kernel.git/fs/afs/server.c, branch linux-4.19.y

afs: fix the usage of read_seqbegin_or_lock() in afs_find_server*()

2024-02-23T07:12:45Z

[ Upstream commit 1702e0654ca9a7bcd7c7619c8a5004db58945b71 ] David Howells says: (5) afs_find_server(). There could be a lot of servers in the list and each server can have multiple addresses, so I think this would be better with an exclusive second pass. The server list isn't likely to change all that often, but when it does change, there's a good chance several servers are going to be added/removed one after the other. Further, this is only going to be used for incoming cache management/callback requests from the server, which hopefully aren't going to happen too often - but it is remotely drivable. (6) afs_find_server_by_uuid(). Similarly to (5), there could be a lot of servers to search through, but they are in a tree not a flat list, so it should be faster to process. Again, it's not likely to change that often and, again, when it does change it's likely to involve multiple changes. This can be driven remotely by an incoming cache management request but is mostly going to be driven by setting up or reconfiguring a volume's server list - something that also isn't likely to happen often. Make the "seq" counter odd on the 2nd pass, otherwise read_seqbegin_or_lock() never takes the lock. Signed-off-by: Oleg Nesterov Signed-off-by: David Howells cc: Marc Dionne cc: linux-afs@lists.infradead.org Link: https://lore.kernel.org/r/20231130115614.GA21581@redhat.com/ Signed-off-by: Sasha Levin

afs: Fix afs_find_server lookups for ipv4 peers

2020-01-09T09:18:55Z

[ Upstream commit 9bd0160d12370a076e44f8d1320cde9c83f2c647 ] afs_find_server tries to find a server that has an address that matches the transport address of an rxrpc peer. The code assumes that the transport address is always ipv6, with ipv4 represented as ipv4 mapped addresses, but that's not the case. If the transport family is AF_INET, srx->transport.sin6.sin6_addr.s6_addr32[] will be beyond the actual ipv4 address and will always be 0, and all ipv4 addresses will be seen as matching. As a result, the first ipv4 address seen on any server will be considered a match, and the server returned may be the wrong one. One of the consequences is that callbacks received over ipv4 will only be correctly applied for the server that happens to have the first ipv4 address on the fs_addresses4 list. Callbacks over ipv4 from all other servers are dropped, causing the client to serve stale data. This is fixed by looking at the transport family, and comparing ipv4 addresses based on a sockaddr_in structure rather than a sockaddr_in6. Fixes: d2ddc776a458 ("afs: Overhaul volume and server record caching and fileserver rotation") Signed-off-by: Marc Dionne Signed-off-by: David Howells Signed-off-by: Sasha Levin

afs: Fix clearance of reply

2018-10-15T13:31:47Z

The recent patch to fix the afs_server struct leak didn't actually fix the bug, but rather fixed some of the symptoms. The problem is that an asynchronous call that holds a resource pointed to by call->reply[0] will find the pointer cleared in the call destructor, thereby preventing the resource from being cleaned up. In the case of the server record leak, the afs_fs_get_capabilities() function in devel code sets up a call with reply[0] pointing at the server record that should be altered when the result is obtained, but this was being cleared before the destructor was called, so the put in the destructor does nothing and the record is leaked. Commit f014ffb025c1 removed the additional ref obtained by afs_install_server(), but the removal of this ref is actually used by the garbage collector to mark a server record as being defunct after the record has expired through lack of use. The offending clearance of call->reply[0] upon completion in afs_process_async_call() has been there from the origin of the code, but none of the asynchronous calls actually use that pointer currently, so it should be safe to remove (note that synchronous calls don't involve this function). Fix this by the following means: (1) Revert commit f014ffb025c1. (2) Remove the clearance of reply[0] from afs_process_async_call(). Without this, afs_manage_servers() will suffer an assertion failure if it sees a server record that didn't get used because the usage count is not 1. Fixes: f014ffb025c1 ("afs: Fix afs_server struct leak") Fixes: 08e0e7c82eea ("[AF_RXRPC]: Make the in-kernel AFS filesystem use AF_RXRPC.") Signed-off-by: David Howells Cc: stable Signed-off-by: Greg Kroah-Hartman

