 +0000 (12:52 +0400)]
SUNRPC: optimize net_ns dereferencing in rpcbind registering calls

Static rpcbind registering functions can be parametrized by network namespace
pointer, calculated only once, instead of using init_net pointer (or taking it
from current when virtualization will be comleted) in many places.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoSUNRPC: optimize net_ns dereferencing in rpcbind creation calls
Stanislav Kinsbursky [Fri, 13 Jan 2012 08:52:18 +0000 (12:52 +0400)]
SUNRPC: optimize net_ns dereferencing in rpcbind creation calls

Static rpcbind creation functions can be parametrized by network namespace
pointer, calculated only once, instead of using init_net pointer (or taking it
from current when virtualization will be comleted) in many places.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoSUNRPC: move rpcbind internals to sunrpc part of network namespace context
Stanislav Kinsbursky [Fri, 13 Jan 2012 08:52:10 +0000 (12:52 +0400)]
SUNRPC: move rpcbind internals to sunrpc part of network namespace context

This patch makes rpcbind logic works in network namespace context. IOW each
network namespace will have it's own unique rpcbind internals (clients and
friends) required for registering svc services per network namespace.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoSUNRPC: Fix potential races in xprt_lock_write_next()
Trond Myklebust [Wed, 18 Jan 2012 03:57:37 +0000 (22:57 -0500)]
SUNRPC: Fix potential races in xprt_lock_write_next()

We have to ensure that the wake up from the waitqueue and the assignment
of xprt->snd_task are atomic. We can do this by assigning the snd_task
while under the waitqueue spinlock.

Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoNFS: Ensure that mmapped pages remain stable during writeback
Trond Myklebust [Wed, 18 Jan 2012 03:04:26 +0000 (22:04 -0500)]
NFS: Ensure that mmapped pages remain stable during writeback

Ensure that nfs_vm_page_mkwrite() waits for the page writeback to
complete before the application is allowed to modify page
contents.
The main reason for wanting to do this in NFS is to ensure that the
server doesn't get confused if we have to resend the RPC request
due to a dropped/missed reply.

Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoNFS: Optimise away unnecessary setattrs for open(O_TRUNC);
Trond Myklebust [Wed, 18 Jan 2012 03:04:26 +0000 (22:04 -0500)]
NFS: Optimise away unnecessary setattrs for open(O_TRUNC);

Currently, we will correctly optimise away a truncate that doesn't
change the file size. However, in the case of open(O_TRUNC), we
also want to optimise away the time changes.

Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoNFS: Move struct nfs_unique_id into struct nfs_seqid_counter
Trond Myklebust [Wed, 18 Jan 2012 03:04:25 +0000 (22:04 -0500)]
NFS: Move struct nfs_unique_id into struct nfs_seqid_counter

Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoNFSv4: Move contents of struct rpc_sequence into struct nfs_seqid_counter
Trond Myklebust [Wed, 18 Jan 2012 03:04:25 +0000 (22:04 -0500)]
NFSv4: Move contents of struct rpc_sequence into struct nfs_seqid_counter

Clean up.

Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoNFSv41: Add a new helper nfs4_init_sequence()
Trond Myklebust [Wed, 18 Jan 2012 03:04:25 +0000 (22:04 -0500)]
NFSv41: Add a new helper nfs4_init_sequence()

Clean up

Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoNFSv4: Replace lock_owner->ld_id with an ida based allocator
Trond Myklebust [Wed, 18 Jan 2012 03:04:25 +0000 (22:04 -0500)]
NFSv4: Replace lock_owner->ld_id with an ida based allocator

Again, We're unlikely to ever need more than 2^31 simultaneous lock
owners, so let's replace the custom allocator.

Now that there are no more users, we can also get rid of the custom
allocator code.

Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoNFSv4: Replace state_owner->so_owner_id with an ida based allocator
Trond Myklebust [Wed, 18 Jan 2012 03:04:24 +0000 (22:04 -0500)]
NFSv4: Replace state_owner->so_owner_id with an ida based allocator

We're unlikely to ever need more than 2^31 simultaneous open owners,
so let's replace the custom allocator with the generic ida allocator.

Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoNFSv4: Clean up nfs4_get_state_owner
Trond Myklebust [Wed, 18 Jan 2012 03:04:24 +0000 (22:04 -0500)]
NFSv4: Clean up nfs4_get_state_owner

Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoNFS: Remove unnecessary includes from linux/nfs_fs_i.h
Trond Myklebust [Wed, 18 Jan 2012 03:04:24 +0000 (22:04 -0500)]
NFS: Remove unnecessary includes from linux/nfs_fs_i.h

Also from linux/nfs_xdr.h.

Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoSUNRPC: kernel PipeFS mount point creation routines removed
Stanislav Kinsbursky [Tue, 10 Jan 2012 13:04:48 +0000 (17:04 +0400)]
SUNRPC: kernel PipeFS mount point creation routines removed

This patch removes static rpc_mnt variable and its creation and destruction
routines, because they are not used anymore.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoNFS: remove RPC PipeFS mount point reference from blocklayout routines
Stanislav Kinsbursky [Tue, 10 Jan 2012 13:04:40 +0000 (17:04 +0400)]
NFS: remove RPC PipeFS mount point reference from blocklayout routines

This is a cleanup patch. We don't need this reference anymore, because
blocklayout pipes dentries now creates and destroys in per-net operations and
on PipeFS mount/umount notification.
Note that nfs4blocklayout_register_net() now returns 0 instead of -ENOENT in
case of PipeFS superblock absence. This is ok, because blocklayout pipe dentry
will be created on PipeFS mount event.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoNFS: blocklayout PipeFS notifier introduced
Stanislav Kinsbursky [Tue, 10 Jan 2012 13:04:32 +0000 (17:04 +0400)]
NFS: blocklayout PipeFS notifier introduced

This patch subscribes blocklayout pipes to RPC pipefs notifications. Notifier
is registering on blocklayout module load. This notifier callback is
responsible for creation/destruction of PipeFS blocklayout pipe dentry.
Note that no locking required in notifier callback because PipeFS superblock
pointer is passed as an argument from it's creation or destruction routine and
thus we can be sure about it's validity.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoNFS: blocklayout pipe creation per network namespace context introduced
Stanislav Kinsbursky [Tue, 10 Jan 2012 13:04:24 +0000 (17:04 +0400)]
NFS: blocklayout pipe creation per network namespace context introduced

This patch implements blocklayout pipe creation and registration per each
existent network namespace.
This was achived by registering NFS per-net operations, responsible for
blocklayout pipe allocation/register and unregister/destruction instead of
initialization and destruction of static "bl_device_pipe" pipe (this one was
removed).
Note, than pointer to network blocklayout pipe is stored in per-net "nfs_net"
structure, because allocating of one more per-net structure for blocklayout
module looks redundant.
This patch also changes dev_remove() function prototype (and all it's callers,
where it' requied) by adding network namespace pointer parameter, which is used
to discover proper blocklayout pipe for rpc_queue_upcall() call.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoNFS: handle blocklayout pipe PipeFS dentry by network namespace aware routines
Stanislav Kinsbursky [Tue, 10 Jan 2012 13:04:16 +0000 (17:04 +0400)]
NFS: handle blocklayout pipe PipeFS dentry by network namespace aware routines

This patch makes blocklayout pipe dentry allocated and destroyed in network
namespace context by PipeFS network namespace aware routines.
Network namespace context is obtained from nfs_client structure.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoNFS: idmap PipeFS notifier introduced
Stanislav Kinsbursky [Tue, 10 Jan 2012 12:13:19 +0000 (16:13 +0400)]
NFS: idmap PipeFS notifier introduced

v2:
1) Added "nfs_idmap_init" and "nfs_idmap_quit" definitions for kernels built
without CONFIG_NFS_V4 option set.

This patch subscribes NFS clients to RPC pipefs notifications. Idmap notifier
is registering on NFS module load. This notifier callback is responsible for
creation/destruction of PipeFS idmap pipe dentry for NFS4 clients.

Since ipdmap pipe is created in rpc client pipefs directory, we have make sure,
that this directory has been created already. IOW RPC client notifier callback
has been called already. To achive this, PipeFS notifier priorities has been
introduced (RPC clients notifier priority is greater than NFS idmap one).
But this approach gives another problem: unlink for RPC client directory will
be called before NFS idmap pipe unlink on UMOUNT event and will fail, because
directory is not empty.
The solution, introduced in this patch, is to try to remove client directory
once again after idmap pipe was unlinked. This looks like ugly hack, so
probably it should be replaced in some more elegant way.

