All the kernel_sendmsg() calls in rxrpc_send_data_packet() need to send
both parts of the iov[] buffer, but one of them does not. Fix it so that
it does.
Without this, short IPv6 rxrpc DATA packets may be seen that have the rxrpc
header included, but no payload.
Fixes: 5a924b8951 ("rxrpc: Don't store the rxrpc header in the Tx queue sk_buffs")
Reported-by: Marc Dionne <marc.dionne@auristor.com>
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
kmalloc() can't always allocate large enough buffers for big_key to use for
crypto (1MB + some metadata) so we cannot use that to allocate the buffer.
Further, vmalloc'd pages can't be passed to sg_init_one() and the aead
crypto accessors cannot be called progressively and must be passed all the
data in one go (which means we can't pass the data in one block at a time).
Fix this by allocating the buffer pages individually and passing them
through a multientry scatterlist to the crypto layer. This has the bonus
advantage that we don't have to allocate a contiguous series of pages.
We then vmap() the page list and pass that through to the VFS read/write
routines.
This can trigger a warning:
WARNING: CPU: 0 PID: 60912 at mm/page_alloc.c:3883 __alloc_pages_nodemask+0xb7c/0x15f8
([<00000000002acbb6>] __alloc_pages_nodemask+0x1ee/0x15f8)
[<00000000002dd356>] kmalloc_order+0x46/0x90
[<00000000002dd3e0>] kmalloc_order_trace+0x40/0x1f8
[<0000000000326a10>] __kmalloc+0x430/0x4c0
[<00000000004343e4>] big_key_preparse+0x7c/0x210
[<000000000042c040>] key_create_or_update+0x128/0x420
[<000000000042e52c>] SyS_add_key+0x124/0x220
[<00000000007bba2c>] system_call+0xc4/0x2b0
from the keyctl/padd/useradd test of the keyutils testsuite on s390x.
Note that it might be better to shovel data through in page-sized lumps
instead as there's no particular need to use a monolithic buffer unless the
kernel itself wants to access the data.
Fixes: 13100a72f4 ("Security: Keys: Big keys stored encrypted")
Reported-by: Paul Bunyan <pbunyan@redhat.com>
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Kirill Marinushkin <k.marinushkin@gmail.com>
Due to a check recently added to copy_to_user(), it's now not permitted to
copy from slab-held data to userspace unless the slab is whitelisted. This
affects rxrpc_recvmsg() when it attempts to place an RXRPC_USER_CALL_ID
control message in the userspace control message buffer. A warning is
generated by usercopy_warn() because the source is the copy of the
user_call_ID retained in the rxrpc_call struct.
Work around the issue by copying the user_call_ID to a variable on the
stack and passing that to put_cmsg().
The warning generated looks like:
Bad or missing usercopy whitelist? Kernel memory exposure attempt detected from SLUB object 'dmaengine-unmap-128' (offset 680, size 8)!
WARNING: CPU: 0 PID: 1401 at mm/usercopy.c:81 usercopy_warn+0x7e/0xa0
...
RIP: 0010:usercopy_warn+0x7e/0xa0
...
Call Trace:
__check_object_size+0x9c/0x1a0
put_cmsg+0x98/0x120
rxrpc_recvmsg+0x6fc/0x1010 [rxrpc]
? finish_wait+0x80/0x80
___sys_recvmsg+0xf8/0x240
? __clear_rsb+0x25/0x3d
? __clear_rsb+0x15/0x3d
? __clear_rsb+0x25/0x3d
? __clear_rsb+0x15/0x3d
? __clear_rsb+0x25/0x3d
? __clear_rsb+0x15/0x3d
? __clear_rsb+0x25/0x3d
? __clear_rsb+0x15/0x3d
? finish_task_switch+0xa6/0x2b0
? trace_hardirqs_on_caller+0xed/0x180
? _raw_spin_unlock_irq+0x29/0x40
? __sys_recvmsg+0x4e/0x90
__sys_recvmsg+0x4e/0x90
do_syscall_64+0x7a/0x220
entry_SYSCALL_64_after_hwframe+0x26/0x9b
Reported-by: Jonathan Billings <jsbillings@jsbillings.org>
Signed-off-by: David Howells <dhowells@redhat.com>
Acked-by: Kees Cook <keescook@chromium.org>
Tested-by: Jonathan Billings <jsbillings@jsbillings.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
AF_RXRPC is incorrectly sending back to the server any abort it receives
for a client connection. This is due to the final-ACK offload to the
connection event processor patch. The abort code is copied into the
last-call information on the connection channel and then the event
processor is set.
Instead, the following should be done:
(1) In the case of a final-ACK for a successful call, the ACK should be
scheduled as before.
(2) In the case of a locally generated ABORT, the ABORT details should be
cached for sending in response to further packets related to that
call and no further action scheduled at call disconnect time.
(3) In the case of an ACK received from the peer, the call should be
considered dead, no ABORT should be transmitted at this time. In
response to further non-ABORT packets from the peer relating to this
call, an RX_USER_ABORT ABORT should be transmitted.
(4) In the case of a call killed due to network error, an RX_USER_ABORT
ABORT should be cached for transmission in response to further
packets, but no ABORT should be sent at this time.
Fixes: 3136ef49a1 ("rxrpc: Delay terminal ACK transmission on a client call")
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Support the AFS dynamic root which is a pseudo-volume that doesn't connect
to any server resource, but rather is just a root directory that
dynamically creates mountpoint directories where the name of such a
directory is the name of the cell.
Such a mount can be created thus:
mount -t afs none /afs -o dyn
Dynamic root superblocks aren't shared except by bind mounts and
propagation. Cell root volumes can then be mounted by referring to them by
name, e.g.:
ls /afs/grand.central.org/
ls /afs/.grand.central.org/
The kernel will upcall to consult the DNS if the address wasn't supplied
directly.
Signed-off-by: David Howells <dhowells@redhat.com>
Rearrange afs_select_fileserver() a little to put the use_server chunk
before the next_server chunk so that with the removal of a couple of gotos
the main path through the function is all one sequence.
Signed-off-by: David Howells <dhowells@redhat.com>
Fix server list handling in the following ways:
(1) In afs_alloc_volume(), remove duplicate server list build code. This
was already done by afs_alloc_server_list() which afs_alloc_volume()
previously called. This just results in twice as many VL RPCs.
(2) In afs_deliver_vl_get_entry_by_name_u(), use the number of server
records indicated by ->nServers in the UVLDB record returned by the
VL.GetEntryByNameU RPC call rather than scanning all NMAXNSERVERS
slots. Unused slots may contain garbage.
(3) In afs_alloc_server_list(), don't stop converting a UVLDB record into
a server list just because we can't look up one of the servers. Just
skip that server and go on to the next. If we can't look up any of
the servers then we'll fail at the end.
Without this patch, an attempt to view the umich.edu root cell using
something like "ls /afs/umich.edu" on a dynamic root (future patch) mount
or an autocell mount will result in ENOMEDIUM. The failure is due to kafs
not stopping after nServers'worth of records have been read, but then
trying to access a server with a garbage UUID and getting an error, which
aborts the server list build.
Fixes: d2ddc776a4 ("afs: Overhaul volume and server record caching and fileserver rotation")
Reported-by: Jonathan Billings <jsbillings@jsbillings.org>
Signed-off-by: David Howells <dhowells@redhat.com>
cc: stable@vger.kernel.org
In afs_select_fileserver(), we need to clear the ->responded flag in the
address list when reusing it. We should also clear it in
afs_select_current_fileserver().
To this end, just memset() the object before initialising it.
Fixes: d2ddc776a4 ("afs: Overhaul volume and server record caching and fileserver rotation")
Signed-off-by: David Howells <dhowells@redhat.com>
cc: stable@vger.kernel.org
afs_select_fileserver() ends the address cursor it is using in the case in
which we get some sort of network error and run out of addresses to iterate
through, before it jumps to try the next server. This also needs to be
done when the server aborts with some sort of error that means we should
try the next server.
Fix this by:
(1) Move the iterate_address afs_end_cursor() call to the next_server
case.
(2) End the cursor in the failed case.
(3) Make afs_end_cursor() clear the ->begun flag and ->addr pointer in the
address cursor.
(4) Make afs_end_cursor() able to be called on an already cleared cursor.
Without this, something like the following oops may occur:
AFS: Assertion failed
18446612134397189888 == 0 is false
0xffff88007c279f00 == 0x0 is false
------------[ cut here ]------------
kernel BUG at fs/afs/rotate.c:360!
