blaster4385/linux-IllusionX/drivers/nvme, branch v6.12.10

nvmet: Don't overflow subsysnqn

2025-01-09T12:33:40Z

[ Upstream commit 4db3d750ac7e894278ef1cb1c53cc7d883060496 ] nvmet_root_discovery_nqn_store treats the subsysnqn string like a fixed size buffer, even though it is dynamically allocated to the size of the string. Create a new string with kstrndup instead of using the old buffer. Reported-by: syzbot+ff4aab278fa7e27e0f9e@syzkaller.appspotmail.com Closes: https://syzkaller.appspot.com/bug?extid=ff4aab278fa7e27e0f9e Fixes: 95409e277d83 ("nvmet: implement unique discovery NQN") Signed-off-by: Leo Stone Reviewed-by: Sagi Grimberg Reviewed-by: Christoph Hellwig Signed-off-by: Keith Busch Signed-off-by: Sasha Levin

nvme-pci: 512 byte aligned dma pool segment quirk

2025-01-09T12:33:34Z

[ Upstream commit ebefac5647968679f6ef5803e5d35a71997d20fa ] We initially introduced a quick fix limiting the queue depth to 1 as experimentation showed that it fixed data corruption on 64GB steamdecks. Further experimentation revealed corruption only happens when the last PRP data element aligns to the end of the page boundary. The device appears to treat this as a PRP chain to a new list instead of the data element that it actually is. This implementation is in violation of the spec. Encountering this errata with the Linux driver requires the host request a 128k transfer and coincidently be handed the last small pool dma buffer within a page. The QD1 quirk effectly works around this because the last data PRP always was at a 248 byte offset from the page start, so it never appeared at the end of the page, but comes at the expense of throttling IO and wasting the remainder of the PRP page beyond 256 bytes. Also to note, the MDTS on these devices is small enough that the "large" prp pool can hold enough PRP elements to never reach the end, so that pool is not a problem either. Introduce a new quirk to ensure the small pool is always aligned such that the last PRP element can't appear a the end of the page. This comes at the expense of wasting 256 bytes per small pool page allocated. Link: https://lore.kernel.org/linux-nvme/20241113043151.GA20077@lst.de/T/#u Fixes: 83bdfcbdbe5d ("nvme-pci: qdepth 1 quirk") Cc: Paweł Anikiel Signed-off-by: Robert Beckett Signed-off-by: Keith Busch Signed-off-by: Sasha Levin

Revert "nvme: make keep-alive synchronous operation"

2024-12-14T19:04:09Z

[ Upstream commit 84488282166de6b6760ada8030e87aaa08bce3aa ] This reverts commit d06923670b5a5f609603d4a9fee4dec02d38de9c. It was realized that the fix implemented to contain the race condition among the keep alive task and the fabric shutdown code path in the commit d06923670b5ia ("nvme: make keep-alive synchronous operation") is not optimal. The reason being keep-alive runs under the workqueue and making it synchronous would waste a workqueue context. Furthermore, we later found that the above race condition is a regression caused due to the changes implemented in commit a54a93d0e359 ("nvme: move stopping keep-alive into nvme_uninit_ctrl()"). So we decided to revert the commit d06923670b5a ("nvme: make keep-alive synchronous operation") and then fix the regression. Link: https://lore.kernel.org/all/196f4013-3bbf-43ff-98b4-9cb2a96c20c2@grimberg.me/ Reviewed-by: Ming Lei Signed-off-by: Nilay Shroff Signed-off-by: Keith Busch Signed-off-by: Sasha Levin

