kernel.git/net/ceph, branch linux-5.10.y

libceph: make calc_target() set t->paused, not just clear it

2026-01-19T12:12:08Z

commit c0fe2994f9a9d0a2ec9e42441ea5ba74b6a16176 upstream. Currently calc_target() clears t->paused if the request shouldn't be paused anymore, but doesn't ever set t->paused even though it's able to determine when the request should be paused. Setting t->paused is left to __submit_request() which is fine for regular requests but doesn't work for linger requests -- since __submit_request() doesn't operate on linger requests, there is nowhere for lreq->t.paused to be set. One consequence of this is that watches don't get reestablished on paused -> unpaused transitions in cases where requests have been paused long enough for the (paused) unwatch request to time out and for the subsequent (re)watch request to enter the paused state. On top of the watch not getting reestablished, rbd_reregister_watch() gets stuck with rbd_dev->watch_mutex held: rbd_register_watch __rbd_register_watch ceph_osdc_watch linger_reg_commit_wait It's waiting for lreq->reg_commit_wait to be completed, but for that to happen the respective request needs to end up on need_resend_linger list and be kicked when requests are unpaused. There is no chance for that if the request in question is never marked paused in the first place. The fact that rbd_dev->watch_mutex remains taken out forever then prevents the image from getting unmapped -- "rbd unmap" would inevitably hang in D state on an attempt to grab the mutex. Cc: stable@vger.kernel.org Reported-by: Raphael Zimmer Signed-off-by: Ilya Dryomov Reviewed-by: Viacheslav Dubeyko Signed-off-by: Greg Kroah-Hartman

libceph: make free_choose_arg_map() resilient to partial allocation

2026-01-19T12:12:08Z

commit e3fe30e57649c551757a02e1cad073c47e1e075e upstream. free_choose_arg_map() may dereference a NULL pointer if its caller fails after a partial allocation. For example, in decode_choose_args(), if allocation of arg_map->args fails, execution jumps to the fail label and free_choose_arg_map() is called. Since arg_map->size is updated to a non-zero value before memory allocation, free_choose_arg_map() will iterate over arg_map->args and dereference a NULL pointer. To prevent this potential NULL pointer dereference and make free_choose_arg_map() more resilient, add checks for pointers before iterating. Cc: stable@vger.kernel.org Co-authored-by: Ilya Dryomov Signed-off-by: Tuo Li Reviewed-by: Viacheslav Dubeyko Signed-off-by: Ilya Dryomov Signed-off-by: Greg Kroah-Hartman

libceph: replace overzealous BUG_ON in osdmap_apply_incremental()

2026-01-19T12:12:08Z

commit e00c3f71b5cf75681dbd74ee3f982a99cb690c2b upstream. If the osdmap is (maliciously) corrupted such that the incremental osdmap epoch is different from what is expected, there is no need to BUG. Instead, just declare the incremental osdmap to be invalid. Cc: stable@vger.kernel.org Reported-by: ziming zhang Signed-off-by: Ilya Dryomov Signed-off-by: Greg Kroah-Hartman

libceph: make decode_pool() more resilient against corrupted osdmaps

2026-01-19T12:11:50Z

commit 8c738512714e8c0aa18f8a10c072d5b01c83db39 upstream. If the osdmap is (maliciously) corrupted such that the encoded length of ceph_pg_pool envelope is less than what is expected for a particular encoding version, out-of-bounds reads may ensue because the only bounds check that is there is based on that length value. This patch adds explicit bounds checks for each field that is decoded or skipped. Cc: stable@vger.kernel.org Reported-by: ziming zhang Signed-off-by: Ilya Dryomov Reviewed-by: Xiubo Li Tested-by: ziming zhang Signed-off-by: Greg Kroah-Hartman

libceph: fix potential use-after-free in have_mon_and_osd_map()