RIP: 0010:afs_select_fileserver+0x79b/0xa30 [kafs]
Call Trace:
afs_statfs+0xcc/0x180 [kafs]
? p9_client_statfs+0x9e/0x110 [9pnet]
? _cond_resched+0x19/0x40
statfs_by_dentry+0x6d/0x90
vfs_statfs+0x1b/0xc0
user_statfs+0x4b/0x80
SYSC_statfs+0x15/0x30
SyS_statfs+0xe/0x10
entry_SYSCALL_64_fastpath+0x20/0x83
Fixes: d2ddc776a4 ("afs: Overhaul volume and server record caching and fileserver rotation")
Reported-by: Marc Dionne <marc.dionne@auristor.com>
Signed-off-by: David Howells <dhowells@redhat.com>
cc: stable@vger.kernel.org
afs_alloc_volume() needs to release the cell ref it obtained in the case of
an error. Fix this by adding an afs_put_cell() call into the error path.
This can triggered when a lookup for a cell in a dynamic root or an
autocell mount returns an error whilst trying to look up the server (such
as ENOMEDIUM). This results in an assertion failure oops when the module
is unloaded due to outstanding refs on a cell record.
Fixes: d2ddc776a4 ("afs: Overhaul volume and server record caching and fileserver rotation")
Signed-off-by: David Howells <dhowells@redhat.com>
cc: stable@vger.kernel.org
There doesn't seem to be any need to have the INIT_SIGNALS and INIT_SIGHAND
macros, so expand them in their single places of use and remove them.
Signed-off-by: David Howells <dhowells@redhat.com>
Tested-by: Tony Luck <tony.luck@intel.com>
Tested-by: Will Deacon <will.deacon@arm.com> (arm64)
Tested-by: Palmer Dabbelt <palmer@sifive.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Construct the init thread stack in the linker script rather than doing it
by means of a union so that ia64's init_task.c can be got rid of.
The following symbols are then made available from INIT_TASK_DATA() linker
script macro:
init_thread_union
init_stack
INIT_TASK_DATA() also expands the region to THREAD_SIZE to accommodate the
size of the init stack. init_thread_union is given its own section so that
it can be placed into the stack space in the right order. I'm assuming
that the ia64 ordering is correct and that the task_struct is first and the
thread_info second.
Signed-off-by: David Howells <dhowells@redhat.com>
Tested-by: Tony Luck <tony.luck@intel.com>
Tested-by: Will Deacon <will.deacon@arm.com> (arm64)
Tested-by: Palmer Dabbelt <palmer@sifive.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
afs_write_end() is missing page unlock and put if afs_fill_page() fails.
Reported-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: David Howells <dhowells@redhat.com>
Repeating creation and deletion of a file on an afs mount will run the box
out of memory, e.g.:
dd if=/dev/zero of=/afs/scratch/m0 bs=$((1024*1024)) count=512
rm /afs/scratch/m0
The problem seems to be that it's not properly decrementing the nlink count
so that the inode can be scrapped.
Note that this doesn't fix local creation followed by remote deletion.
That's harder to handle and will require a separate patch as we're not told
that the file has been deleted - only that the directory has changed.
Reported-by: Marc Dionne <marc.dionne@auristor.com>
Signed-off-by: David Howells <dhowells@redhat.com>
Fix the default for fscache_maybe_release_page() for when the cookie isn't
valid or the page isn't cached. It mustn't return false as that indicates
the page cannot yet be freed.
The problem with the default is that if, say, there's no cache, but a
network filesystem's pages are using up almost all the available memory, a
system can OOM because the filesystem ->releasepage() op will not allow
them to be released as fscache_maybe_release_page() incorrectly prevents
it.
This can be tested by writing a sequence of 512MiB files to an AFS mount.
It does not affect NFS or CIFS because both of those wrap the call in a
check of PG_fscache and it shouldn't bother Ceph as that only has
PG_private set whilst writeback is in progress. This might be an issue for
9P, however.
Note that the pages aren't entirely stuck. Removing a file or unmounting
will clear things because that uses ->invalidatepage() instead.
Fixes: 201a15428b ("FS-Cache: Handle pages pending storage that get evicted under OOM conditions")
Reported-by: Marc Dionne <marc.dionne@auristor.com>
Signed-off-by: David Howells <dhowells@redhat.com>
Reviewed-by: Jeff Layton <jlayton@redhat.com>
Acked-by: Al Viro <viro@zeniv.linux.org.uk>
Tested-by: Marc Dionne <marc.dionne@auristor.com>
cc: stable@vger.kernel.org # 2.6.32+
In rxrpc_release_sock() there may be no rx->local value to access, so we
can't unconditionally follow it to the rxrpc network namespace information
to poke the connection reapers.
Instead, use the socket's namespace pointer to find the namespace.
This unfixed code causes the following static checker warning:
net/rxrpc/af_rxrpc.c:898 rxrpc_release_sock()
error: we previously assumed 'rx->local' could be null (see line 887)
Fixes: 3d18cbb7fd ("rxrpc: Fix conn expiry timers")
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When an AFS inode is allocated by afs_alloc_inode(), the allocated
afs_vnode struct isn't necessarily reset from the last time it was used as
an inode because the slab constructor is only invoked once when the memory
is obtained from the page allocator.
This means that information can leak from one inode to the next because
we're not calling kmem_cache_zalloc(). Some of the information isn't
reset, in particular the permit cache pointer.
Bring the clearances up to date.
Signed-off-by: David Howells <dhowells@redhat.com>
Tested-by: Marc Dionne <marc.dionne@auristor.com>
Fix four refcount bugs in afs_cache_permit():
(1) When checking the result of the kzalloc(), we can't just return, but
must put 'permits'.
(2) We shouldn't put permits immediately after hashing a new permit as we
need to keep the pointer stable so that we can check to see if
vnode->permit_cache has changed before we decide whether to assign to
it.
(3) 'permits' is being put twice.
(4) We need to put either the replacement or the thing replaced after the
assignment to vnode->permit_cache.
Without this, lots of the following are seen:
Kernel BUG at ffffffffa039857b [verbose debug info unavailable]
------------[ cut here ]------------
Kernel BUG at ffffffffa039858a [verbose debug info unavailable]
------------[ cut here ]------------
The addresses are in the .text..refcount section of the kafs.ko module.
Following the relocation records for the __ex_table section shows one to be
due to the decrement in afs_put_permits() and the other to be key_get() in
afs_cache_permit().
Occasionally, the following is seen:
refcount_t overflow at afs_cache_permit+0x57d/0x5c0 [kafs] in cc1[562], uid/euid: 0/0
WARNING: CPU: 0 PID: 562 at kernel/panic.c:657 refcount_error_report+0x9c/0xac
...
Reported-by: Marc Dionne <marc.dionne@auristor.com>
Signed-off-by: David Howells <dhowells@redhat.com>
Tested-by: Marc Dionne <marc.dionne@auristor.com>
Repeat terminal ACKs and now terminal ACKs are now generated from the
connection event processor rather from call handling as this allows us to
discard client call structures as soon as possible and free up the channel
for a follow on call.
However, in ACKs so generated, the additional information trailer is
malformed because the padding that's meant to be in the middle isn't
included in what's transmitted.
Fix it so that the 3 bytes of padding are included in the transmission.
Further, the trailer is misaligned because of the padding, so assigment to
the u16 and u32 fields inside it might cause problems on some arches, so
fix this by breaking the padding and the trailer out of the packed struct.
(This also deals with potential compiler weirdies where some of the nested
structs are packed and some aren't).
The symptoms can be seen in wireshark as terminal DUPLICATE or IDLE ACK
packets in which the Max MTU, Interface MTU and rwind fields have weird
values and the Max Packets field is apparently missing.
Reported-by: Jeffrey Altman <jaltman@auristor.com>
Signed-off-by: David Howells <dhowells@redhat.com>
afs_mkdir(), afs_create(), afs_link() and afs_symlink() all need to drop
the target dentry if a signal causes the operation to be killed immediately
before we try to contact the server.
Signed-off-by: David Howells <dhowells@redhat.com>
Fix some of dentry handling in AFS directory ops:
(1) Do d_drop() on the new_dentry before assigning a new inode to it in
afs_vnode_new_inode(). It's fine to do this before calling afs_iget()
because the operation has taken place on the server.
(2) Replace d_instantiate()/d_rehash() with d_add().
(3) Don't d_drop() the new_dentry in afs_rename() on error.
Also fix afs_link() and afs_rename() to call key_put() on all error paths
where the key is taken.
Signed-off-by: David Howells <dhowells@redhat.com>
Make afs_write_begin() wait for a page that's marked PG_writeback because:
(1) We need to avoid interference with the data being stored so that the
data on the server ends up in a defined state.
(2) page->private is used to track the window of dirty data within a page,
but it's also used by the storage code to track what's being written,
being cleared by the completion notification. Ownership can't be
relinquished by the storage code until completion because it a store
fails, the data must be remarked dirty.