afs: Fix afs_server struct leak

2018-10-12T15:36:40Z

Fix a leak of afs_server structs. The routine that installs them in the various lookup lists and trees gets a ref on leaving the function, whether it added the server or a server already exists. It shouldn't increment the refcount if it added the server. The effect of this that "rmmod kafs" will hang waiting for the leaked server to become unused. Fixes: d2ddc776a458 ("afs: Overhaul volume and server record caching and fileserver rotation") Signed-off-by: David Howells Signed-off-by: Greg Kroah-Hartman

afs: Optimise callback breaking by not repeating volume lookup

2018-06-15T14:27:09Z

At the moment, afs_break_callbacks calls afs_break_one_callback() for each separate FID it was given, and the latter looks up the volume individually for each one. However, this is inefficient if two or more FIDs have the same vid as we could reuse the volume. This is complicated by cell aliasing whereby we may have multiple cells sharing a volume and can therefore have multiple callback interests for any particular volume ID. At the moment afs_break_one_callback() scans the entire list of volumes we're getting from a server and breaks the appropriate callback in every matching volume, regardless of cell. This scan is done for every FID. Optimise callback breaking by the following means: (1) Sort the FID list by vid so that all FIDs belonging to the same volume are clumped together. This is done through the use of an indirection table as we cannot do an insertion sort on the afs_callback_break array as we decode FIDs into it as we subsequently also have to decode callback info into it that corresponds by array index only. We also don't really want to bubblesort afterwards if we can avoid it. (2) Sort the server->cb_interests array by vid so that all the matching volumes are grouped together. This permits the scan to stop after finding a record that has a higher vid. (3) When breaking FIDs, we try to keep server->cb_break_lock as long as possible, caching the start point in the array for that volume group as long as possible. It might make sense to add another layer in that list and have a refcounted volume ID anchor that has the matching interests attached to it rather than being in the list. This would allow the lock to be dropped without losing the cursor. Signed-off-by: David Howells

afs: Fix afs_find_server search loop

2018-05-14T14:15:18Z

The code that looks up servers by addresses makes the assumption that the list of addresses for a server is sorted. It exits the loop if it finds that the target address is larger than the current candidate. As the list is not currently sorted, this can lead to a failure to find a matching server, which can cause callbacks from that server to be ignored. Remove the early exit case so that the complete list is searched. Fixes: d2ddc776a458 ("afs: Overhaul volume and server record caching and fileserver rotation") Signed-off-by: Marc Dionne Signed-off-by: David Howells

afs: Fix giving up callbacks on server destruction

2018-05-14T12:17:35Z

When a server record is destroyed, we want to send a message to the server telling it that we're giving up all the callbacks it has promised us. Apply two fixes to this: (1) Only send the FS.GiveUpAllCallBacks message if we actually got a callback from that server. We assume this to be the case if we performed at least one successful FS operation on that server. (2) Send it to the address last used for that server rather than always picking the first address in the list (which might be unreachable). Fixes: d2ddc776a458 ("afs: Overhaul volume and server record caching and fileserver rotation") Signed-off-by: David Howells