Note that no locking required in notifier callback because PipeFS superblock
pointer is passed as an argument from it's creation or destruction routine and
thus we can be sure about it's validity.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoNFS: handle NFS idmap pipe PipeFS dentries by network namespace aware routines
Stanislav Kinsbursky [Tue, 10 Jan 2012 12:13:11 +0000 (16:13 +0400)]
NFS: handle NFS idmap pipe PipeFS dentries by network namespace aware routines

This patch makes NFS idmap pipes dentries allocated and destroyed in network
namespace context by PipeFS network namespace aware routines.
Network namespace context is obtained from nfs_client structure.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoNFS: create callback transports in parent transport network namespace
Stanislav Kinsbursky [Tue, 10 Jan 2012 12:13:03 +0000 (16:13 +0400)]
NFS: create callback transports in parent transport network namespace

This patch replaces static "init_net" references with parent transport xprt_net
reference. Thus callback transports will be created in the same network
namespace as respective NFS mount point was created.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoNFS: pass NFS client owner network namespace to RPC client creation routine
Stanislav Kinsbursky [Tue, 10 Jan 2012 12:12:54 +0000 (16:12 +0400)]
NFS: pass NFS client owner network namespace to RPC client creation routine

This patch replaces static "init_net" with nfs_client->net pointer in RPC
client creation calls.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoNFS: make NFS client allocated per network namespace context
Stanislav Kinsbursky [Tue, 10 Jan 2012 12:12:46 +0000 (16:12 +0400)]
NFS: make NFS client allocated per network namespace context

This patch adds new net variable to nfs_client structure. This variable is set
on NFS client creation and cheched during matching NFS client search.
Initially current->nsproxy->net_ns is used as network namespace owner for new
NFS client to create. This network namespace pointer is set during mount
options parsing and thus can be passed from user-spave utils in future if will
be necessary.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoSUNRPC: fix pipe->ops cleanup on pipe dentry unlink
Stanislav Kinsbursky [Tue, 10 Jan 2012 12:12:38 +0000 (16:12 +0400)]
SUNRPC: fix pipe->ops cleanup on pipe dentry unlink

This patch looks late due to GSS AUTH patches sent already. But it fixes a flaw
in RPC PipeFS pipes handling.
I've added this patch in the series, because this series related to pipes. But
it should be a part of previous series named "SUNPRC: cleanup PipeFS for
network-namespace-aware users".

Pipe dentry can be created and destroyed many times during pipe life cycle.
This actually means, that we can't set pipe->ops to NULL in rpc_close_pipes()
and use this variable as a flag, indicating, that pipe's dentry is unlinking.
To follow this restriction, this patch replaces "pipe->ops = NULL" assignment
and checks for NULL with "pipe->dentry = NULL" assignment and checks for
NULL respectively.
This patch also removes check for non-NULL pipe->ops (or pipe->dentry) in
rpc_close_pipes() because it always non-NULL now.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoNFS: remove RPC PipeFS mount point references from NFS cache routines
Stanislav Kinsbursky [Fri, 25 Nov 2011 14:13:20 +0000 (17:13 +0300)]
NFS: remove RPC PipeFS mount point references from NFS cache routines

This is a cleanup patch. We don't need this reference anymore, because DNS
resolver cache now creates it's dentries in per-net operations and on PipeFS
mount/umount notification.
Note that nfs_cache_register_net() now returns 0 instead of -ENOENT in case of
PiepFS superblock absence. This is ok, Dns resolver cache will be regestered on
PipeFS mount event.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoNFS: DNS resolver PipeFS notifier introduced
Stanislav Kinsbursky [Fri, 25 Nov 2011 14:13:12 +0000 (17:13 +0300)]
NFS: DNS resolver PipeFS notifier introduced

This patch subscribes DNS resolver caches to RPC pipefs notifications. Notifier
is registering on NFS module load. This notifier callback is responsible for
creation/destruction of PipeFS DNS resolver cache directory.
Note that no locking required in notifier callback because PipeFS superblock
pointer is passed as an argument from it's creation or destruction routine and
thus we can be sure about it's validity.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoNFS: DNS resolver cache per network namespace context introduced
Stanislav Kinsbursky [Fri, 25 Nov 2011 14:13:04 +0000 (17:13 +0300)]
NFS: DNS resolver cache per network namespace context introduced

This patch implements DNS resolver cache creation and registration for each
alive network namespace context.
This was done by registering NFS per-net operations, responsible for DNS cache
allocation/register and unregister/destructioning instead of initialization and
destruction of static "nfs_dns_resolve" cache detail (this one was removed).
Pointer to network dns resolver cache is stored in new per-net "nfs_net"
structure.
This patch also changes nfs_dns_resolve_name() function prototype (and it's
calls) by adding network pointer parameter, which is used to get proper DNS
resolver cache pointer for do_cache_lookup_wait() call.

Note: empty nfs_dns_resolver_init() and nfs_dns_resolver_destroy() functions
will be used in next patch in the series.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoNFS: handle NFS caches dentries by network namespace aware routines
Stanislav Kinsbursky [Fri, 25 Nov 2011 14:12:56 +0000 (17:12 +0300)]
NFS: handle NFS caches dentries by network namespace aware routines

This patch makes NFS caches PipeFS dentries allocated and destroyed in network
namespace context by PipeFS network namespace aware routines.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoNFS: split cache creation and PipeFS registration
Stanislav Kinsbursky [Fri, 25 Nov 2011 14:12:48 +0000 (17:12 +0300)]
NFS: split cache creation and PipeFS registration

This precursor patch splits NFS cache creation and PipeFS registartion.
It's required for latter split of NFS DNS resolver cache creation per network
namespace context and PipeFS registration/unregistration on MOUNT/UMOUNT
events.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoSUNRPC: split cache creation and PipeFS registration
Stanislav Kinsbursky [Fri, 25 Nov 2011 14:12:40 +0000 (17:12 +0300)]
SUNRPC: split cache creation and PipeFS registration

This precursor patch splits SUNRPC cache creation and PipeFS registartion.
It's required for latter split of NFS DNS resolver cache creation per network
namespace context and PipeFS registration/unregistration on MOUNT/UMOUNT
events.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoSUNRPC: remove RPC PipeFS mount point reference from RPC client
Stanislav Kinsbursky [Wed, 11 Jan 2012 15:18:42 +0000 (19:18 +0400)]
SUNRPC: remove RPC PipeFS mount point reference from RPC client

This is a cleanup patch. We don't need this reference anymore.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoSUNRPC: remove RPC pipefs mount point manipulations from RPC clients code
Stanislav Kinsbursky [Wed, 11 Jan 2012 15:18:34 +0000 (19:18 +0400)]
SUNRPC: remove RPC pipefs mount point manipulations from RPC clients code

v2:
1) Updated due to changes in the first patch of the series.

Now, with RPC pipefs mount notifications handling in RPC clients, we can remove
mount point creation and destruction. RPC clients dentries will be created on
PipeFS mount event and removed on umount event.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoSUNRPC: remove RPC client pipefs dentries after unregister
Stanislav Kinsbursky [Wed, 11 Jan 2012 15:18:26 +0000 (19:18 +0400)]
SUNRPC: remove RPC client pipefs dentries after unregister

Without this patch we have races:

rpc_fill_super rpc_free_client
rpc_pipefs_event(MOUNT) rpc_remove_pipedir
spin_lock(&rpc_client_lock);
rpc_setup_pipedir_sb
spin_unlock(&rpc_client_lock);
spin_lock(&rpc_client_lock);
(remove from list)
spin_unlock(&rpc_client_lock);
MEAMORY LEAKED

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoSUNRPC: subscribe RPC clients to pipefs notifications
Stanislav Kinsbursky [Wed, 11 Jan 2012 15:18:17 +0000 (19:18 +0400)]
SUNRPC: subscribe RPC clients to pipefs notifications

This patch subscribes RPC clients to RPC pipefs notifications. RPC clients
notifier block is registering with pipefs initialization during SUNRPC module
init.
This notifier callback is responsible for RPC client PipeFS directory and GSS
pipes creation. For pipes creation and destruction two additional callbacks
were added to struct rpc_authops.
Note that no locking required in notifier callback because PipeFS superblock
pointer is passed as an argument from it's creation or destruction routine and
thus we can be sure about it's validity.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoSUNRPC: make SUNPRC clients list per network namespace context
Stanislav Kinsbursky [Thu, 12 Jan 2012 18:07:51 +0000 (22:07 +0400)]
SUNRPC: make SUNPRC clients list per network namespace context

This patch moves static SUNRPC clients list and it's lock to sunrpc_net
structure.
Currently this list is used only for debug purposes. But later it will be used
also for selecting clients by networks namespace on PipeFS mount/umount events.
Per-network namespace lists will make this faster and simplier.

Note: client list is taken from "init_net" network namespace context in
rpc_show_tasks(). This will be changed some day later with making SUNRPC
sysctl's per network namespace context.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoSUNRPC: handle GSS AUTH pipes by network namespace aware routines
Stanislav Kinsbursky [Wed, 11 Jan 2012 15:18:09 +0000 (19:18 +0400)]
SUNRPC: handle GSS AUTH pipes by network namespace aware routines

This patch makes RPC GSS PipeFs pipes allocated in it's RPC client owner
network namespace context.
Pipes creation and destruction now done in separated functions, which takes
care about PipeFS superblock locking.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoSUNRPC: handle RPC client pipefs dentries by network namespace aware routines
Stanislav Kinsbursky [Wed, 11 Jan 2012 15:18:01 +0000 (19:18 +0400)]
SUNRPC: handle RPC client pipefs dentries by network namespace aware routines

v2:
1) "Over-put" of PipeFS mount point fixed. Fix is ugly, but allows to bisect
the patch set. And it will be removed later in the series.