Tracing shows something like the following (edited):
x86_64-linux-gn-15940 [1] afs_page_dirty: vn=ffff8800bef33800 9c75 begin 0-125
kworker/u8:3-114 [2] afs_page_dirty: vn=ffff8800bef33800 9c75 store+ 0-125
x86_64-linux-gn-15940 [1] afs_page_dirty: vn=ffff8800bef33800 9c75 begin 0-2052
kworker/u8:3-114 [2] afs_page_dirty: vn=ffff8800bef33800 9c75 clear 0-2052
kworker/u8:3-114 [2] afs_page_dirty: vn=ffff8800bef33800 9c75 store 0-0
kworker/u8:3-114 [2] afs_page_dirty: vn=ffff8800bef33800 9c75 WARN 0-0
The clear (completion) corresponding to the store+ (store continuation from
a previous page) happens between the second begin (afs_write_begin) and the
store corresponding to that. This results in the second store not seeing
any data to write back, leading to the following warning:
WARNING: CPU: 2 PID: 114 at ../fs/afs/write.c:403 afs_write_back_from_locked_page+0x19d/0x76c [kafs]
Modules linked in: kafs(E)
CPU: 2 PID: 114 Comm: kworker/u8:3 Tainted: G E 4.14.0-fscache+ #242
Hardware name: ASUS All Series/H97-PLUS, BIOS 2306 10/09/2014
Workqueue: writeback wb_workfn (flush-afs-2)
task: ffff8800cad72600 task.stack: ffff8800cad44000
RIP: 0010:afs_write_back_from_locked_page+0x19d/0x76c [kafs]
RSP: 0018:ffff8800cad47aa0 EFLAGS: 00010246
RAX: 0000000000000001 RBX: ffff8800bef33a20 RCX: 0000000000000000
RDX: 000000000000000f RSI: ffffffff81c5d0e0 RDI: ffff8800cad72e78
RBP: ffff8800d31ea1e8 R08: ffff8800c1358000 R09: ffff8800ca00e400
R10: ffff8800cad47a38 R11: ffff8800c5d9e400 R12: 0000000000000000
R13: ffffea0002d9df00 R14: ffffffffa0023c1c R15: 0000000000007fdf
FS: 0000000000000000(0000) GS:ffff8800ca700000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00007f85ac6c4000 CR3: 0000000001c10001 CR4: 00000000001606e0
Call Trace:
? clear_page_dirty_for_io+0x23a/0x267
afs_writepages_region+0x1be/0x286 [kafs]
afs_writepages+0x60/0x127 [kafs]
do_writepages+0x36/0x70
__writeback_single_inode+0x12f/0x635
writeback_sb_inodes+0x2cc/0x452
__writeback_inodes_wb+0x68/0x9f
wb_writeback+0x208/0x470
? wb_workfn+0x22b/0x565
wb_workfn+0x22b/0x565
? worker_thread+0x230/0x2ac
process_one_work+0x2cc/0x517
? worker_thread+0x230/0x2ac
worker_thread+0x1d4/0x2ac
? rescuer_thread+0x29b/0x29b
kthread+0x15d/0x165
? kthread_create_on_node+0x3f/0x3f
? call_usermodehelper_exec_async+0x118/0x11f
ret_from_fork+0x24/0x30
Signed-off-by: David Howells <dhowells@redhat.com>
Fix the rxrpc connection expiry timers so that connections for closed
AF_RXRPC sockets get deleted in a more timely fashion, freeing up the
transport UDP port much more quickly.
(1) Replace the delayed work items with work items plus timers so that
timer_reduce() can be used to shorten them and so that the timer
doesn't requeue the work item if the net namespace is dead.
(2) Don't use queue_delayed_work() as that won't alter the timeout if the
timer is already running.
(3) Don't rearm the timers if the network namespace is dead.
Signed-off-by: David Howells <dhowells@redhat.com>
RxRPC service endpoints expire like they're supposed to by the following
means:
(1) Mark dead rxrpc_net structs (with ->live) rather than twiddling the
global service conn timeout, otherwise the first rxrpc_net struct to
die will cause connections on all others to expire immediately from
then on.
(2) Mark local service endpoints for which the socket has been closed
(->service_closed) so that the expiration timeout can be much
shortened for service and client connections going through that
endpoint.
(3) rxrpc_put_service_conn() needs to schedule the reaper when the usage
count reaches 1, not 0, as idle conns have a 1 count.
(4) The accumulator for the earliest time we might want to schedule for
should be initialised to jiffies + MAX_JIFFY_OFFSET, not ULONG_MAX as
the comparison functions use signed arithmetic.
(5) Simplify the expiration handling, adding the expiration value to the
idle timestamp each time rather than keeping track of the time in the
past before which the idle timestamp must go to be expired. This is
much easier to read.
(6) Ignore the timeouts if the net namespace is dead.
(7) Restart the service reaper work item rather the client reaper.
Signed-off-by: David Howells <dhowells@redhat.com>
We need to transmit a packet every so often to act as a keepalive for the
peer (which has a timeout from the last time it received a packet) and also
to prevent any intervening firewalls from closing the route.
Do this by resetting a timer every time we transmit a packet. If the timer
ever expires, we transmit a PING ACK packet and thereby also elicit a PING
RESPONSE ACK from the other side - which prevents our last-rx timeout from
expiring.
The timer is set to 1/6 of the last-rx timeout so that we can detect the
other side going away if it misses 6 replies in a row.
This is particularly necessary for servers where the processing of the
service function may take a significant amount of time.
Signed-off-by: David Howells <dhowells@redhat.com>
Add an extra timeout that is set/updated when we send a DATA packet that
has the request-ack flag set. This allows us to detect if we don't get an
ACK in response to the latest flagged packet.
The ACK packet is adjudged to have been lost if it doesn't turn up within
2*RTT of the transmission.
If the timeout occurs, we schedule the sending of a PING ACK to find out
the state of the other side. If a new DATA packet is ready to go sooner,
we cancel the sending of the ping and set the request-ack flag on that
instead.
If we get back a PING-RESPONSE ACK that indicates a lower tx_top than what
we had at the time of the ping transmission, we adjudge all the DATA
packets sent between the response tx_top and the ping-time tx_top to have
been lost and retransmit immediately.
Rather than sending a PING ACK, we could just pick a DATA packet and
speculatively retransmit that with request-ack set. It should result in
either a REQUESTED ACK or a DUPLICATE ACK which we can then use in lieu the
a PING-RESPONSE ACK mentioned above.
Signed-off-by: David Howells <dhowells@redhat.com>
Express protocol timeouts for data retransmission and deferred ack
generation in terms on RTT rather than specified timeouts once we have
sufficient RTT samples.
For the moment, this requires just one RTT sample to be able to use this
for ack deferral and two for data retransmission.
The data retransmission timeout is set at RTT*1.5 and the ACK deferral
timeout is set at RTT.
Note that the calculated timeout is limited to a minimum of 4ns to make
sure it doesn't happen too quickly.
Signed-off-by: David Howells <dhowells@redhat.com>
Don't transmit a DELAY ACK immediately on proposal when the Rx window is
rotated, but rather defer it to the work function. This means that we have
a chance to queue/consume more received packets before we actually send the
DELAY ACK, or even cancel it entirely, thereby reducing the number of
packets transmitted.
We do, however, want to continue sending other types of packet immediately,
particularly REQUESTED ACKs, as they may be used for RTT calculation by the
other side.
Signed-off-by: David Howells <dhowells@redhat.com>
Fix the rxrpc call expiration timeouts and make them settable from
userspace. By analogy with other rx implementations, there should be three
timeouts:
(1) "Normal timeout"
This is set for all calls and is triggered if we haven't received any
packets from the peer in a while. It is measured from the last time
we received any packet on that call. This is not reset by any
connection packets (such as CHALLENGE/RESPONSE packets).
If a service operation takes a long time, the server should generate
PING ACKs at a duration that's substantially less than the normal
timeout so is to keep both sides alive. This is set at 1/6 of normal
timeout.
(2) "Idle timeout"
This is set only for a service call and is triggered if we stop
receiving the DATA packets that comprise the request data. It is
measured from the last time we received a DATA packet.
(3) "Hard timeout"
This can be set for a call and specified the maximum lifetime of that
call. It should not be specified by default. Some operations (such
as volume transfer) take a long time.
Allow userspace to set/change the timeouts on a call with sendmsg, using a
control message:
RXRPC_SET_CALL_TIMEOUTS
The data to the message is a number of 32-bit words, not all of which need
be given:
u32 hard_timeout; /* sec from first packet */
u32 idle_timeout; /* msec from packet Rx */
u32 normal_timeout; /* msec from data Rx */
This can be set in combination with any other sendmsg() that affects a
call.