nvme-rdma: unquiesce admin_q before destroy it

2024-12-14T19:03:17Z

[ Upstream commit 5858b687559809f05393af745cbadf06dee61295 ] Kernel will hang on destroy admin_q while we create ctrl failed, such as following calltrace: PID: 23644 TASK: ff2d52b40f439fc0 CPU: 2 COMMAND: "nvme" #0 [ff61d23de260fb78] __schedule at ffffffff8323bc15 #1 [ff61d23de260fc08] schedule at ffffffff8323c014 #2 [ff61d23de260fc28] blk_mq_freeze_queue_wait at ffffffff82a3dba1 #3 [ff61d23de260fc78] blk_freeze_queue at ffffffff82a4113a #4 [ff61d23de260fc90] blk_cleanup_queue at ffffffff82a33006 #5 [ff61d23de260fcb0] nvme_rdma_destroy_admin_queue at ffffffffc12686ce #6 [ff61d23de260fcc8] nvme_rdma_setup_ctrl at ffffffffc1268ced #7 [ff61d23de260fd28] nvme_rdma_create_ctrl at ffffffffc126919b #8 [ff61d23de260fd68] nvmf_dev_write at ffffffffc024f362 #9 [ff61d23de260fe38] vfs_write at ffffffff827d5f25 RIP: 00007fda7891d574 RSP: 00007ffe2ef06958 RFLAGS: 00000202 RAX: ffffffffffffffda RBX: 000055e8122a4d90 RCX: 00007fda7891d574 RDX: 000000000000012b RSI: 000055e8122a4d90 RDI: 0000000000000004 RBP: 00007ffe2ef079c0 R8: 000000000000012b R9: 000055e8122a4d90 R10: 0000000000000000 R11: 0000000000000202 R12: 0000000000000004 R13: 000055e8122923c0 R14: 000000000000012b R15: 00007fda78a54500 ORIG_RAX: 0000000000000001 CS: 0033 SS: 002b This due to we have quiesced admi_q before cancel requests, but forgot to unquiesce before destroy it, as a result we fail to drain the pending requests, and hang on blk_mq_freeze_queue_wait() forever. Here try to reuse nvme_rdma_teardown_admin_queue() to fix this issue and simplify the code. Fixes: 958dc1d32c80 ("nvme-rdma: add clean action for failed reconnection") Reported-by: Yingfu.zhou Signed-off-by: Chunguang.xu Signed-off-by: Yue.zhao Reviewed-by: Christoph Hellwig Reviewed-by: Hannes Reinecke Signed-off-by: Keith Busch Signed-off-by: Sasha Levin

nvme-tcp: fix the memleak while create new ctrl failed

2024-12-14T19:03:17Z

[ Upstream commit fec55c29e54d3ca6fe9d7d7d9266098b4514fd34 ] Now while we create new ctrl failed, we have not free the tagset occupied by admin_q, here try to fix it. Fixes: fd1418de10b9 ("nvme-tcp: avoid open-coding nvme_tcp_teardown_admin_queue()") Signed-off-by: Chunguang.xu Reviewed-by: Christoph Hellwig Reviewed-by: Hannes Reinecke Signed-off-by: Keith Busch Signed-off-by: Sasha Levin

nvme-fabrics: handle zero MAXCMD without closing the connection

2024-12-14T19:03:17Z

[ Upstream commit 88c23a32b851e36adc4ab36f796d9b711f47e2bb ] The NVMe specification states that MAXCMD is mandatory for NVMe-over-Fabrics implementations. However, some NVMe/TCP and NVMe/FC arrays from major vendors have buggy firmware that reports MAXCMD as zero in the Identify Controller data structure. Currently, the implementation closes the connection in such cases, completely preventing the host from connecting to the target. Fix the issue by printing a clear error message about the firmware bug and allowing the connection to proceed. It assumes that the target supports a MAXCMD value of SQSIZE + 1. If any issues arise, the user can manually adjust SQSIZE to mitigate them. Fixes: 4999568184e5 ("nvme-fabrics: check max outstanding commands") Signed-off-by: Maurizio Lombardi Reviewed-by: Laurence Oberman Reviewed-by: Christoph Hellwig Signed-off-by: Keith Busch Signed-off-by: Sasha Levin

nvme: don't apply NVME_QUIRK_DEALLOCATE_ZEROES when DSM is not supported

2024-12-14T19:03:16Z

[ Upstream commit 58a0c875ce028678c9594c7bdf3fe33462392808 ] Commit 63dfa1004322 ("nvme: move NVME_QUIRK_DEALLOCATE_ZEROES out of nvme_config_discard") started applying the NVME_QUIRK_DEALLOCATE_ZEROES quirk even then the Dataset Management is not supported. It turns out that there versions of these old Intel SSDs that have DSM support disabled in the firmware, which will now lead to errors everytime a Write Zeroes command is issued. Fix this by checking for DSM support before applying the quirk. Reported-by: Saeed Mirzamohammadi Fixes: 63dfa1004322 ("nvme: move NVME_QUIRK_DEALLOCATE_ZEROES out of nvme_config_discard") Tested-by: Saeed Mirzamohammadi Signed-off-by: Christoph Hellwig Reviewed-by: Nitesh Shetty Reviewed-by: Chaitanya Kulkarni Signed-off-by: Keith Busch Signed-off-by: Sasha Levin