2025-12-06T21:08:24Z

commit 076381c261374c587700b3accf410bdd2dba334e upstream. The wait loop in __ceph_open_session() can race with the client receiving a new monmap or osdmap shortly after the initial map is received. Both ceph_monc_handle_map() and handle_one_map() install a new map immediately after freeing the old one kfree(monc->monmap); monc->monmap = monmap; ceph_osdmap_destroy(osdc->osdmap); osdc->osdmap = newmap; under client->monc.mutex and client->osdc.lock respectively, but because neither is taken in have_mon_and_osd_map() it's possible for client->monc.monmap->epoch and client->osdc.osdmap->epoch arms in client->monc.monmap && client->monc.monmap->epoch && client->osdc.osdmap && client->osdc.osdmap->epoch; condition to dereference an already freed map. This happens to be reproducible with generic/395 and generic/397 with KASAN enabled: BUG: KASAN: slab-use-after-free in have_mon_and_osd_map+0x56/0x70 Read of size 4 at addr ffff88811012d810 by task mount.ceph/13305 CPU: 2 UID: 0 PID: 13305 Comm: mount.ceph Not tainted 6.14.0-rc2-build2+ #1266 ... Call Trace: have_mon_and_osd_map+0x56/0x70 ceph_open_session+0x182/0x290 ceph_get_tree+0x333/0x680 vfs_get_tree+0x49/0x180 do_new_mount+0x1a3/0x2d0 path_mount+0x6dd/0x730 do_mount+0x99/0xe0 __do_sys_mount+0x141/0x180 do_syscall_64+0x9f/0x100 entry_SYSCALL_64_after_hwframe+0x76/0x7e Allocated by task 13305: ceph_osdmap_alloc+0x16/0x130 ceph_osdc_init+0x27a/0x4c0 ceph_create_client+0x153/0x190 create_fs_client+0x50/0x2a0 ceph_get_tree+0xff/0x680 vfs_get_tree+0x49/0x180 do_new_mount+0x1a3/0x2d0 path_mount+0x6dd/0x730 do_mount+0x99/0xe0 __do_sys_mount+0x141/0x180 do_syscall_64+0x9f/0x100 entry_SYSCALL_64_after_hwframe+0x76/0x7e Freed by task 9475: kfree+0x212/0x290 handle_one_map+0x23c/0x3b0 ceph_osdc_handle_map+0x3c9/0x590 mon_dispatch+0x655/0x6f0 ceph_con_process_message+0xc3/0xe0 ceph_con_v1_try_read+0x614/0x760 ceph_con_workfn+0x2de/0x650 process_one_work+0x486/0x7c0 process_scheduled_works+0x73/0x90 worker_thread+0x1c8/0x2a0 kthread+0x2ec/0x300 ret_from_fork+0x24/0x40 ret_from_fork_asm+0x1a/0x30 Rewrite the wait loop to check the above condition directly with client->monc.mutex and client->osdc.lock taken as appropriate. While at it, improve the timeout handling (previously mount_timeout could be exceeded in case wait_event_interruptible_timeout() slept more than once) and access client->auth_err under client->monc.mutex to match how it's set in finish_auth(). monmap_show() and osdmap_show() now take the respective lock before accessing the map as well. Cc: stable@vger.kernel.org Reported-by: David Howells Signed-off-by: Ilya Dryomov Reviewed-by: Viacheslav Dubeyko Signed-off-by: Greg Kroah-Hartman

libceph: fix race between delayed_work() and ceph_monc_stop()

2024-07-18T11:05:49Z

commit 69c7b2fe4c9cc1d3b1186d1c5606627ecf0de883 upstream. The way the delayed work is handled in ceph_monc_stop() is prone to races with mon_fault() and possibly also finish_hunting(). Both of these can requeue the delayed work which wouldn't be canceled by any of the following code in case that happens after cancel_delayed_work_sync() runs -- __close_session() doesn't mess with the delayed work in order to avoid interfering with the hunting interval logic. This part was missed in commit b5d91704f53e ("libceph: behave in mon_fault() if cur_mon < 0") and use-after-free can still ensue on monc and objects that hang off of it, with monc->auth and monc->monmap being particularly susceptible to quickly being reused. To fix this: - clear monc->cur_mon and monc->hunting as part of closing the session in ceph_monc_stop() - bail from delayed_work() if monc->cur_mon is cleared, similar to how it's done in mon_fault() and finish_hunting() (based on monc->hunting) - call cancel_delayed_work_sync() after the session is closed Cc: stable@vger.kernel.org Link: https://tracker.ceph.com/issues/66857 Signed-off-by: Ilya Dryomov Reviewed-by: Xiubo Li Signed-off-by: Greg Kroah-Hartman

libceph: use kernel_connect()