Signed-off-by: David Howells <dhowells@redhat.com>
When rxrpc_sendmsg() parses the control message buffer, it places the
parameters extracted into a structure, but lumps together call parameters
(such as user call ID) with operation parameters (such as whether to send
data, send an abort or accept a call).
Split the call parameters out into their own structure, a copy of which is
then embedded in the operation parameters struct.
The call parameters struct is then passed down into the places that need it
instead of passing the individual parameters. This allows for extra call
parameters to be added.
Signed-off-by: David Howells <dhowells@redhat.com>
Delay terminal ACK transmission on a client call by deferring it to the
connection processor. This allows it to be skipped if we can send the next
call instead, the first DATA packet of which will implicitly ack this call.
Signed-off-by: David Howells <dhowells@redhat.com>
Don't set upgrade by default when creating a call from sendmsg(). This is
a holdover from when I was testing the code.
Signed-off-by: David Howells <dhowells@redhat.com>
The caller of rxrpc_accept_call() must release the lock on call->user_mutex
returned by that function.
Signed-off-by: David Howells <dhowells@redhat.com>
Fix the AFS file locking whereby the use of the big kernel lock (which
could be slept with) was replaced by a spinlock (which couldn't). The
problem is that the AFS code was doing stuff inside the critical section
that might call schedule(), so this is a broken transformation.
Fix this by the following means:
(1) Use a state machine with a proper state that can only be changed under
the spinlock rather than using a collection of bit flags.
(2) Cache the key used for the lock and the lock type in the afs_vnode
struct so that the manager work function doesn't have to refer to a
file_lock struct that's been dequeued. This makes signal handling
safer.
(4) Move the unlock from afs_do_unlk() to afs_fl_release_private() which
means that unlock is achieved in other circumstances too.
(5) Unlock the file on the server before taking the next conflicting lock.
Also change:
(1) Check the permits on a file before actually trying the lock.
(2) fsync the file before effecting an explicit unlock operation. We
don't fsync if the lock is erased otherwise as we might not be in a
context where we can actually do that.
Further fixes:
(1) Fixed-fileserver address rotation is made to work. It's only used by
the locking functions, so couldn't be tested before.
Fixes: 72f98e7255 ("locks: turn lock_flocks into a spinlock")
Signed-off-by: David Howells <dhowells@redhat.com>
cc: jlayton@redhat.com
Set the module licence to prevent the kernel from being tainted if loaded
as a module.
Reported-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: David Howells <dhowells@redhat.com>
Protect call->state changes against the call being prematurely terminated
due to a signal.
What can happen is that a signal causes afs_wait_for_call_to_complete() to
abort an afs_call because it's not yet complete whilst afs_deliver_to_call()
is delivering data to that call.
If the data delivery causes the state to change, this may overwrite the state
of the afs_call, making it not-yet-complete again - but no further
notifications will be forthcoming from AF_RXRPC as the rxrpc call has been
aborted and completed, so kAFS will just hang in various places waiting for
that call or on page bits that need clearing by that call.
A tracepoint to monitor call state changes is also provided.
Signed-off-by: David Howells <dhowells@redhat.com>
Get rid of the afs_writeback record that kAFS is using to match keys with
writes made by that key.
Instead, keep a list of keys that have a file open for writing and/or
sync'ing and iterate through those.
Signed-off-by: David Howells <dhowells@redhat.com>
Introduce a file-private data record for kAFS and put the key into it
rather than storing the key in file->private_data.
Signed-off-by: David Howells <dhowells@redhat.com>
Because parsing of the directory wasn't being done under any sort of lock,
the pages holding the directory content can get invalidated whilst the
parsing is ongoing.
Further, the directory page check function gets called outside of the page
lock, so if the page gets cleared or updated, this may return reports of
bad magic numbers in the directory page.
Also, the directory may change size whilst checking and parsing are
ongoing, so more care needs to be taken here.
Fix this by:
(1) Perform the page check from the page filling function before we set
PageUptodate and drop the page lock.
(2) Check for the file having shrunk and the page having been abandoned
before checking the page contents.
(3) Lock the page whilst parsing it for the directory iterator.
Whilst we're at it, add a tracepoint to report check failure.
Signed-off-by: David Howells <dhowells@redhat.com>
Add a pair of tracepoints to log the sending of pages for an FS.StoreData
or FS.StoreData64 operation.
Tracepoint afs_send_pages notes each set of pages added to the operation.
There may be several of these per operation as we get up at most 8
contiguous pages in one go because the bvec we're using is on the stack.
Tracepoint afs_sent_pages notes the end of adding data from a whole run of
pages to the operation and the completion of the request phase.
Signed-off-by: David Howells <dhowells@redhat.com>
Add tracepoints to trace the initiation and completion of client calls
within the kafs filesystem.
The afs_make_vl_call tracepoint watches calls to the volume location
database server.
The afs_make_fs_call tracepoint watches calls to the file server.
The afs_call_done tracepoint watches for call completion.
Signed-off-by: David Howells <dhowells@redhat.com>
The documentation that describes the #-prefix and the %-prefix used when
specifying the source to mount is has the descriptions the wrong way
round. Switch them over.
Reported-by: Marc Dionne <marc.dionne@auristor.com>
Signed-off-by: David Howells <dhowells@redhat.com>
Fix the total-length calculation in afs_make_call() when the operation
being dispatched has data from a series of pages attached.
Despite the patched code looking like that it should reduce mathematically
to the current code, it doesn't because the 32-bit unsigned arithmetic
being used to calculate the page-offset-difference doesn't correctly extend
to a 64-bit value when the result is effectively negative.
Without this, some FS.StoreData operations that span multiple pages fail,
reporting too little or too much data.
Signed-off-by: David Howells <dhowells@redhat.com>
Only progress the AFS call state at the end of Tx phase from the callback
passed to rxrpc_kernel_send_data() rather than setting it before the last
data send call.
Signed-off-by: David Howells <dhowells@redhat.com>
YFS VL servers offer an upgraded Volume Location service that can return
IPv6 addresses to fileservers and volume servers in addition to IPv4
addresses using the YFSVL.GetEndpoints operation which we should use if
it's available.
To this end:
(1) Make rxrpc_kernel_recv_data() return the call's current service ID so
that the caller can detect service upgrade and see what the service
was upgraded to.
(2) When we see a VL server address we haven't seen before, send a
VL.GetCapabilities operation to it with the service upgrade bit set.
If we get an upgrade to the YFS VL service, change the service ID in
the address list for that address to use the upgraded service and set
a flag to note that this appears to be a YFS-compatible server.
(3) If, when a server's addresses are being looked up, we note that we
previously detected a YFS-compatible server, then send the
YFSVL.GetEndpoints operation rather than VL.GetAddrsU.
(4) Build a fileserver address list from the reply of YFSVL.GetEndpoints,
including both IPv4 and IPv6 addresses. Volume server addresses are
discarded.
(5) The address list is sorted by address and port now, instead of just
address. This allows multiple servers on the same host sitting on
different ports.
Signed-off-by: David Howells <dhowells@redhat.com>
The current code assumes that volumes and servers are per-cell and are
never shared, but this is not enforced, and, indeed, public cells do exist
that are aliases of each other. Further, an organisation can, say, set up
a public cell and a private cell with overlapping, but not identical, sets
of servers. The difference is purely in the database attached to the VL
servers.
The current code will malfunction if it sees a server in two cells as it
assumes global address -> server record mappings and that each server is in
just one cell.
Further, each server may have multiple addresses - and may have addresses
of different families (IPv4 and IPv6, say).
To this end, the following structural changes are made:
(1) Server record management is overhauled:
(a) Server records are made independent of cell. The namespace keeps
track of them, volume records have lists of them and each vnode
has a server on which its callback interest currently resides.
(b) The cell record no longer keeps a list of servers known to be in
that cell.
(c) The server records are now kept in a flat list because there's no
single address to sort on.
(d) Server records are now keyed by their UUID within the namespace.
(e) The addresses for a server are obtained with the VL.GetAddrsU
rather than with VL.GetEntryByName, using the server's UUID as a
parameter.
(f) Cached server records are garbage collected after a period of
non-use and are counted out of existence before purging is allowed
to complete. This protects the work functions against rmmod.
(g) The servers list is now in /proc/fs/afs/servers.
(2) Volume record management is overhauled:
(a) An RCU-replaceable server list is introduced. This tracks both
servers and their coresponding callback interests.
(b) The superblock is now keyed on cell record and numeric volume ID.
(c) The volume record is now tied to the superblock which mounts it,
and is activated when mounted and deactivated when unmounted.
This makes it easier to handle the cache cookie without causing a
double-use in fscache.
(d) The volume record is loaded from the VLDB using VL.GetEntryByNameU
to get the server UUID list.