afs: Fix server record deletion

2018-04-20T16:59:33Z

AFS server records get removed from the net->fs_servers tree when they're deleted, but not from the net->fs_addresses{4,6} lists, which can lead to an oops in afs_find_server() when a server record has been removed, for instance during rmmod. Fix this by deleting the record from the by-address lists before posting it for RCU destruction. The reason this hasn't been noticed before is that the fileserver keeps probing the local cache manager, thereby keeping the service record alive, so the oops would only happen when a fileserver eventually gets bored and stops pinging or if the module gets rmmod'd and a call comes in from the fileserver during the window between the server records being destroyed and the socket being closed. The oops looks something like: BUG: unable to handle kernel NULL pointer dereference at 000000000000001c ... Workqueue: kafsd afs_process_async_call [kafs] RIP: 0010:afs_find_server+0x271/0x36f [kafs] ... Call Trace: afs_deliver_cb_init_call_back_state3+0x1f2/0x21f [kafs] afs_deliver_to_call+0x1ee/0x5e8 [kafs] afs_process_async_call+0x5b/0xd0 [kafs] process_one_work+0x2c2/0x504 worker_thread+0x1d4/0x2ac kthread+0x11f/0x127 ret_from_fork+0x24/0x30 Fixes: d2ddc776a458 ("afs: Overhaul volume and server record caching and fileserver rotation") Signed-off-by: David Howells Signed-off-by: Linus Torvalds

afs: Fix checker warnings

2018-04-09T20:12:31Z

Fix warnings raised by checker, including: (*) Warnings raised by unequal comparison for the purposes of sorting, where the endianness doesn't matter: fs/afs/addr_list.c:246:21: warning: restricted __be16 degrades to integer fs/afs/addr_list.c:246:30: warning: restricted __be16 degrades to integer fs/afs/addr_list.c:248:21: warning: restricted __be32 degrades to integer fs/afs/addr_list.c:248:49: warning: restricted __be32 degrades to integer fs/afs/addr_list.c:283:21: warning: restricted __be16 degrades to integer fs/afs/addr_list.c:283:30: warning: restricted __be16 degrades to integer (*) afs_set_cb_interest() is not actually used and can be removed. (*) afs_cell_gc_delay() should be provided with a sysctl. (*) afs_cell_destroy() needs to use rcu_access_pointer() to read cell->vl_addrs. (*) afs_init_fs_cursor() should be static. (*) struct afs_vnode::permit_cache needs to be marked __rcu. (*) afs_server_rcu() needs to use rcu_access_pointer(). (*) afs_destroy_server() should use rcu_access_pointer() on server->addresses as the server object is no longer accessible. (*) afs_find_server() casts __be16/__be32 values to int in order to directly compare them for the purpose of finding a match in a list, but is should also annotate the cast with __force to avoid checker warnings. (*) afs_check_permit() accesses vnode->permit_cache outside of the RCU readlock, though it doesn't then access the value; the extraneous access is deleted. False positives: (*) Conditional locking around the code in xdr_decode_AFSFetchStatus. This can be dealt with in a separate patch. fs/afs/fsclient.c:148:9: warning: context imbalance in 'xdr_decode_AFSFetchStatus' - different lock contexts for basic block (*) Incorrect handling of seq-retry lock context balance: fs/afs/inode.c:455:38: warning: context imbalance in 'afs_getattr' - different lock contexts for basic block fs/afs/server.c:52:17: warning: context imbalance in 'afs_find_server' - different lock contexts for basic block fs/afs/server.c:128:17: warning: context imbalance in 'afs_find_server_by_uuid' - different lock contexts for basic block Errors: (*) afs_lookup_cell_rcu() needs to break out of the seq-retry loop, not go round again if it successfully found the workstation cell. (*) Fix UUID decode in afs_deliver_cb_probe_uuid(). (*) afs_cache_permit() has a missing rcu_read_unlock() before one of the jumps to the someone_else_changed_it label. Move the unlock to after the label. (*) afs_vl_get_addrs_u() is using ntohl() rather than htonl() when encoding to XDR. (*) afs_deliver_yfsvl_get_endpoints() is using htonl() rather than ntohl() when decoding from XDR. Signed-off-by: David Howells

sched/wait, fs/afs: Convert wait_on_atomic_t() usage to the new wait_var_event() API

2018-03-20T07:23:19Z

The old wait_on_atomic_t() is going to get removed, use the more flexible wait_var_event() API instead. No change in functionality. Signed-off-by: Peter Zijlstra (Intel) Cc: David Howells Cc: Linus Torvalds Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: linux-kernel@vger.kernel.org Signed-off-by: Ingo Molnar