This patch makes RPC clients PipeFs dentries allocations in it's owner network
namespace context.
RPC client pipefs dentries creation logic has been changed:
1) Pipefs dentries creation by sb was moved to separated function, which will
be used for handling PipeFS mount notification.
2) Initial value of RPC client PipeFS dir dentry is set no NULL now.

RPC client pipefs dentries cleanup logic has been changed:
1) Cleanup is done now in separated rpc_remove_pipedir() function, which takes
care about pipefs superblock locking.

Also this patch removes slashes from cb_program.pipe_dir_name and from
NFS_PIPE_DIRNAME to make rpc_d_lookup_sb() work. This doesn't affect
vfs_path_lookup() results in nfs4blocklayout_init() since this slash is cutted
off anyway in link_path_walk().

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoSUNRPC: split SUNPRC PipeFS dentry and private pipe data creation
Stanislav Kinsbursky [Mon, 26 Dec 2011 12:44:06 +0000 (15:44 +0300)]
SUNRPC: split SUNPRC PipeFS dentry and private pipe data creation

This patch is a final step towards to removing PipeFS inode references from
kernel code other than PipeFS itself. It makes all kernel SUNRPC PipeFS users
depends on pipe private data, which state depend on their specific operations,
etc.
This patch completes SUNRPC PipeFS preparations and allows to create pipe
private data and PipeFS dentries independently.
Next step will be making SUNPRC PipeFS dentries allocated by SUNRPC PipeFS
network namespace aware routines.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoSUNRPC: cleanup GSS pipes usage
Stanislav Kinsbursky [Mon, 26 Dec 2011 12:43:57 +0000 (15:43 +0300)]
SUNRPC: cleanup GSS pipes usage

Currently gss auth holds RPC inode pointer which is now redundant since it
requires only pipes operations which takes private pipe data as an argument.
Thus this code can be cleaned and all references to RPC inode can be replaced
with privtae pipe data references.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoSUNPRC: cleanup RPC PipeFS pipes upcall interface
Stanislav Kinsbursky [Mon, 26 Dec 2011 12:43:49 +0000 (15:43 +0300)]
SUNPRC: cleanup RPC PipeFS pipes upcall interface

RPC pipe upcall doesn't requires only private pipe data. Thus RPC inode
references in this code can be removed.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoSUNRPC: cleanup PipeFS redundant RPC inode usage
Stanislav Kinsbursky [Mon, 26 Dec 2011 12:43:41 +0000 (15:43 +0300)]
SUNRPC: cleanup PipeFS redundant RPC inode usage

This patch removes redundant RPC inode references from PipeFS. These places are
actually where pipes operations are performed.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoSUNRPC: split SUNPRC PipeFS pipe data and inode creation
Stanislav Kinsbursky [Mon, 26 Dec 2011 12:43:32 +0000 (15:43 +0300)]
SUNRPC: split SUNPRC PipeFS pipe data and inode creation

Generally, pipe data is used only for pipes, and thus allocating space for it
on every RPC inode allocation is redundant. This patch splits private SUNRPC
PipeFS pipe data and inode, makes pipe data allocated only for pipe inodes.
This patch is also is a next step towards to to removing PipeFS inode
references from kernel code other than PipeFS itself.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoSUNRPC: replace inode lock with pipe lock for RPC PipeFS operations
Stanislav Kinsbursky [Mon, 26 Dec 2011 12:43:23 +0000 (15:43 +0300)]
SUNRPC: replace inode lock with pipe lock for RPC PipeFS operations

Currenly, inode i_lock is used to provide concurrent access to SUNPRC PipeFS
pipes. It looks redundant, since now other use of inode is present in most of
these places and thus can be easely replaced, which will allow to remove most
of inode references from PipeFS code. This is a first step towards to removing
PipeFS inode references from kernel code other than PipeFS itself.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoSUNRPC: added debug messages to RPC pipefs
Stanislav Kinsbursky [Mon, 26 Dec 2011 12:39:47 +0000 (15:39 +0300)]
SUNRPC: added debug messages to RPC pipefs

This patch adds debug messages for notification events.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoSUNRPC: pipefs per-net operations helper introduced
Stanislav Kinsbursky [Mon, 26 Dec 2011 12:39:39 +0000 (15:39 +0300)]
SUNRPC: pipefs per-net operations helper introduced

During per-net pipes creation and destruction we have to make sure, that pipefs
sb exists for the whole creation/destruction cycle. This is done by using
special mutex which controls pipefs sb reference on network namespace context.
Helper consists of two parts: first of them (rpc_get_dentry_net) searches for
dentry with specified name and returns with mutex taken on success. When pipe
creation or destructions is completed, caller should release this mutex by
rpc_put_dentry_net call.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoSUNRPC: put pipefs superblock link on network namespace
Stanislav Kinsbursky [Mon, 26 Dec 2011 12:39:30 +0000 (15:39 +0300)]
SUNRPC: put pipefs superblock link on network namespace

We have modules (like, pNFS blocklayout module) which creates pipes on
rpc_pipefs. Thus we need per-net operations for them. To make it possible we
require appropriate super block. So we have to put sb link on network namespace
context. Note, that it's not strongly required to create pipes in per-net
operations. IOW, if pipefs wasn't mounted yet, that no sb link reference will
present on network namespace and in this case we need just need to pass through
pipe creation. Pipe dentry will be created during pipefs mount notification.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoSUNRPC: pipefs dentry lookup helper introduced
Stanislav Kinsbursky [Mon, 26 Dec 2011 12:39:22 +0000 (15:39 +0300)]
SUNRPC: pipefs dentry lookup helper introduced

In all places, where pipefs dentries are created, only directory inode is
actually required to create new dentry. And all this directories has root
pipefs dentry as their parent. So we actually don't need this pipefs mount
point at all if some pipefs lookup method will be provided.
IOW, all we really need is just superblock and simple lookup method to find
root's child dentry with appropriate name. And this patch introduces this
method.
Note, that no locking implemented in rpc_d_lookup_sb(). So it can be used only
in case of assurance, that pipefs superblock still exist. IOW, we can use this
method only in pipefs mount-umount notification subscribers callbacks.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoSUNRPC: send notification events on pipefs sb creation and destruction
Stanislav Kinsbursky [Mon, 26 Dec 2011 12:39:13 +0000 (15:39 +0300)]
SUNRPC: send notification events on pipefs sb creation and destruction

They will be used to notify subscribers about pipefs superblock creation and
destruction.
Subcribers will have to create their dentries on passed superblock on mount
event and destroy otherwise.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoSUNRPC: hold current network namespace while pipefs superblock is active
Stanislav Kinsbursky [Mon, 26 Dec 2011 12:39:04 +0000 (15:39 +0300)]
SUNRPC: hold current network namespace while pipefs superblock is active

We want to be sure that network namespace is still alive while we have pipefs
mounted.
This will be required later, when RPC pipefs will be mounting only from
user-space context.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoSUNRPC: create RPC pipefs superblock per network namespace context
Stanislav Kinsbursky [Mon, 26 Dec 2011 12:38:56 +0000 (15:38 +0300)]
SUNRPC: create RPC pipefs superblock per network namespace context

This is the initial step of RPC pipefs virtualization. It changes nothing to
current pipefs behaviour except that mount of pipefs in other than init_net
network namespace context will provide only root tree. No other dentries will
be visible.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoSUNRPC: remove non-exclusive pipe creation from RPC pipefs
Stanislav Kinsbursky [Tue, 8 Nov 2011 12:09:19 +0000 (15:09 +0300)]
SUNRPC: remove non-exclusive pipe creation from RPC pipefs

This patch-set was created in context of clone of git branch:
git://git.linux-nfs.org/projects/trondmy/nfs-2.6.git.

v2:
1) Rebased of current repo state (i.e. all commits were pulled before apply)

I feel it is ready for inclusion if no objections will appear.

SUNRPC pipefs non-exclusive pipe creation code looks obsolete. IOW, as I see
it, all pipes are creating with unique full path and only once.

Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
12 years agoLinux 3.3-rc2
Linus Torvalds [Tue, 31 Jan 2012 21:31:54 +0000 (13:31 -0800)]
Linux 3.3-rc2

12 years agoMerge tag 'for-linus' of git://github.com/prasad-joshi/logfs_upstream
Linus Torvalds [Tue, 31 Jan 2012 17:23:59 +0000 (09:23 -0800)]
Merge tag 'for-linus' of git://github.com/prasad-joshi/logfs_upstream

There are few important bug fixes for LogFS

* tag 'for-linus' of git://github.com/prasad-joshi/logfs_upstream:
  Logfs: Allow NULL block_isbad() methods
  logfs: Grow inode in delete path
  logfs: Free areas before calling generic_shutdown_super()
  logfs: remove useless BUG_ON
  MAINTAINERS: Add Prasad Joshi in LogFS maintiners
  logfs: Propagate page parameter to __logfs_write_inode
  logfs: set superblock shutdown flag after generic sb shutdown
  logfs: take write mutex lock during fsync and sync
  logfs: Prevent memory corruption
  logfs: update page reference count for pined pages

Fix up conflict in fs/logfs/dev_mtd.c due to semantic change in what
"mtd->block_isbad" means in commit f2933e86ad93: "Logfs: Allow NULL
block_isbad() methods" clashing with the abstraction changes in the
commits 7086c19d0742: "mtd: introduce mtd_block_isbad interface" and
d58b27ed58a3: "logfs: do not use 'mtd->block_isbad' directly".