(e) The volume name is updated if it is seen to have changed when the
volume is updated (the update is keyed on the volume ID).
(3) The vlocation record is got rid of and VLDB records are no longer
cached. Sufficient information is stored in the volume record, though
an update to a volume record is now no longer shared between related
volumes (volumes come in bundles of three: R/W, R/O and backup).
and the following procedural changes are made:
(1) The fileserver cursor introduced previously is now fleshed out and
used to iterate over fileservers and their addresses.
(2) Volume status is checked during iteration, and the server list is
replaced if a change is detected.
(3) Server status is checked during iteration, and the address list is
replaced if a change is detected.
(4) The abort code is saved into the address list cursor and -ECONNABORTED
returned in afs_make_call() if a remote abort happened rather than
translating the abort into an error message. This allows actions to
be taken depending on the abort code more easily.
(a) If a VMOVED abort is seen then this is handled by rechecking the
volume and restarting the iteration.
(b) If a VBUSY, VRESTARTING or VSALVAGING abort is seen then this is
handled by sleeping for a short period and retrying and/or trying
other servers that might serve that volume. A message is also
displayed once until the condition has cleared.
(c) If a VOFFLINE abort is seen, then this is handled as VBUSY for the
moment.
(d) If a VNOVOL abort is seen, the volume is rechecked in the VLDB to
see if it has been deleted; if not, the fileserver is probably
indicating that the volume couldn't be attached and needs
salvaging.
(e) If statfs() sees one of these aborts, it does not sleep, but
rather returns an error, so as not to block the umount program.
(5) The fileserver iteration functions in vnode.c are now merged into
their callers and more heavily macroised around the cursor. vnode.c
is removed.
(6) Operations on a particular vnode are serialised on that vnode because
the server will lock that vnode whilst it operates on it, so a second
op sent will just have to wait.
(7) Fileservers are probed with FS.GetCapabilities before being used.
This is where service upgrade will be done.
(8) A callback interest on a fileserver is set up before an FS operation
is performed and passed through to afs_make_call() so that it can be
set on the vnode if the operation returns a callback. The callback
interest is passed through to afs_iget() also so that it can be set
there too.
In general, record updating is done on an as-needed basis when we try to
access servers, volumes or vnodes rather than offloading it to work items
and special threads.
Notes:
(1) Pre AFS-3.4 servers are no longer supported, though this can be added
back if necessary (AFS-3.4 was released in 1998).
(2) VBUSY is retried forever for the moment at intervals of 1s.
(3) /proc/fs/afs/<cell>/servers no longer exists.
Signed-off-by: David Howells <dhowells@redhat.com>
Add an RCU replaceable address list structure to hold a list of server
addresses. The list also holds the
To this end:
(1) A cell's VL server address list can be loaded directly via insmod or
echo to /proc/fs/afs/cells or dynamically from a DNS query for AFSDB
or SRV records.
(2) Anyone wanting to use a cell's VL server address must wait until the
cell record comes online and has tried to obtain some addresses.
(3) An FS server's address list, for the moment, has a single entry that
is the key to the server list. This will change in the future when a
server is instead keyed on its UUID and the VL.GetAddrsU operation is
used.
(4) An 'address cursor' concept is introduced to handle iteration through
the address list. This is passed to the afs_make_call() as, in the
future, stuff (such as abort code) that doesn't outlast the call will
be returned in it.
In the future, we might want to annotate the list with information about
how each address fares. We might then want to propagate such annotations
over address list replacement.
Whilst we're at it, we allow IPv6 addresses to be specified in
colon-delimited lists by enclosing them in square brackets.
Signed-off-by: David Howells <dhowells@redhat.com>
Overhaul the way that the in-kernel AFS client keeps track of cells in the
following manner:
(1) Cells are now held in an rbtree to make walking them quicker and RCU
managed (though this is probably overkill).
(2) Cells now have a manager work item that:
(A) Looks after fetching and refreshing the VL server list.
(B) Manages cell record lifetime, including initialising and
destruction.
(B) Manages cell record caching whereby threads are kept around for a
certain time after last use and then destroyed.
(C) Manages the FS-Cache index cookie for a cell. It is not permitted
for a cookie to be in use twice, so we have to be careful to not
allow a new cell record to exist at the same time as an old record
of the same name.
(3) Each AFS network namespace is given a manager work item that manages
the cells within it, maintaining a single timer to prod cells into
updating their DNS records.
This uses the reduce_timer() facility to make the timer expire at the
soonest timed event that needs happening.
(4) When a module is being unloaded, cells and cell managers are now
counted out using dec_after_work() to make sure the module text is
pinned until after the data structures have been cleaned up.
(5) Each cell's VL server list is now protected by a seqlock rather than a
semaphore.
Signed-off-by: David Howells <dhowells@redhat.com>
Overhaul permit caching in AFS by making it per-vnode and sharing permit
lists where possible.
When most of the fileserver operations are called, they return a status
structure indicating the (revised) details of the vnode or vnodes involved
in the operation. This includes the access mark derived from the ACL
(named CallerAccess in the protocol definition file). This is cacheable
and if the ACL changes, the server will tell us that it is breaking the
callback promise, at which point we can discard the currently cached
permits.
With this patch, the afs_permits structure has, at the end, an array of
{ key, CallerAccess } elements, sorted by key pointer. This is then cached
in a hash table so that it can be shared between vnodes with the same
access permits.
Permit lists can only be shared if they contain the exact same set of
key->CallerAccess mappings.
Note that that table is global rather than being per-net_ns. If the keys
in a permit list cross net_ns boundaries, there is no problem sharing the
cached permits, since the permits are just integer masks.
Since permit lists pin keys, the permit cache also makes it easier for a
future patch to find all occurrences of a key and remove them by means of
setting the afs_permits::invalidated flag and then clearing the appropriate
key pointer. In such an event, memory barriers will need adding.
Lastly, the permit caching is skipped if the server has sent either a
vnode-specific or an entire-server callback since the start of the
operation.
Signed-off-by: David Howells <dhowells@redhat.com>
Overhaul the AFS callback handling by the following means:
(1) Don't give up callback promises on vnodes that we are no longer using,
rather let them just expire on the server or let the server break
them. This is actually more efficient for the server as the callback
lookup is expensive if there are lots of extant callbacks.
(2) Only give up the callback promises we have from a server when the
server record is destroyed. Then we can just give up *all* the
callback promises on it in one go.
(3) Servers can end up being shared between cells if cells are aliased, so
don't add all the vnodes being backed by a particular server into a
big FID-indexed tree on that server as there may be duplicates.
Instead have each volume instance (~= superblock) register an interest
in a server as it starts to make use of it and use this to allow the
processor for callbacks from the server to find the superblock and
thence the inode corresponding to the FID being broken by means of
ilookup_nowait().
(4) Rather than iterating over the entire callback list when a mass-break
comes in from the server, maintain a counter of mass-breaks in
afs_server (cb_seq) and make afs_validate() check it against the copy
in afs_vnode.
It would be nice not to have to take a read_lock whilst doing this,
but that's tricky without using RCU.
(5) Save a ref on the fileserver we're using for a call in the afs_call
struct so that we can access its cb_s_break during call decoding.
(6) Write-lock around callback and status storage in a vnode and read-lock
around getattr so that we don't see the status mid-update.
This has the following consequences:
(1) Data invalidation isn't seen until someone calls afs_validate() on a
vnode. Unfortunately, we need to use a key to query the server, but
getting one from a background thread is tricky without caching loads
of keys all over the place.
(2) Mass invalidation isn't seen until someone calls afs_validate().
(3) Callback breaking is going to hit the inode_hash_lock quite a bit.
Could this be replaced with rcu_read_lock() since inodes are destroyed
under RCU conditions.
Signed-off-by: David Howells <dhowells@redhat.com>
Rename the server member of struct afs_call to cm_server as we're only
going to be using it for incoming calls for the Cache Manager service.
This makes it easier to differentiate from the pointer to the target server
for the client, which will point to a different structure to allow for
callback handling.
Signed-off-by: David Howells <dhowells@redhat.com>
In AFS's encoding of a UUID, the eight 'char' fields are all signed, so
represent them with __s8 rather than __u8. This makes the compiler
sign-extend them correctly when XDR-encoding them.
Signed-off-by: David Howells <dhowells@redhat.com>
The handler for the CB.ProbeUuid operation in the cache manager is
implemented, but isn't listed in the switch-statement of operation
selection, so won't be used. Fix this by adding it.
Signed-off-by: David Howells <dhowells@redhat.com>
If call->ret_reply0 is set, return call->reply[0] on success. Change the
return type of afs_make_call() to long so that this can be passed back
without bit loss and then cast to a pointer if required.
Signed-off-by: David Howells <dhowells@redhat.com>
The AFS abort code space is shared across all services, so there's no need
for separate abort_to_error translators for each service.