nvme-fabrics: fix kernel crash while shutting down controller

2024-12-05T13:03:06Z

[ Upstream commit e9869c85c81168a1275f909d5972a3fc435304be ] The nvme keep-alive operation, which executes at a periodic interval, could potentially sneak in while shutting down a fabric controller. This may lead to a race between the fabric controller admin queue destroy code path (invoked while shutting down controller) and hw/hctx queue dispatcher called from the nvme keep-alive async request queuing operation. This race could lead to the kernel crash shown below: Call Trace: autoremove_wake_function+0x0/0xbc (unreliable) __blk_mq_sched_dispatch_requests+0x114/0x24c blk_mq_sched_dispatch_requests+0x44/0x84 blk_mq_run_hw_queue+0x140/0x220 nvme_keep_alive_work+0xc8/0x19c [nvme_core] process_one_work+0x200/0x4e0 worker_thread+0x340/0x504 kthread+0x138/0x140 start_kernel_thread+0x14/0x18 While shutting down fabric controller, if nvme keep-alive request sneaks in then it would be flushed off. The nvme_keep_alive_end_io function is then invoked to handle the end of the keep-alive operation which decrements the admin->q_usage_counter and assuming this is the last/only request in the admin queue then the admin->q_usage_counter becomes zero. If that happens then blk-mq destroy queue operation (blk_mq_destroy_ queue()) which could be potentially running simultaneously on another cpu (as this is the controller shutdown code path) would forward progress and deletes the admin queue. So, now from this point onward we are not supposed to access the admin queue resources. However the issue here's that the nvme keep-alive thread running hw/hctx queue dispatch operation hasn't yet finished its work and so it could still potentially access the admin queue resource while the admin queue had been already deleted and that causes the above crash. The above kernel crash is regression caused due to changes implemented in commit a54a93d0e359 ("nvme: move stopping keep-alive into nvme_uninit_ctrl()"). Ideally we should stop keep-alive before destroyin g the admin queue and freeing the admin tagset so that it wouldn't sneak in during the shutdown operation. However we removed the keep alive stop operation from the beginning of the controller shutdown code path in commit a54a93d0e359 ("nvme: move stopping keep-alive into nvme_uninit_ctrl()") and added it under nvme_uninit_ctrl() which executes very late in the shutdown code path after the admin queue is destroyed and its tagset is removed. So this change created the possibility of keep-alive sneaking in and interfering with the shutdown operation and causing observed kernel crash. To fix the observed crash, we decided to move nvme_stop_keep_alive() from nvme_uninit_ctrl() to nvme_remove_admin_tag_set(). This change would ensure that we don't forward progress and delete the admin queue until the keep- alive operation is finished (if it's in-flight) or cancelled and that would help contain the race condition explained above and hence avoid the crash. Moving nvme_stop_keep_alive() to nvme_remove_admin_tag_set() instead of adding nvme_stop_keep_alive() to the beginning of the controller shutdown code path in nvme_stop_ctrl(), as was the case earlier before commit a54a93d0e359 ("nvme: move stopping keep-alive into nvme_uninit_ctrl()"), would help save one callsite of nvme_stop_keep_alive(). Fixes: a54a93d0e359 ("nvme: move stopping keep-alive into nvme_uninit_ctrl()") Link: https://lore.kernel.org/all/1a21f37b-0f2a-4745-8c56-4dc8628d3983@linux.ibm.com/ Reviewed-by: Ming Lei Signed-off-by: Nilay Shroff Signed-off-by: Keith Busch Signed-off-by: Sasha Levin

nvme/multipath: Fix RCU list traversal to use SRCU primitive

2024-12-05T13:03:05Z

[ Upstream commit 5dd18f09ce7399df6fffe80d1598add46c395ae9 ] The code currently uses list_for_each_entry_rcu() while holding an SRCU lock, triggering false positive warnings with CONFIG_PROVE_RCU=y enabled: drivers/nvme/host/multipath.c:168 RCU-list traversed in non-reader section!! drivers/nvme/host/multipath.c:227 RCU-list traversed in non-reader section!! drivers/nvme/host/multipath.c:260 RCU-list traversed in non-reader section!! While the list is properly protected by SRCU lock, the code uses the wrong list traversal primitive. Replace list_for_each_entry_rcu() with list_for_each_entry_srcu() to correctly indicate SRCU-based protection and eliminate the false warning. Signed-off-by: Breno Leitao Fixes: be647e2c76b2 ("nvme: use srcu for iterating namespace list") Signed-off-by: Keith Busch Signed-off-by: Sasha Levin

nvme-pci: reverse request order in nvme_queue_rqs

2024-12-05T13:01:15Z

[ Upstream commit beadf0088501d9dcf2454b05d90d5d31ea3ba55f ] blk_mq_flush_plug_list submits requests in the reverse order that they were submitted, which leads to a rather suboptimal I/O pattern especially in rotational devices. Fix this by rewriting nvme_queue_rqs so that it always pops the requests from the passed in request list, and then adds them to the head of a local submit list. This actually simplifies the code a bit as it removes the complicated list splicing, at the cost of extra updates of the rq_next pointer. As that should be cache hot anyway it should be an easy price to pay. Fixes: d62cbcf62f2f ("nvme: add support for mq_ops->queue_rqs()") Signed-off-by: Christoph Hellwig Link: https://lore.kernel.org/r/20241113152050.157179-2-hch@lst.de Signed-off-by: Jens Axboe Signed-off-by: Sasha Levin