This resolution takes the semantics from commit f2933e86ad93, and just
makes mtd_block_isbad() return zero (false) if the 'block_isbad'
function is NULL.  But that also means that now "mtd_can_have_bb()"
always returns 0.

Now, "mtd_block_markbad()" will obviously return an error if the
low-level driver doesn't support bad blocks, so this is somewhat
non-symmetric, but it actually makes sense if a NULL "block_isbad"
function is considered to mean "I assume that all my blocks are always
good".

12 years agoMerge branch 'hwmon-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/groec...
Linus Torvalds [Tue, 31 Jan 2012 01:08:40 +0000 (17:08 -0800)]
Merge branch 'hwmon-for-linus' of git://git./linux/kernel/git/groeck/linux-staging

* 'hwmon-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/groeck/linux-staging:
  hwmon: (w83627ehf) Disable setting DC mode for pwm2, pwm3 on NCT6776F
  hwmon: (sht15) fix bad error code
  MAINTAINERS: Drop maintainer for MAX1668 hwmon driver
  MAINTAINERS: Add hwmon entries for Wolfson
  hwmon: (f71805f) Fix clamping of temperature limits

12 years agoMerge branch 'for-torvalds' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw...
Linus Torvalds [Tue, 31 Jan 2012 01:06:26 +0000 (17:06 -0800)]
Merge branch 'for-torvalds' of git://git./linux/kernel/git/linusw/linux-pinctrl

Here are some fixes to the pin control system that has accumulated since
-rc1.  Mainly Tony Lindgren fixed the module load/unload logic and the
rest are minor fixes and documentation.

* 'for-torvalds' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw/linux-pinctrl:
  pinctrl: add checks for empty function names
  pinctrl: fix pinmux_hog_maps when ctrl_dev_name is not set
  pinctrl: fix some pinmux typos
  pinctrl: free debugfs entries when unloading a pinmux driver
  pinctrl: unbreak error messages
  Documentation/pinctrl: fix a few syntax errors in code examples
  pinctrl: fix pinconf_pins_show iteration

12 years agoMerge tag 'tty-3.3-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty
Linus Torvalds [Mon, 30 Jan 2012 23:17:21 +0000 (15:17 -0800)]
Merge tag 'tty-3.3-rc1' of git://git./linux/kernel/git/gregkh/tty

Here are some tty/serial patches for 3.3-rc1

Big thing here is the movement of the 8250 serial drivers to their own
directory, now that the patch churn has calmed down.

Other than that, only minor stuff (omap patches were reverted as they
were found to be wrong), and another broken driver removed from the
system.

Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
* tag 'tty-3.3-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty:
  serial: Kill off Moorestown code
  Revert "tty: serial: OMAP: ensure FIFO levels are set correctly in non-DMA mode"
  Revert "tty: serial: OMAP: transmit FIFO threshold interrupts don't wake the chip"
  serial: Fix wakeup init logic to speed up startup
  docbook: don't use serial_core.h in device-drivers book
  serial: amba-pl011: lock console writes against interrupts
  amba-pl011: do not disable RTS during shutdown
  tty: serial: OMAP: transmit FIFO threshold interrupts don't wake the chip
  tty: serial: OMAP: ensure FIFO levels are set correctly in non-DMA mode
  omap-serial: make serial_omap_restore_context depend on CONFIG_PM_RUNTIME
  omap-serial :Make the suspend/resume functions depend on CONFIG_PM_SLEEP.
  TTY: fix UV serial console regression
  jsm: Fixed EEH recovery error
  Updated TTY MAINTAINERS info
  serial: group all the 8250 related code together

12 years agoMerge tag 'usb-3.3-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb
Linus Torvalds [Mon, 30 Jan 2012 19:38:28 +0000 (11:38 -0800)]
Merge tag 'usb-3.3-rc1' of git://git./linux/kernel/git/gregkh/usb

Here are a bunch of USB patches for 3.3-rc1.

Nothing major, largest thing here is the removal of some drivers that
did not work at all.  Other than that, the normal collection of bugfixes
and new device ids.

Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
* tag 'usb-3.3-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb: (52 commits)
  uwb & wusb: fix kconfig error
  USB: Realtek cr: fix autopm scheduling while atomic
  USB: ftdi_sio: Add more identifiers
  xHCI: Cleanup isoc transfer ring when TD length mismatch found
  usb: musb: omap2430: minor cleanups.
  qcaux: add more Pantech UML190 and UML290 ports
  Revert "drivers: usb: Fix dependency for USB_HWA_HCD"
  usb: mv-otg - Fix build if CONFIG_USB is not set
  USB: cdc-wdm: Avoid hanging on interface with no USB_CDC_DMM_TYPE
  usb: add support for STA2X11 host driver
  drivers: usb: Fix dependency for USB_HWA_HCD
  kernel-doc: fix new warning in usb.h
  USB: OHCI: fix new compiler warnings
  usb: serial: kobil_sct: fix compile warning:
  drivers/usb/host/ehci-fsl.c: add missing iounmap
  USB: cdc-wdm: better allocate a buffer that is at least as big as we tell the USB core
  USB: cdc-wdm: call wake_up_all to allow driver to shutdown on device removal
  USB: cdc-wdm: use two mutexes to allow simultaneous read and write
  USB: cdc-wdm: updating desc->length must be protected by spin_lock
  USB: usbsevseg: fix max length
  ...

12 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Linus Torvalds [Mon, 30 Jan 2012 18:53:20 +0000 (10:53 -0800)]
Merge git://git./linux/kernel/git/davem/net

1) Setting link attributes can modify the size of the attributes that
   would be reported on a subsequent getlink netlink operation,
   therefore min_ifinfo_dump_size needs to be adjusted.  From Stefan
   Gula.

2) Resegmentation of TSO frames while trimming can violate invariants
   expected by callers, namely that the number of segments can only stay
   the same or decrease, never increase.  If MSS changes, however, we
   can trim data but then end up with more segments.  Fix this by only
   segmenting to the MSS already recorded in the SKB.  That's the
   simplest fix for now and if we want to get more fancy in the future
   that's a more involved change.

   This probably explains some retransmit counter inaccuracies.

   From Neal Cardwell.

3) Fix too-many-wakeups in POLL with AF_UNIX sockets, from Eric Dumazet.

4) Fix CAIF crashes wrt.  namespace handling.  From Eric Dumazet and
   Eric W. Biederman.

5) TCP port selection fixes from Flavio Leitner.

6) More socket memory cgroup build fixes in certain randonfig
   situations.  From Glauber Costa.

7) Fix TCP memory sysctl regression reported by Ingo Molnar, also from
   Glauber Costa.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net:
  af_unix: fix EPOLLET regression for stream sockets
  tcp: fix tcp_trim_head() to adjust segment count with skb MSS
  net/tcp: Fix tcp memory limits initialization when !CONFIG_SYSCTL
  net caif: Register properly as a pernet subsystem.
  netns: Fail conspicously if someone uses net_generic at an inappropriate time.
  net: explicitly add jump_label.h header to sock.h
  net: RTNETLINK adjusting values of min_ifinfo_dump_size
  ipv6: Fix ip_gre lockless xmits.
  xen-netfront: correct MAX_TX_TARGET calculation.
  netns: fix net_alloc_generic()
  tcp: bind() optimize port allocation
  tcp: bind() fix autoselection to share ports
  l2tp: l2tp_ip - fix possible oops on packet receive
  iwlwifi: fix PCI-E transport "inta" race
  mac80211: set bss_conf.idle when vif is connected
  mac80211: update oper_channel on ibss join

12 years agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regulator
Linus Torvalds [Mon, 30 Jan 2012 18:16:25 +0000 (10:16 -0800)]
Merge tag 'for-linus' of git://git./linux/kernel/git/broonie/regulator

This fixes an integration issue with the regulator device tree bindings
which shook out in -rc.  The bindings were overly enthusiatic when
deciding to set a voltage on a regulator and would try to set zero volts
on an unconfigured regulator which isn't supported.

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regulator:
  regulator: Set apply_uV only when min and max voltages are defined

12 years agoaf_unix: fix EPOLLET regression for stream sockets
Eric Dumazet [Sat, 28 Jan 2012 16:11:03 +0000 (16:11 +0000)]
af_unix: fix EPOLLET regression for stream sockets

Commit 0884d7aa24 (AF_UNIX: Fix poll blocking problem when reading from
a stream socket) added a regression for epoll() in Edge Triggered mode
(EPOLLET)

Appropriate fix is to use skb_peek()/skb_unlink() instead of
skb_dequeue(), and only call skb_unlink() when skb is fully consumed.

This remove the need to requeue a partial skb into sk_receive_queue head
and the extra sk->sk_data_ready() calls that added the regression.

This is safe because once skb is given to sk_receive_queue, it is not
modified by a writer, and readers are serialized by u->readlock mutex.

This also reduce number of spinlock acquisition for small reads or
MSG_PEEK users so should improve overall performance.