Consolidate them into a single function and remove the function pointers
for them.
Signed-off-by: David Howells <dhowells@redhat.com>
Allow VL server specifications to be given IPv6 addresses as well as IPv4
addresses, for example as:
echo add foo.org 1111:2222:3333:0:4444:5555:6666:7777 >/proc/fs/afs/cells
Note that ':' is the expected separator for separating IPv4 addresses, but
if a ',' is detected or no '.' is detected in the string, the delimiter is
switched to ','.
This also works with DNS AFSDB or SRV record strings fetched by upcall from
userspace.
Signed-off-by: David Howells <dhowells@redhat.com>
Keep and pass sockaddr_rxrpc addresses around rather than keeping and
passing in_addr addresses to allow for the use of IPv6 and non-standard
port numbers in future.
This also allows the port and service_id fields to be removed from the
afs_call struct.
Signed-off-by: David Howells <dhowells@redhat.com>
Update the cache index structure in the following ways:
(1) Don't use the volume name followed by the volume type as levels in the
cache index. Volumes can be renamed. Use the volume ID instead.
(2) Don't store the VLDB data for a volume in the tree. If the volume
database should be cached locally, then it should be done in a separate
tree.
(3) Expand the volume ID stored in the cache to 64 bits.
(4) Expand the file/vnode ID stored in the cache to 96 bits.
(5) Increment the cache structure version number to 1.
Signed-off-by: David Howells <dhowells@redhat.com>
Add some protocol definitions, including max field lengths, flag defs, an
XDR-encoded UUID def, more VL operation IDs and more fileserver abort
codes.
Signed-off-by: David Howells <dhowells@redhat.com>
Push the network namespace pointer to more places in AFS, including the
afs_server structure (which doesn't hold a ref on the netns).
In particular, afs_put_cell() now takes requires a net ns parameter so that
it can safely alter the netns after decrementing the cell usage count - the
cell will be deallocated by a background thread after being cached for a
period, which means that it's not safe to access it after reducing its
usage count.
Signed-off-by: David Howells <dhowells@redhat.com>
Keep a reference to the cell in the superblock info structure in addition
to the volume and net pointers. This will make it easier to clean up in a
future patch in which afs_put_volume() will need the cell pointer.
Whilst we're at it, make the cell and volume getting functions return a
pointer to the object got to make the call sites look neater.
Signed-off-by: David Howells <dhowells@redhat.com>
Fix server reaping and make sure it's all done before we start trying to
purge cells, given that servers currently pin cells.
Signed-off-by: David Howells <dhowells@redhat.com>
Close the rxrpc socket only after we've purged the server records (and also
cell and volume records which might refer to servers) so that we can give
up the callbacks on each server.
Signed-off-by: David Howells <dhowells@redhat.com>
Lay the groundwork for supporting network namespaces (netns) to the AFS
filesystem by moving various global features to a network-namespace struct
(afs_net) and providing an instance of this as a temporary global variable
that everything uses via accessor functions for the moment.
The following changes have been made:
(1) Store the netns in the superblock info. This will be obtained from
the mounter's nsproxy on a manual mount and inherited from the parent
superblock on an automount.
(2) The cell list is made per-netns. It can be viewed through
/proc/net/afs/cells and also be modified by writing commands to that
file.
(3) The local workstation cell is set per-ns in /proc/net/afs/rootcell.
This is unset by default.
(4) The 'rootcell' module parameter, which sets a cell and VL server list
modifies the init net namespace, thereby allowing an AFS root fs to be
theoretically used.
(5) The volume location lists and the file lock manager are made
per-netns.
(6) The AF_RXRPC socket and associated I/O bits are made per-ns.
The various workqueues remain global for the moment.
Changes still to be made:
(1) /proc/fs/afs/ should be moved to /proc/net/afs/ and a symlink emplaced
from the old name.
(2) A per-netns subsys needs to be registered for AFS into which it can
store its per-netns data.
(3) Rather than the AF_RXRPC socket being opened on module init, it needs
to be opened on the creation of a superblock in that netns.
(4) The socket needs to be closed when the last superblock using it is
destroyed and all outstanding client calls on it have been completed.
This prevents a reference loop on the namespace.
(5) It is possible that several namespaces will want to use AFS, in which
case each one will need its own UDP port. These can either be set
through /proc/net/afs/cm_port or the kernel can pick one at random.
The init_ns gets 7001 by default.
Other issues that need resolving:
(1) The DNS keyring needs net-namespacing.
(2) Where do upcalls go (eg. DNS request-key upcall)?
(3) Need something like open_socket_in_file_ns() syscall so that AFS
command line tools attempting to operate on an AFS file/volume have
their RPC calls go to the right place.
Signed-off-by: David Howells <dhowells@redhat.com>
Make wait_on_atomic_t() pass the TASK_* mode onto its action function as an
extra argument and make it 'unsigned int throughout.
Also, consolidate a bunch of identical action functions into a default
function that can do the appropriate thing for the mode.
Also, change the argument name in the bit_wait*() function declarations to
reflect the fact that it's the mode and not the bit number.
[Peter Z gives this a grudging ACK, but thinks that the whole atomic_t wait
should be done differently, though he's not immediately sure as to how]
Signed-off-by: David Howells <dhowells@redhat.com>
Acked-by: Peter Zijlstra <peterz@infradead.org>
cc: Ingo Molnar <mingo@kernel.org>
Add a function, similar to mod_timer(), that will start a timer if it isn't
running and will modify it if it is running and has an expiry time longer
than the new time. If the timer is running with an expiry time that's the
same or sooner, no change is made.
The function looks like:
int timer_reduce(struct timer_list *timer, unsigned long expires);
This can be used by code such as networking code to make it easier to share
a timer for multiple timeouts. For instance, in upcoming AF_RXRPC code,
the rxrpc_call struct will maintain a number of timeouts:
unsigned long ack_at;
unsigned long resend_at;
unsigned long ping_at;
unsigned long expect_rx_by;
unsigned long expect_req_by;
unsigned long expect_term_by;
each of which is set independently of the others. With timer reduction
available, when the code needs to set one of the timeouts, it only needs to
look at that timeout and then call timer_reduce() to modify the timer,
starting it or bringing it forward if necessary. There is no need to refer
to the other timeouts to see which is earliest and no need to take any lock
other than, potentially, the timer lock inside timer_reduce().
Note, that this does not protect against concurrent invocations of any of
the timer functions.
As an example, the expect_rx_by timeout above, which terminates a call if
we don't get a packet from the server within a certain time window, would
be set something like this:
unsigned long now = jiffies;
unsigned long expect_rx_by = now + packet_receive_timeout;
WRITE_ONCE(call->expect_rx_by, expect_rx_by);
timer_reduce(&call->timer, expect_rx_by);
The timer service code (which might, say, be in a work function) would then
check all the timeouts to see which, if any, had triggered, deal with
those:
t = READ_ONCE(call->ack_at);
if (time_after_eq(now, t)) {
cmpxchg(&call->ack_at, t, now + MAX_JIFFY_OFFSET);
set_bit(RXRPC_CALL_EV_ACK, &call->events);
}
and then restart the timer if necessary by finding the soonest timeout that
hasn't yet passed and then calling timer_reduce().
The disadvantage of doing things this way rather than comparing the timers
each time and calling mod_timer() is that you *will* take timer events
unless you can finish what you're doing and delete the timer in time.
The advantage of doing things this way is that you don't need to use a lock
to work out when the next timer should be set, other than the timer's own
lock - which you might not have to take.
[ tglx: Fixed weird formatting and adopted it to pending changes ]
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: keyrings@vger.kernel.org
Cc: linux-afs@lists.infradead.org
Link: https://lkml.kernel.org/r/151023090769.23050.1801643667223880753.stgit@warthog.procyon.org.uk
Fix call expiry handling in the following ways
(1) If all the request data from a client call is acked, don't send a
follow up IDLE ACK with firstPacket == 1 and previousPacket == 0 as
this appears to fool some servers into thinking everything has been
accepted.
(2) Never send an abort back to the server once it has ACK'd all the
request packets; rather just try to reuse the channel for the next
call. The first request DATA packet of the next call on the same
channel will implicitly ACK the entire reply of the dead call - even
if we haven't transmitted it yet.
(3) Don't send RX_CALL_TIMEOUT in an ABORT packet, librx uses abort codes
to pass local errors to the caller in addition to remote errors, and
this is meant to be local only.
The following also need to be addressed in future patches:
(4) Service calls should send PING ACKs as 'keep alives' if the server is
still processing the call.
(5) VERSION REPLY packets should be sent to the peers of service
connections to act as keep-alives. This is used to keep firewall
routes in place. The AFS CM should enable this.