2023-10-25T09:54:15Z

commit 7563cf17dce0a875ba3d872acdc63a78ea344019 upstream. Direct calls to ops->connect() can overwrite the address parameter when used in conjunction with BPF SOCK_ADDR hooks. Recent changes to kernel_connect() ensure that callers are insulated from such side effects. This patch wraps the direct call to ops->connect() with kernel_connect() to prevent unexpected changes to the address passed to ceph_tcp_connect(). This change was originally part of a larger patch targeting the net tree addressing all instances of unprotected calls to ops->connect() throughout the kernel, but this change was split up into several patches targeting various trees. Cc: stable@vger.kernel.org Link: https://lore.kernel.org/netdev/20230821100007.559638-1-jrife@google.com/ Link: https://lore.kernel.org/netdev/9944248dba1bce861375fcce9de663934d933ba9.camel@redhat.com/ Fixes: d74bad4e74ee ("bpf: Hooks for sys_connect") Signed-off-by: Jordan Rife Reviewed-by: Ilya Dryomov Signed-off-by: Ilya Dryomov Signed-off-by: Greg Kroah-Hartman

libceph, rbd: ignore addr->type while comparing in some cases

2023-08-30T14:23:11Z

[ Upstream commit 313771e80fd253d4b5472e61a2d12b03c5293aa9 ] For libceph, this ensures that libceph instance sharing (share option) continues to work. For rbd, this avoids blocklisting alive lock owners (locker addr is always LEGACY, while watcher addr is ANY in nautilus). Signed-off-by: Ilya Dryomov Stable-dep-of: 588159009d5b ("rbd: retrieve and check lock owner twice before blocklisting") Signed-off-by: Sasha Levin

libceph: fix potential hang in ceph_osdc_notify()

2023-08-11T09:57:51Z

commit e6e2843230799230fc5deb8279728a7218b0d63c upstream. If the cluster becomes unavailable, ceph_osdc_notify() may hang even with osd_request_timeout option set because linger_notify_finish_wait() waits for MWatchNotify NOTIFY_COMPLETE message with no associated OSD request in flight -- it's completely asynchronous. Introduce an additional timeout, derived from the specified notify timeout. While at it, switch both waits to killable which is more correct. Cc: stable@vger.kernel.org Signed-off-by: Ilya Dryomov Reviewed-by: Dongsheng Yang Reviewed-by: Xiubo Li Signed-off-by: Greg Kroah-Hartman

libceph: fix potential use-after-free on linger ping and resends

2022-05-25T07:17:56Z

commit 75dbb685f4e8786c33ddef8279bab0eadfb0731f upstream. request_reinit() is not only ugly as the comment rightfully suggests, but also unsafe. Even though it is called with osdc->lock held for write in all cases, resetting the OSD request refcount can still race with handle_reply() and result in use-after-free. Taking linger ping as an example: handle_timeout thread handle_reply thread down_read(&osdc->lock) req = lookup_request(...) ... finish_request(req) # unregisters up_read(&osdc->lock) __complete_request(req) linger_ping_cb(req) # req->r_kref == 2 because handle_reply still holds its ref down_write(&osdc->lock) send_linger_ping(lreq) req = lreq->ping_req # same req # cancel_linger_request is NOT # called - handle_reply already # unregistered request_reinit(req) WARN_ON(req->r_kref != 1) # fires request_init(req) kref_init(req->r_kref) # req->r_kref == 1 after kref_init ceph_osdc_put_request(req) kref_put(req->r_kref) # req->r_kref == 0 after kref_put, req is freed !!! This happens because send_linger_ping() always (re)uses the same OSD request for watch ping requests, relying on cancel_linger_request() to unregister it from the OSD client and rip its messages out from the messenger. send_linger() does the same for watch/notify registration and watch reconnect requests. Unfortunately cancel_request() doesn't guarantee that after it returns the OSD client would be completely done with the OSD request -- a ref could still be held and the callback (if specified) could still be invoked too. The original motivation for request_reinit() was inability to deal with allocation failures in send_linger() and send_linger_ping(). Switching to using osdc->req_mempool (currently only used by CephFS) respects that and allows us to get rid of request_reinit(). Cc: stable@vger.kernel.org Signed-off-by: Ilya Dryomov Reviewed-by: Xiubo Li Acked-by: Jeff Layton Signed-off-by: Greg Kroah-Hartman