Reported-by: Nick Mathewson <nickm@freehaven.net>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Alexey Moiseytsev <himeraster@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agotcp: fix tcp_trim_head() to adjust segment count with skb MSS
Neal Cardwell [Sat, 28 Jan 2012 17:29:46 +0000 (17:29 +0000)]
tcp: fix tcp_trim_head() to adjust segment count with skb MSS

This commit fixes tcp_trim_head() to recalculate the number of
segments in the skb with the skb's existing MSS, so trimming the head
causes the skb segment count to be monotonically non-increasing - it
should stay the same or go down, but not increase.

Previously tcp_trim_head() used the current MSS of the connection. But
if there was a decrease in MSS between original transmission and ACK
(e.g. due to PMTUD), this could cause tcp_trim_head() to
counter-intuitively increase the segment count when trimming bytes off
the head of an skb. This violated assumptions in tcp_tso_acked() that
tcp_trim_head() only decreases the packet count, so that packets_acked
in tcp_tso_acked() could underflow, leading tcp_clean_rtx_queue() to
pass u32 pkts_acked values as large as 0xffffffff to
ca_ops->pkts_acked().

As an aside, if tcp_trim_head() had really wanted the skb to reflect
the current MSS, it should have called tcp_set_skb_tso_segs()
unconditionally, since a decrease in MSS would mean that a
single-packet skb should now be sliced into multiple segments.

Signed-off-by: Neal Cardwell <ncardwell@google.com>
Acked-by: Nandita Dukkipati <nanditad@google.com>
Acked-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agonet/tcp: Fix tcp memory limits initialization when !CONFIG_SYSCTL
Glauber Costa [Mon, 30 Jan 2012 01:20:17 +0000 (01:20 +0000)]
net/tcp: Fix tcp memory limits initialization when !CONFIG_SYSCTL

sysctl_tcp_mem() initialization was moved to sysctl_tcp_ipv4.c
in commit 3dc43e3e4d0b52197d3205214fe8f162f9e0c334, since it
became a per-ns value.

That code, however, will never run when CONFIG_SYSCTL is
disabled, leading to bogus values on those fields - causing hung
TCP sockets.

This patch fixes it by keeping an initialization code in
tcp_init(). It will be overwritten by the first net namespace
init if CONFIG_SYSCTL is compiled in, and do the right thing if
it is compiled out.

It is also named properly as tcp_init_mem(), to properly signal
its non-sysctl side effect on TCP limits.

Reported-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Glauber Costa <glommer@parallels.com>
Cc: David S. Miller <davem@davemloft.net>
Link: http://lkml.kernel.org/r/4F22D05A.8030604@parallels.com
[ renamed the function, tidied up the changelog a bit ]
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux
Linus Torvalds [Mon, 30 Jan 2012 17:02:10 +0000 (09:02 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/s390/linux

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux:
  [S390] dasd: revalidate server for new pathgroup
  [S390] dasd: revert LCU optimization
  [S390] cleanup entry point definition

12 years agoMerge branch 'next' of git://git.monstr.eu/linux-2.6-microblaze
Linus Torvalds [Mon, 30 Jan 2012 16:59:46 +0000 (08:59 -0800)]
Merge branch 'next' of git://git.monstr.eu/linux-2.6-microblaze

* 'next' of git://git.monstr.eu/linux-2.6-microblaze:
  microblaze: generic atomic64 support

12 years agoMerge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux
Linus Torvalds [Mon, 30 Jan 2012 16:56:41 +0000 (08:56 -0800)]
Merge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux

* 'drm-fixes' of git://people.freedesktop.org/~airlied/linux:
  vmwgfx: Fix assignment in vmw_framebuffer_create_handle
  drm/radeon/kms: Fix device tree linkage of i2c buses
  drm: Pass the real error code back during GEM bo initialisation
  Revert "drm/i810: cleanup reclaim_buffers"

12 years agoMerge tag 'nfs-for-3.3-3' of git://git.linux-nfs.org/projects/trondmy/linux-nfs
Linus Torvalds [Mon, 30 Jan 2012 16:47:49 +0000 (08:47 -0800)]
Merge tag 'nfs-for-3.3-3' of git://git.linux-nfs.org/projects/trondmy/linux-nfs

NFS client bugfixes for Linux 3.3 (pull 3)

* tag 'nfs-for-3.3-3' of git://git.linux-nfs.org/projects/trondmy/linux-nfs:
  SUNRPC: Fix machine creds in generic_create_cred and generic_match

12 years agoMerge tag 'pm-fix-for-3.3-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/rafae...
Linus Torvalds [Mon, 30 Jan 2012 16:33:40 +0000 (08:33 -0800)]
Merge tag 'pm-fix-for-3.3-rc2' of git://git./linux/kernel/git/rafael/linux-pm

Power management fix for 3.3-rc2

Fix for a hibernate (s2disk) regression introduced during the 3.2
merge window that causes s2disk to trigger BUG_ON() in
freeze_workqueues_begin() if there is not enough swap space to save
the image.

* tag 'pm-fix-for-3.3-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  PM / Hibernate: Fix s2disk regression related to freezing workqueues

12 years agovmwgfx: Fix assignment in vmw_framebuffer_create_handle
Ryan Mallon [Fri, 27 Jan 2012 21:51:40 +0000 (08:51 +1100)]
vmwgfx: Fix assignment in vmw_framebuffer_create_handle

The assignment of handle in vmw_framebuffer_create_handle doesn't actually do anything useful and is incorrectly assigning an integer value to a pointer argument. It appears that this is a typo and should be dereferencing handle rather than assigning to it directly. This fixes a bug where an undefined handle value is potentially returned to user-space.

Signed-off-by: Ryan Mallon <rmallon@gmail.com>
Reviewed-by: Jakob Bornecrantz<jakob@vmware.com>
Cc: stable@vger.kernel.org
Signed-off-by: Dave Airlie <airlied@redhat.com>
12 years agodrm/radeon/kms: Fix device tree linkage of i2c buses
Jean Delvare [Sat, 28 Jan 2012 10:10:38 +0000 (11:10 +0100)]
drm/radeon/kms: Fix device tree linkage of i2c buses

Properly set the parent device of i2c buses before registering them so
that they will show at the right place in the device tree (rather than
in /sys/devices directly.)

Signed-off-by: Jean Delvare <jdelvare@suse.de>
Cc: Dave Airlie <airlied@gmail.com>
Reviewed-by: Alex Deucher <alexdeucher@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
12 years agodrm: Pass the real error code back during GEM bo initialisation
Chris Wilson [Sun, 29 Jan 2012 16:45:32 +0000 (16:45 +0000)]
drm: Pass the real error code back during GEM bo initialisation

In particular, I found I was hitting the max-file limit in the VFS,
and the EFILE was being magically transformed into ENOMEM. Confusion
reigns.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
12 years agoRevert "drm/i810: cleanup reclaim_buffers"
Daniel Vetter [Sun, 29 Jan 2012 16:05:52 +0000 (17:05 +0100)]
Revert "drm/i810: cleanup reclaim_buffers"

This reverts commit 87499ffdcb1c70f66988cd8febc4ead0ba2f9118.

Where is that paper bag ... ah here.

I've failed to take an odd interaction between my other cleanups and
this reclaim_buffers patch into account and also failed to properly
test it. Looks like there are more dragons and hidden trapdoors in the
drm release path than actual lines of code.

Until I get a clue, let's just revert this.

Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
12 years agohwmon: (w83627ehf) Disable setting DC mode for pwm2, pwm3 on NCT6776F
Guenter Roeck [Sat, 28 Jan 2012 01:56:06 +0000 (17:56 -0800)]
hwmon: (w83627ehf) Disable setting DC mode for pwm2, pwm3 on NCT6776F

NCT6776F only supports pwm mode for pwm2 and pwm3. Return error if an attempt
is made to set those pwm channels to DC mode.

Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Acked-by: Jean Delvare <khali@linux-fr.org>
Cc: stable@vger.kernel.org # 3.0+
Signed-off-by: Guenter Roeck <guenter.roeck@ericsson.com>
12 years agoPM / Hibernate: Fix s2disk regression related to freezing workqueues
Rafael J. Wysocki [Sun, 29 Jan 2012 19:35:52 +0000 (20:35 +0100)]
PM / Hibernate: Fix s2disk regression related to freezing workqueues

Commit 2aede851ddf08666f68ffc17be446420e9d2a056

  PM / Hibernate: Freeze kernel threads after preallocating memory

introduced a mechanism by which kernel threads were frozen after
the preallocation of hibernate image memory to avoid problems with
frozen kernel threads not responding to memory freeing requests.
However, it overlooked the s2disk code path in which the
SNAPSHOT_CREATE_IMAGE ioctl was run directly after SNAPSHOT_FREE,
which caused freeze_workqueues_begin() to BUG(), because it saw
that worqueues had been already frozen.

Although in principle this issue might be addressed by removing
the relevant BUG_ON() from freeze_workqueues_begin(), that would
reintroduce the very problem that commit 2aede851ddf08666f68ffc17be4
attempted to avoid into that particular code path.  For this reason,
to fix the issue at hand, introduce thaw_kernel_threads() and make
the SNAPSHOT_FREE ioctl execute it.

Special thanks to Srivatsa S. Bhat for detailed analysis of the
problem.