Signed-off-by: David Howells <dhowells@redhat.com>
rxrpc_fill_out_ack() needs to be passed the connection pointer from its
caller rather than using call->conn as the call may be disconnected in
parallel with it, clearing call->conn, leading to:
BUG: unable to handle kernel NULL pointer dereference at 0000000000000010
IP: rxrpc_send_ack_packet+0x231/0x6a4
Signed-off-by: David Howells <dhowells@redhat.com>
Place a spinlock around the invocation of call->notify_rx() for a kernel
service call and lock again when ending the call and replace the
notification pointer with a pointer to a dummy function.
This is required because it's possible for rxrpc_notify_socket() to be
called after the call has been ended by the kernel service if called from
the asynchronous work function rxrpc_process_call().
However, rxrpc_notify_socket() currently only holds the RCU read lock when
invoking ->notify_rx(), which means that the afs_call struct would need to
be disposed of by call_rcu() rather than by kfree().
But we shouldn't see any notifications from a call after calling
rxrpc_kernel_end_call(), so a lock is required in rxrpc code.
Without this, we may see the call wait queue as having a corrupt spinlock:
BUG: spinlock bad magic on CPU#0, kworker/0:2/1612
general protection fault: 0000 [#1] SMP
...
Workqueue: krxrpcd rxrpc_process_call
task: ffff88040b83c400 task.stack: ffff88040adfc000
RIP: 0010:spin_bug+0x161/0x18f
RSP: 0018:ffff88040adffcc0 EFLAGS: 00010002
RAX: 0000000000000032 RBX: 6b6b6b6b6b6b6b6b RCX: ffffffff81ab16cf
RDX: ffff88041fa14c01 RSI: ffff88041fa0ccb8 RDI: ffff88041fa0ccb8
RBP: ffff88040adffcd8 R08: 00000000ffffffff R09: 00000000ffffffff
R10: ffff88040adffc60 R11: 000000000000022c R12: ffff88040aca2208
R13: ffffffff81a58114 R14: 0000000000000000 R15: 0000000000000000
....
Call Trace:
do_raw_spin_lock+0x1d/0x89
_raw_spin_lock_irqsave+0x3d/0x49
? __wake_up_common_lock+0x4c/0xa7
__wake_up_common_lock+0x4c/0xa7
? __lock_is_held+0x47/0x7a
__wake_up+0xe/0x10
afs_wake_up_call_waiter+0x11b/0x122 [kafs]
rxrpc_notify_socket+0x12b/0x258
rxrpc_process_call+0x18e/0x7d0
process_one_work+0x298/0x4de
? rescuer_thread+0x280/0x280
worker_thread+0x1d1/0x2ae
? rescuer_thread+0x280/0x280
kthread+0x12c/0x134
? kthread_create_on_node+0x3a/0x3a
ret_from_fork+0x27/0x40
In this case, note the corrupt data in EBX. The address of the offending
afs_call is in R12, plus the offset to the spinlock.
Signed-off-by: David Howells <dhowells@redhat.com>
This fixes CVE-2017-12193.
Fix a case in the assoc_array implementation in which a new leaf is
added that needs to go into a node that happens to be full, where the
existing leaves in that node cluster together at that level to the
exclusion of new leaf.
What needs to happen is that the existing leaves get moved out to a new
node, N1, at level + 1 and the existing node needs replacing with one,
N0, that has pointers to the new leaf and to N1.
The code that tries to do this gets this wrong in two ways:
(1) The pointer that should've pointed from N0 to N1 is set to point
recursively to N0 instead.
(2) The backpointer from N0 needs to be set correctly in the case N0 is
either the root node or reached through a shortcut.
Fix this by removing this path and using the split_node path instead,
which achieves the same end, but in a more general way (thanks to Eric
Biggers for spotting the redundancy).
The problem manifests itself as:
BUG: unable to handle kernel NULL pointer dereference at 0000000000000010
IP: assoc_array_apply_edit+0x59/0xe5
Fixes: 3cb989501c ("Add a generic associative array implementation.")
Reported-and-tested-by: WU Fan <u3536072@connect.hku.hk>
Signed-off-by: David Howells <dhowells@redhat.com>
Cc: stable@vger.kernel.org [v3.13-rc1+]
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Don't release call mutex at the end of rxrpc_kernel_begin_call() if the
call pointer actually holds an error value.
Fixes: 540b1c48c3 ("rxrpc: Fix deadlock between call creation and sendmsg/recvmsg")
Reported-by: Marc Dionne <marc.dionne@auristor.com>
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Make AF_RXRPC accept MSG_WAITALL as a flag to sendmsg() to tell it to
ignore signals whilst loading up the message queue, provided progress is
being made in emptying the queue at the other side.
Progress is defined as the base of the transmit window having being
advanced within 2 RTT periods. If the period is exceeded with no progress,
sendmsg() will return anyway, indicating how much data has been copied, if
any.
Once the supplied buffer is entirely decanted, the sendmsg() will return.
Signed-off-by: David Howells <dhowells@redhat.com>
Provide a couple of functions to allow cleaner handling of signals in a
kernel service. They are:
(1) rxrpc_kernel_get_rtt()
This allows the kernel service to find out the RTT time for a call, so
as to better judge how large a timeout to employ.
Note, though, that whilst this returns a value in nanoseconds, the
timeouts can only actually be in jiffies.
(2) rxrpc_kernel_check_life()
This returns a number that is updated when ACKs are received from the
peer (notably including PING RESPONSE ACKs which we can elicit by
sending PING ACKs to see if the call still exists on the server).
The caller should compare the numbers of two calls to see if the call
is still alive.
These can be used to provide an extending timeout rather than returning
immediately in the case that a signal occurs that would otherwise abort an
RPC operation. The timeout would be extended if the server is still
responsive and the call is still apparently alive on the server.
For most operations this isn't that necessary - but for FS.StoreData it is:
OpenAFS writes the data to storage as it comes in without making a backup,
so if we immediately abort it when partially complete on a CTRL+C, say, we
have no idea of the state of the file after the abort.
Signed-off-by: David Howells <dhowells@redhat.com>
Provide support for a kernel service to make use of the service upgrade
facility. This involves:
(1) Pass an upgrade request flag to rxrpc_kernel_begin_call().
(2) Make rxrpc_kernel_recv_data() return the call's current service ID so
that the caller can detect service upgrade and see what the service
was upgraded to.
Signed-off-by: David Howells <dhowells@redhat.com>
Currently, when passed a key that already exists, add_key() will call the
key's ->update() method if such exists. But this is heavily broken in the
case where the key is uninstantiated because it doesn't call
__key_instantiate_and_link(). Consequently, it doesn't do most of the
things that are supposed to happen when the key is instantiated, such as
setting the instantiation state, clearing KEY_FLAG_USER_CONSTRUCT and
awakening tasks waiting on it, and incrementing key->user->nikeys.
It also never takes key_construction_mutex, which means that
->instantiate() can run concurrently with ->update() on the same key. In
the case of the "user" and "logon" key types this causes a memory leak, at
best. Maybe even worse, the ->update() methods of the "encrypted" and
"trusted" key types actually just dereference a NULL pointer when passed an
uninstantiated key.
Change key_create_or_update() to wait interruptibly for the key to finish
construction before continuing.
This patch only affects *uninstantiated* keys. For now we still allow a
negatively instantiated key to be updated (thereby positively
instantiating it), although that's broken too (the next patch fixes it)
and I'm not sure that anyone actually uses that functionality either.