Reported-and-tested-by: Jiri Slaby <jslaby@suse.cz>
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Acked-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Cc: stable@kernel.org
12 years agohwmon: (sht15) fix bad error code
Vivien Didelot [Thu, 26 Jan 2012 20:59:00 +0000 (15:59 -0500)]
hwmon: (sht15) fix bad error code

When no platform data was supplied, returned error code was 0.

Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Cc: stable@vger.kernel.org # 2.6.32+
Signed-off-by: Guenter Roeck <guenter.roeck@ericsson.com>
12 years agoMerge tag 'driver-core-3.3-rc1-bugfixes' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 29 Jan 2012 02:20:48 +0000 (18:20 -0800)]
Merge tag 'driver-core-3.3-rc1-bugfixes' of git://git./linux/kernel/git/gregkh/driver-core

Here are some patches for the 3.3-rc1 tree.

It contains the removal of the sysdev code, now that all users of it are
gone, as well as some sysfs bugfixes that have been reported by users.
There are also some documentation updates here as well.

* tag 'driver-core-3.3-rc1-bugfixes' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core:
  sysfs: Complain bitterly about attempts to remove files from nonexistent directories.
  stable: update documentation to ask for kernel version
  base/core.c:fix typo in comment in function device_add
  Documentation: devres: add allocation functions to list of supported calls
  Documentation update for the driver model core
  kernel-doc: fix new warnings in driver-core
  kernel-doc: fix new warnings in debugfs
  kernel-doc: fix new warnings in device.h
  driver core: remove drivers/base/sys.c and include/linux/sysdev.h

12 years agoMerge tag 'for-linus' of git://github.com/rustyrussell/linux
Linus Torvalds [Sun, 29 Jan 2012 02:16:09 +0000 (18:16 -0800)]
Merge tag 'for-linus' of git://github.com/rustyrussell/linux

* tag 'for-linus' of git://github.com/rustyrussell/linux:
  lguest: remove reference from Documentation/virtual/00-INDEX
  virtio: correct the memory barrier in virtqueue_kick_prepare()
  virtio: fix typos of memory barriers

12 years agoMerge branch 'stable/for-linus-fixes-3.3' of git://git.kernel.org/pub/scm/linux/kerne...
Linus Torvalds [Sun, 29 Jan 2012 02:15:33 +0000 (18:15 -0800)]
Merge branch 'stable/for-linus-fixes-3.3' of git://git./linux/kernel/git/konrad/xen

* 'stable/for-linus-fixes-3.3' of git://git.kernel.org/pub/scm/linux/kernel/git/konrad/xen:
  xen/granttable: Disable grant v2 for HVM domains.
  x86: xen: size struct xen_spinlock to always fit in arch_spinlock_t

12 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux...
Linus Torvalds [Sun, 29 Jan 2012 01:00:19 +0000 (17:00 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/mason/linux-btrfs

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs:
  Btrfs: fix reservations in btrfs_page_mkwrite
  Btrfs: advance window_start if we're using a bitmap
  btrfs: mask out gfp flags in releasepage
  Btrfs: fix enospc error caused by wrong checks of the chunk
  Btrfs: do not defrag a file partially
  Btrfs: fix warning for 32-bit build of fs/btrfs/check-integrity.c
  Btrfs: use cluster->window_start when allocating from a cluster bitmap
  Btrfs: Check for NULL page in extent_range_uptodate
  btrfs: Fix busyloops in transaction waiting code
  Btrfs: make sure a bitmap has enough bytes
  Btrfs: fix uninit warning in backref.c

12 years agoMerge git://www.linux-watchdog.org/linux-watchdog
Linus Torvalds [Sun, 29 Jan 2012 00:57:15 +0000 (16:57 -0800)]
Merge git://www.linux-watchdog.org/linux-watchdog

* git://www.linux-watchdog.org/linux-watchdog:
  watchdog: iTCO_wdt: add Intel Lynx Point DeviceIDs
  watchdog: via_wdt: Set min_timeout and max_timeout for wdt_dev
  watchdog: Fix typo "unexpectdly"
  watchdog: wafer5823wdt: Fix handling WDIOS_DISABLECARD/WDIOS_ENABLECARD options
  watchdog: wm8350_wdt: Fix handling WDIOS_DISABLECARD/WDIOS_ENABLECARD options
  watchdog: Return proper error in nuc900wdt_probe if misc_register fails
  watchdog: Staticise nuc900_wdt
  watchdog: via_wdt: Staticise wdt_pci_table
  watchdog: omap_wdt.c: Fix the mismatch of pm_runtime enable and disable
  watchdog: dw_wdt.c: use devm_request_and_ioremap
  watchdog: imx2_wdt.c: use devm_request_and_ioremap

12 years agoMerge branch 'fixes' of git://git.linaro.org/people/rmk/linux-arm
Linus Torvalds [Sat, 28 Jan 2012 21:27:10 +0000 (13:27 -0800)]
Merge branch 'fixes' of git://git.linaro.org/people/rmk/linux-arm

* 'fixes' of git://git.linaro.org/people/rmk/linux-arm: (31 commits)
  ARM: 7304/1: ioremap: fix boundary check when reusing static mapping
  ARM: 7301/1: Rename the T() macro to TUSER() to avoid namespace conflicts
  ARM: 7299/1: ftrace: clear zero bit in reported IPs for Thumb-2
  ARM: 7298/1: realview: fix mapping of MPCore private memory region
  PCMCIA: fix sa1111 oops on remove
  ARM: 7288/1: mach-sa1100: add missing module_init() call
  ARM: 7297/1: smp_twd: make sure timer is stopped before registering it
  ARM: 7296/1: proc-v7.S: remove HARVARD_CACHE preprocessor guards
  ARM: 7295/1: cortex-a7: move proc_info out of !CONFIG_ARM_LPAE block
  ARM: 7293/1: logical_cpu_map: decouple CPU mapping from SMP
  ARM: 7291/1: cache: assume 64-byte L1 cachelines for ARMv7 CPUs
  ARM: 7290/1: vmlinux.lds.S: align the exception fixup table to a 4-byte boundary
  ARM: 7289/1: vmlinux.lds.S: do not hardcode cacheline size as 32 bytes
  MFD: ucb1x00-ts: fix resume failure
  MFD: ucb1x00-core: fix gpiolib direction_output handling
  MFD: ucb1x00-core: fix missing restore of io output data on resume
  MFD: mcp-core: fix mcp_priv() to be more type safe
  MFD: mcp-core: fix complaints from the genirq layer
  Revert "ARM: sa11x0: Implement autoloading of codec and codec pdata for mcp bus."
  Revert "ARM: sa1100: Refactor mcp-sa11x0 to use platform resources."
  ...

Fix up conflict due to arch/arm/mach-mx5/Kconfig having been merged into
mach-imx5 (commit 784a90c0a7d8: "ARM i.MX: Merge i.MX5 support into
mach-imx"), but the ARM_L1_CACHE_SHIFT_6 entry was moved to be driven by
the CPU_V7 logic from it in the old location in rmk's branch (commit
a092f2b15399: "ARM: 7291/1: cache: assume 64-byte L1 cachelines for
ARMv7 CPUs").

12 years agoMerge tag 'fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arm...
Linus Torvalds [Sat, 28 Jan 2012 21:21:54 +0000 (13:21 -0800)]
Merge tag 'fixes-for-linus' of git://git./linux/kernel/git/arm/arm-soc

arm-soc fixes for 3.3-rc:

AT91 needed reset fixes which resulted in some minor code refactoring,
it also adds a feature-removal for one of their platforms for 3.4.
The USB patches have been acked by Greg K-H.

i.MX and ux500 both have some minor fixes, nothing controversial.

* tag 'fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc:
  arch/arm/mach-imx/mach-mx53_ard.c: add missing iounmap
  ARM: imx: iomux-v1.h: Fix build error due to __init annotation
  ARM: at91: Fix at91sam9g45 and at91cap9 reset
  ARM: at91: make rstc soc independent
  ARM: at91: introduce AT91_SAM9_ALT_RESET to select the at91sam9 alternative reset
  ARM: at91: merge at91cap9_ddrsdr.h in at91sam9_ddrsdr.h
  ARM: at91: fix cap9 ddrsdr register
  ARM/USB: at91/ohci-at91: rename vbus_pin_inverted to vbus_pin_active_low
  USB: at91: fix clk_get error handling
  ARM: at91: removal of CAP9 SoC family
  ARM: at91: fix at91rm9200 soc subtype handling
  mach-ux500: no MMC_CAP_SD_HIGHSPEED on Snowball
  mach-ux500: enable ARM errata 764369
  mach-ux500: do not override outer.inv_all
  mach-ux500: musb: now musb is always in OTG mode
  ARM: imx6: add missing twd_clk for imx6q clock

12 years agoLogfs: Allow NULL block_isbad() methods
Joern Engel [Fri, 5 Aug 2011 09:09:55 +0000 (11:09 +0200)]
Logfs: Allow NULL block_isbad() methods

Not all mtd drivers define block_isbad().  Let's assume no bad blocks
instead of refusing to mount.

Signed-off-by: Joern Engel <joern@logfs.org>
12 years agologfs: Grow inode in delete path
Joern Engel [Fri, 5 Aug 2011 09:13:30 +0000 (11:13 +0200)]
logfs: Grow inode in delete path

Can be necessary if an inode gets deleted (through -ENOSPC) before being
written.  Might be better to move this into logfs_write_rec(), but for
now go with the stupid&safe patch.

Signed-off-by: Joern Engel <joern@logfs.org>
12 years agologfs: Free areas before calling generic_shutdown_super()
Joern Engel [Fri, 5 Aug 2011 09:18:19 +0000 (11:18 +0200)]
logfs: Free areas before calling generic_shutdown_super()

Or hit an assertion in map_invalidatepage() instead.

Signed-off-by: Joern Engel <joern@logfs.org>
12 years agologfs: remove useless BUG_ON
Joern Engel [Mon, 12 Sep 2011 15:39:16 +0000 (21:09 +0530)]
logfs: remove useless BUG_ON

It prevents write sizes >4k.

Signed-off-by: Joern Engel <joern@logfs.org>
12 years agoMAINTAINERS: Add Prasad Joshi in LogFS maintiners
Prasad Joshi [Tue, 13 Sep 2011 17:34:11 +0000 (23:04 +0530)]
MAINTAINERS: Add Prasad Joshi in LogFS maintiners

Acked-by: Joern Engel <joern@logfs.org>
Signed-off-by: Prasad Joshi <prasadjoshi.linux@gmail.com>
12 years agologfs: Propagate page parameter to __logfs_write_inode
Prasad Joshi [Sun, 2 Oct 2011 18:16:51 +0000 (23:46 +0530)]
logfs: Propagate page parameter to __logfs_write_inode

During GC LogFS has to rewrite each valid block to a separate segment.
Rewrite operation reads data from an old segment and writes it to a
newly allocated segment. Since every write operation changes data
block pointers maintained in inode, inode should also be rewritten.

In GC path to avoid AB-BA deadlock LogFS marks a page with
PG_pre_locked in addition to locking the page (PG_locked). The page
lock is ignored iff the page is pre-locked.

LogFS uses a special file called segment file. The segment file
maintains an 8 bytes entry for every segment. It keeps track of erase
count, level etc. for every segment.

Bad things happen with a segment belonging to the segment file is GCed

 ------------[ cut here ]------------
kernel BUG at /home/prasad/logfs/readwrite.c:297!
invalid opcode: 0000 [#1] SMP
Modules linked in: logfs joydev usbhid hid psmouse e1000 i2c_piix4
serio_raw [last unloaded: logfs]
Pid: 20161, comm: mount Not tainted 3.1.0-rc3+ #3 innotek GmbH
VirtualBox
EIP: 0060:[<f809132a>] EFLAGS: 00010292 CPU: 0
EIP is at logfs_lock_write_page+0x6a/0x70 [logfs]
EAX: 00000027 EBX: f73f5b20 ECX: c16007c8 EDX: 00000094
ESI: 00000000 EDI: e59be6e4 EBP: c7337b28 ESP: c7337b18
DS: 007b ES: 007b FS: 00d8 GS: 00e0 SS: 0068
Process mount (pid: 20161, ti=c7336000 task=eb323f70 task.ti=c7336000)
Stack:
f8099a3d c7337b24 f73f5b20 00001002 c7337b50 f8091f6d f8099a4d f80994e4
00000003 00000000 c7337b68 00000000 c67e4400 00001000 c7337b80 f80935e5
00000000 00000000 00000000 00000000 e1fcf000 0000000f e59be618 c70bf900
Call Trace:
[<f8091f6d>] logfs_get_write_page.clone.16+0xdd/0x100 [logfs]
[<f80935e5>] logfs_mod_segment_entry+0x55/0x110 [logfs]
[<f809460d>] logfs_get_segment_entry+0x1d/0x20 [logfs]
[<f8091060>] ? logfs_cleanup_journal+0x50/0x50 [logfs]
[<f809521b>] ostore_get_erase_count+0x1b/0x40 [logfs]
[<f80965b8>] logfs_open_area+0xc8/0x150 [logfs]
[<c141a7ec>] ? kmemleak_alloc+0x2c/0x60
[<f809668e>] __logfs_segment_write.clone.16+0x4e/0x1b0 [logfs]
[<c10dd563>] ? mempool_kmalloc+0x13/0x20
[<c10dd563>] ? mempool_kmalloc+0x13/0x20
[<f809696f>] logfs_segment_write+0x17f/0x1d0 [logfs]
[<f8092e8c>] logfs_write_i0+0x11c/0x180 [logfs]
[<f8092f35>] logfs_write_direct+0x45/0x90 [logfs]
[<f80934cd>] __logfs_write_buf+0xbd/0xf0 [logfs]
[<c102900e>] ? kmap_atomic_prot+0x4e/0xe0
[<f809424b>] logfs_write_buf+0x3b/0x60 [logfs]
[<f80947a9>] __logfs_write_inode+0xa9/0x110 [logfs]
[<f8094cb0>] logfs_rewrite_block+0xc0/0x110 [logfs]
[<f8095300>] ? get_mapping_page+0x10/0x60 [logfs]
[<f8095aa0>] ? logfs_load_object_aliases+0x2e0/0x2f0 [logfs]
[<f808e57d>] logfs_gc_segment+0x2ad/0x310 [logfs]
[<f808e62a>] __logfs_gc_once+0x4a/0x80 [logfs]
[<f808ed43>] logfs_gc_pass+0x683/0x6a0 [logfs]
[<f8097a89>] logfs_mount+0x5a9/0x680 [logfs]
[<c1126b21>] mount_fs+0x21/0xd0
[<c10f6f6f>] ? __alloc_percpu+0xf/0x20
[<c113da41>] ? alloc_vfsmnt+0xb1/0x130
[<c113db4b>] vfs_kern_mount+0x4b/0xa0
[<c113e06e>] do_kern_mount+0x3e/0xe0
[<c113f60d>] do_mount+0x34d/0x670
[<c10f2749>] ? strndup_user+0x49/0x70
[<c113fcab>] sys_mount+0x6b/0xa0
[<c142d87c>] syscall_call+0x7/0xb
Code: f8 e8 8b 93 39 c9 8b 45 f8 3e 0f ba 28 00 19 d2 85 d2 74 ca eb d0 0f 0b 8d 45 fc 89 44 24 04 c7 04 24 3d 9a 09 f8 e8 09 92 39 c9 <0f> 0b 8d 74 26 00 55 89 e5 3e 8d 74 26 00 8b 10 80 e6 01 74 09
EIP: [<f809132a>] logfs_lock_write_page+0x6a/0x70 [logfs] SS:ESP 0068:c7337b18
---[ end trace 96e67d5b3aa3d6ca ]---

The patch passes locked page to __logfs_write_inode. It calls function
logfs_get_wblocks() to pre-lock the page. This ensures any further
attempts to lock the page are ignored (esp from get_erase_count).

Acked-by: Joern Engel <joern@logfs.org>
Signed-off-by: Prasad Joshi <prasadjoshi.linux@gmail.com>
12 years agologfs: set superblock shutdown flag after generic sb shutdown
Prasad Joshi [Sun, 30 Oct 2011 16:45:32 +0000 (22:15 +0530)]
logfs: set superblock shutdown flag after generic sb shutdown

While unmounting the file system LogFS calls generic_shutdown_super.
The function does file system independent superblock shutdown.
However, it might result in call file system specific inode eviction.

LogFS marks FS shutting down by setting bit LOGFS_SB_FLAG_SHUTDOWN in
super->s_flags. Since, inode eviction might call truncate on inode,
following BUG is observed when file system is unmounted:

------------[ cut here ]------------
kernel BUG at /home/prasad/logfs/segment.c:362!
invalid opcode: 0000 [#1] PREEMPT SMP
CPU 3
Modules linked in: logfs binfmt_misc ppdev virtio_blk parport_pc lp
parport psmouse floppy virtio_pci serio_raw virtio_ring virtio