Here is a simple reproducer for the bug using the "encrypted" key type
(requires CONFIG_ENCRYPTED_KEYS=y), though as noted above the bug
pertained to more than just the "encrypted" key type:
#include <stdlib.h>
#include <unistd.h>
#include <keyutils.h>
int main(void)
{
int ringid = keyctl_join_session_keyring(NULL);
if (fork()) {
for (;;) {
const char payload[] = "update user:foo 32";
usleep(rand() % 10000);
add_key("encrypted", "desc", payload, sizeof(payload), ringid);
keyctl_clear(ringid);
}
} else {
for (;;)
request_key("encrypted", "desc", "callout_info", ringid);
}
}
It causes:
BUG: unable to handle kernel NULL pointer dereference at 0000000000000018
IP: encrypted_update+0xb0/0x170
PGD 7a178067 P4D 7a178067 PUD 77269067 PMD 0
PREEMPT SMP
CPU: 0 PID: 340 Comm: reproduce Tainted: G D 4.14.0-rc1-00025-g428490e38b2e #796
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011
task: ffff8a467a39a340 task.stack: ffffb15c40770000
RIP: 0010:encrypted_update+0xb0/0x170
RSP: 0018:ffffb15c40773de8 EFLAGS: 00010246
RAX: 0000000000000000 RBX: ffff8a467a275b00 RCX: 0000000000000000
RDX: 0000000000000005 RSI: ffff8a467a275b14 RDI: ffffffffb742f303
RBP: ffffb15c40773e20 R08: 0000000000000000 R09: ffff8a467a275b17
R10: 0000000000000020 R11: 0000000000000000 R12: 0000000000000000
R13: 0000000000000000 R14: ffff8a4677057180 R15: ffff8a467a275b0f
FS: 00007f5d7fb08700(0000) GS:ffff8a467f200000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000000000000018 CR3: 0000000077262005 CR4: 00000000001606f0
Call Trace:
key_create_or_update+0x2bc/0x460
SyS_add_key+0x10c/0x1d0
entry_SYSCALL_64_fastpath+0x1f/0xbe
RIP: 0033:0x7f5d7f211259
RSP: 002b:00007ffed03904c8 EFLAGS: 00000246 ORIG_RAX: 00000000000000f8
RAX: ffffffffffffffda RBX: 000000003b2a7955 RCX: 00007f5d7f211259
RDX: 00000000004009e4 RSI: 00000000004009ff RDI: 0000000000400a04
RBP: 0000000068db8bad R08: 000000003b2a7955 R09: 0000000000000004
R10: 000000000000001a R11: 0000000000000246 R12: 0000000000400868
R13: 00007ffed03905d0 R14: 0000000000000000 R15: 0000000000000000
Code: 77 28 e8 64 34 1f 00 45 31 c0 31 c9 48 8d 55 c8 48 89 df 48 8d 75 d0 e8 ff f9 ff ff 85 c0 41 89 c4 0f 88 84 00 00 00 4c 8b 7d c8 <49> 8b 75 18 4c 89 ff e8 24 f8 ff ff 85 c0 41 89 c4 78 6d 49 8b
RIP: encrypted_update+0xb0/0x170 RSP: ffffb15c40773de8
CR2: 0000000000000018
Cc: <stable@vger.kernel.org> # v2.6.12+
Reported-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Eric Biggers <ebiggers@google.com>
Consolidate KEY_FLAG_INSTANTIATED, KEY_FLAG_NEGATIVE and the rejection
error into one field such that:
(1) The instantiation state can be modified/read atomically.
(2) The error can be accessed atomically with the state.
(3) The error isn't stored unioned with the payload pointers.
This deals with the problem that the state is spread over three different
objects (two bits and a separate variable) and reading or updating them
atomically isn't practical, given that not only can uninstantiated keys
change into instantiated or rejected keys, but rejected keys can also turn
into instantiated keys - and someone accessing the key might not be using
any locking.
The main side effect of this problem is that what was held in the payload
may change, depending on the state. For instance, you might observe the
key to be in the rejected state. You then read the cached error, but if
the key semaphore wasn't locked, the key might've become instantiated
between the two reads - and you might now have something in hand that isn't
actually an error code.
The state is now KEY_IS_UNINSTANTIATED, KEY_IS_POSITIVE or a negative error
code if the key is negatively instantiated. The key_is_instantiated()
function is replaced with key_is_positive() to avoid confusion as negative
keys are also 'instantiated'.
Additionally, barriering is included:
(1) Order payload-set before state-set during instantiation.
(2) Order state-read before payload-read when using the key.
Further separate barriering is necessary if RCU is being used to access the
payload content after reading the payload pointers.
Fixes: 146aa8b145 ("KEYS: Merge the type-specific data with the payload data")
Cc: stable@vger.kernel.org # v4.4+
Reported-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: David Howells <dhowells@redhat.com>
Reviewed-by: Eric Biggers <ebiggers@google.com>
When an RxRPC service packet comes in, the target connection is looked up
by an rb-tree search under RCU and a read-locked seqlock; the seqlock retry
check is, however, currently skipped if we got a match, but probably
shouldn't be in case the connection we found gets replaced whilst we're
doing a search.
Make the lookup procedure always go through need_seqretry(), even if the
lookup was successful. This makes sure we always pick up on a write-lock
event.
On the other hand, since we don't take a ref on the object, but rely on RCU
to prevent its destruction after dropping the seqlock, I'm not sure this is
necessary.
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Allow a client call that failed on network error to be retried, provided
that the Tx queue still holds DATA packet 1. This allows an operation to
be submitted to another server or another address for the same server
without having to repackage and re-encrypt the data so far processed.
Two new functions are provided:
(1) rxrpc_kernel_check_call() - This is used to find out the completion
state of a call to guess whether it can be retried and whether it
should be retried.
(2) rxrpc_kernel_retry_call() - Disconnect the call from its current
connection, reset the state and submit it as a new client call to a
new address. The new address need not match the previous address.
A call may be retried even if all the data hasn't been loaded into it yet;
a partially constructed will be retained at the same point it was at when
an error condition was detected. msg_data_left() can be used to find out
how much data was packaged before the error occurred.
Signed-off-by: David Howells <dhowells@redhat.com>
Add a callback to rxrpc_kernel_send_data() so that a kernel service can get
a notification that the AF_RXRPC call has transitioned out the Tx phase and
is now waiting for a reply or a final ACK.
This is called from AF_RXRPC with the call state lock held so the
notification is guaranteed to come before any reply is passed back.
Further, modify the AFS filesystem to make use of this so that we don't have
to change the afs_call state before sending the last bit of data.
Signed-off-by: David Howells <dhowells@redhat.com>
call->error is stored as 0 or a negative error code. Don't negate this
value (ie. make it positive) before returning it from a kernel function
(though it should still be negated before passing to userspace through a
control message).
Signed-off-by: David Howells <dhowells@redhat.com>
Fix IPv6 support in AF_RXRPC in the following ways:
(1) When extracting the address from a received IPv4 packet, if the local
transport socket is open for IPv6 then fill out the sockaddr_rxrpc
struct for an IPv4-mapped-to-IPv6 AF_INET6 transport address instead
of an AF_INET one.
(2) When sending CHALLENGE or RESPONSE packets, the transport length needs
to be set from the sockaddr_rxrpc::transport_len field rather than
sizeof() on the IPv4 transport address.
(3) When processing an IPv4 ICMP packet received by an IPv6 socket, set up
the address correctly before searching for the affected peer.
Signed-off-by: David Howells <dhowells@redhat.com>
When an XDR-encoded Kerberos 5 ticket is added as an rxrpc-type key, the
expiry time should be drawn from the k5 part of the token union (which was
what was filled in), rather than the kad part of the union.
Reported-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David Howells <dhowells@redhat.com>
rxrpc_service_prealloc_one() doesn't set the socket pointer on any new call
it preallocates, but does add it to the rxrpc net namespace call list.
This, however, causes rxrpc_put_call() to oops when the call is discarded
when the socket is closed. rxrpc_put_call() needs the socket to be able to
reach the namespace so that it can use a lock held therein.
Fix this by setting a call's socket pointer immediately before discarding
it.
This can be triggered by unloading the kafs module, resulting in an oops
like the following:
BUG: unable to handle kernel NULL pointer dereference at 0000000000000030
IP: rxrpc_put_call+0x1e2/0x32d
PGD 0
P4D 0
Oops: 0000 [#1] SMP
Modules linked in: kafs(E-)
CPU: 3 PID: 3037 Comm: rmmod Tainted: G E 4.12.0-fscache+ #213
Hardware name: ASUS All Series/H97-PLUS, BIOS 2306 10/09/2014
task: ffff8803fc92e2c0 task.stack: ffff8803fef74000
RIP: 0010:rxrpc_put_call+0x1e2/0x32d
RSP: 0018:ffff8803fef77e08 EFLAGS: 00010282
RAX: 0000000000000000 RBX: ffff8803fab99ac0 RCX: 000000000000000f
RDX: ffffffff81c50a40 RSI: 000000000000000c RDI: ffff8803fc92ea88
RBP: ffff8803fef77e30 R08: ffff8803fc87b941 R09: ffffffff82946d20
R10: ffff8803fef77d10 R11: 00000000000076fc R12: 0000000000000005
R13: ffff8803fab99c20 R14: 0000000000000001 R15: ffffffff816c6aee
FS: 00007f915a059700(0000) GS:ffff88041fb80000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000000000000030 CR3: 00000003fef39000 CR4: 00000000001406e0
Call Trace:
rxrpc_discard_prealloc+0x325/0x341
rxrpc_listen+0xf9/0x146
kernel_listen+0xb/0xd
afs_close_socket+0x3e/0x173 [kafs]
afs_exit+0x1f/0x57 [kafs]
SyS_delete_module+0x10f/0x19a
do_syscall_64+0x8a/0x149
entry_SYSCALL64_slow_path+0x25/0x25
Fixes: 2baec2c3f8 ("rxrpc: Support network namespacing")
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
David Howells