Pid: 1933, comm: umount Not tainted 3.0.0+ #4 Bochs Bochs
RIP: 0010:[<ffffffffa008c841>]  [<ffffffffa008c841>]
logfs_segment_write+0x211/0x230 [logfs]
RSP: 0018:ffff880062d7b9e8  EFLAGS: 00010202
RAX: 000000000000000e RBX: ffff88006eca9000 RCX: 0000000000000000
RDX: ffff88006fd87c40 RSI: ffffea00014ff468 RDI: ffff88007b68e000
RBP: ffff880062d7ba48 R08: 8000000020451430 R09: 0000000000000000
R10: dead000000100100 R11: 0000000000000000 R12: ffff88006fd87c40
R13: ffffea00014ff468 R14: ffff88005ad0a460 R15: 0000000000000000
FS:  00007f25d50ea760(0000) GS:ffff88007fd80000(0000)
knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 000000008005003b
CR2: 0000000000d05e48 CR3: 0000000062c72000 CR4: 00000000000006e0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
Process umount (pid: 1933, threadinfo ffff880062d7a000,
task ffff880070b44500)
Stack:
ffff880062d7ba38 ffff88005ad0a508 0000000000001000 0000000000000000
8000000020451430 ffffea00014ff468 ffff880062d7ba48 ffff88005ad0a460
ffff880062d7bad8 ffffea00014ff468 ffff88006fd87c40 0000000000000000
Call Trace:
[<ffffffffa0088fee>] logfs_write_i0+0x12e/0x190 [logfs]
[<ffffffffa0089360>] __logfs_write_rec+0x140/0x220 [logfs]
[<ffffffffa0089312>] __logfs_write_rec+0xf2/0x220 [logfs]
[<ffffffffa00894a4>] logfs_write_rec+0x64/0xd0 [logfs]
[<ffffffffa0089616>] __logfs_write_buf+0x106/0x110 [logfs]
[<ffffffffa008a19e>] logfs_write_buf+0x4e/0x80 [logfs]
[<ffffffffa008a6b8>] __logfs_write_inode+0x98/0x110 [logfs]
[<ffffffffa008a7c4>] logfs_truncate+0x54/0x290 [logfs]
[<ffffffffa008abfc>] logfs_evict_inode+0xdc/0x190 [logfs]
[<ffffffff8115eef5>] evict+0x85/0x170
[<ffffffff8115f126>] iput+0xe6/0x1b0
[<ffffffff8115b4a8>] shrink_dcache_for_umount_subtree+0x218/0x280
[<ffffffff8115ce91>] shrink_dcache_for_umount+0x51/0x90
[<ffffffff8114796c>] generic_shutdown_super+0x2c/0x100
[<ffffffffa008cc47>] logfs_kill_sb+0x57/0xf0 [logfs]
[<ffffffff81147de5>] deactivate_locked_super+0x45/0x70
[<ffffffff811487ea>] deactivate_super+0x4a/0x70
[<ffffffff81163934>] mntput_no_expire+0xa4/0xf0
[<ffffffff8116469f>] sys_umount+0x6f/0x380
[<ffffffff814dd46b>] system_call_fastpath+0x16/0x1b
Code: 55 c8 49 8d b6 a8 00 00 00 45 89 f9 45 89 e8 4c 89 e1 4c 89 55
b8 c7 04 24 00 00 00 00 e8 68 fc ff ff 4c 8b 55 b8 e9 3c ff ff ff <0f>
0b 0f 0b c7 45 c0 00 00 00 00 e9 44 fe ff ff 66 66 66 66 66
RIP  [<ffffffffa008c841>] logfs_segment_write+0x211/0x230 [logfs]
RSP <ffff880062d7b9e8>
---[ end trace fe6b040cea952290 ]---

Therefore, move super->s_flags setting after the fs-indenpendent work
has been finished.

Reviewed-by: Joern Engel <joern@logfs.org>
Signed-off-by: Prasad Joshi <prasadjoshi.linux@gmail.com>
12 years agologfs: take write mutex lock during fsync and sync
Prasad Joshi [Sat, 28 Jan 2012 06:06:06 +0000 (11:36 +0530)]
logfs: take write mutex lock during fsync and sync

LogFS uses super->s_write_mutex while writing data to disk. Taking the
same mutex lock in sync and fsync code path solves the following BUG:

------------[ cut here ]------------
kernel BUG at /home/prasad/logfs/dev_bdev.c:134!

Pid: 2387, comm: flush-253:16 Not tainted 3.0.0+ #4 Bochs Bochs
RIP: 0010:[<ffffffffa007deed>]  [<ffffffffa007deed>]
                bdev_writeseg+0x25d/0x270 [logfs]
Call Trace:
[<ffffffffa007c381>] logfs_open_area+0x91/0x150 [logfs]
[<ffffffff8128dcb2>] ? find_level.clone.9+0x62/0x100
[<ffffffffa007c49c>] __logfs_segment_write.clone.20+0x5c/0x190 [logfs]
[<ffffffff810ef005>] ? mempool_kmalloc+0x15/0x20
[<ffffffff810ef383>] ? mempool_alloc+0x53/0x130
[<ffffffffa007c7a4>] logfs_segment_write+0x1d4/0x230 [logfs]
[<ffffffffa0078f8e>] logfs_write_i0+0x12e/0x190 [logfs]
[<ffffffffa0079300>] __logfs_write_rec+0x140/0x220 [logfs]
[<ffffffffa0079444>] logfs_write_rec+0x64/0xd0 [logfs]
[<ffffffffa00795b6>] __logfs_write_buf+0x106/0x110 [logfs]
[<ffffffffa007a13e>] logfs_write_buf+0x4e/0x80 [logfs]
[<ffffffffa0073e33>] __logfs_writepage+0x23/0x80 [logfs]
[<ffffffffa007410c>] logfs_writepage+0xdc/0x110 [logfs]
[<ffffffff810f5ba7>] __writepage+0x17/0x40
[<ffffffff810f6208>] write_cache_pages+0x208/0x4f0
[<ffffffff810f5b90>] ? set_page_dirty+0x70/0x70
[<ffffffff810f653a>] generic_writepages+0x4a/0x70
[<ffffffff810f75d1>] do_writepages+0x21/0x40
[<ffffffff8116b9d1>] writeback_single_inode+0x101/0x250
[<ffffffff8116bdbd>] writeback_sb_inodes+0xed/0x1c0
[<ffffffff8116c5fb>] writeback_inodes_wb+0x7b/0x1e0
[<ffffffff8116cc23>] wb_writeback+0x4c3/0x530
[<ffffffff814d984d>] ? sub_preempt_count+0x9d/0xd0
[<ffffffff8116cd6b>] wb_do_writeback+0xdb/0x290
[<ffffffff814d984d>] ? sub_preempt_count+0x9d/0xd0
[<ffffffff814d6208>] ? _raw_spin_unlock_irqrestore+0x18/0x40
[<ffffffff8105aa5a>] ? del_timer+0x8a/0x120
[<ffffffff8116cfac>] bdi_writeback_thread+0x8c/0x2e0
[<ffffffff8116cf20>] ? wb_do_writeback+0x290/0x290
[<ffffffff8106d2e6>] kthread+0x96/0xa0
[<ffffffff814de514>] kernel_thread_helper+0x4/0x10
[<ffffffff8106d250>] ? kthread_worker_fn+0x190/0x190
[<ffffffff814de510>] ? gs_change+0xb/0xb
RIP  [<ffffffffa007deed>] bdev_writeseg+0x25d/0x270 [logfs]
---[ end trace 0211ad60a57657c4 ]---

Reviewed-by: Joern Engel <joern@logfs.org>
Signed-off-by: Prasad Joshi <prasadjoshi.linux@gmail.com>
12 years agologfs: Prevent memory corruption
Joern Engel [Sun, 20 Nov 2011 16:59:01 +0000 (22:29 +0530)]
logfs: Prevent memory corruption

This is a bad one.  I wonder whether we were so far protected by
no_free_segments(sb) usually being smaller than LOGFS_NO_AREAS.

Found by Dan Carpenter <dan.carpenter@oracle.com> using smatch.

Signed-off-by: Joern Engel <joern@logfs.org>
Signed-off-by: Prasad Joshi <prasadjoshi.linux@gmail.com>
12 years agologfs: update page reference count for pined pages
Prasad Joshi [Sat, 26 Nov 2011 05:30:47 +0000 (11:00 +0530)]
logfs: update page reference count for pined pages

LogFS sets PG_private flag to indicate a pined page. We assumed that
marking a page as private is enough to ensure its existence. But
instead it is necessary to hold a reference count to the page.

The change resolves the following BUG

BUG: Bad page state in process flush-253:16  pfn:6a6d0
page flags: 0x100000000000808(uptodate|private)

Suggested-and-Acked-by: Joern Engel <joern@logfs.org>
Signed-off-by: Prasad Joshi <prasadjoshi.linux@gmail.com>
12 years agonet caif: Register properly as a pernet subsystem.
Eric W. Biederman [Thu, 26 Jan 2012 14:04:53 +0000 (14:04 +0000)]
net caif: Register properly as a pernet subsystem.

caif is a subsystem and as such it needs to register with
register_pernet_subsys instead of register_pernet_device.

Among other problems using register_pernet_device was resulting in
net_generic being called before the caif_net structure was allocated.
Which has been causing net_generic to fail with either BUG_ON's or by
return NULL pointers.

A more ugly problem that could be caused is packets in flight why the
subsystem is shutting down.

To remove confusion also remove the cruft cause by inappropriately
trying to fix this bug.

With the aid of the previous patch I have tested this patch and
confirmed that using register_pernet_subsys makes the failure go away as
it should.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Acked-by: Sjur Brændeland <sjur.brandeland@stericsson.com>
Tested-by: Sasha Levin <levinsasha928@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agonetns: Fail conspicously if someone uses net_generic at an inappropriate time.
Eric W. Biederman [Thu, 26 Jan 2012 14:02:55 +0000 (14:02 +0000)]
netns: Fail conspicously if someone uses net_generic at an inappropriate time.

By definition net_generic should never be called when it can return
NULL.  Fail conspicously with a BUG_ON to make it clear when people mess
up that a NULL return should never happen.

Recently there was a bug in the CAIF subsystem where it was registered
with register_pernet_device instead of register_pernet_subsys.  It was
erroneously concluded that net_generic could validly return NULL and
that net_assign_generic was buggy (when it was just inefficient).
Hopefully this BUG_ON will prevent people to coming to similar erroneous
conclusions in the futrue.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Tested-by: Sasha Levin <levinsasha928@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>