kernel/linux.git/drivers/nvme, branch v6.6.39

nvmet: fix a possible leak when destroy a ctrl during qp establishment

2024-07-11T10:49:21+00:00

[ Upstream commit c758b77d4a0a0ed3a1292b3fd7a2aeccd1a169a4 ] In nvmet_sq_destroy we capture sq->ctrl early and if it is non-NULL we know that a ctrl was allocated (in the admin connect request handler) and we need to release pending AERs, clear ctrl->sqs and sq->ctrl (for nvme-loop primarily), and drop the final reference on the ctrl. However, a small window is possible where nvmet_sq_destroy starts (as a result of the client giving up and disconnecting) concurrently with the nvme admin connect cmd (which may be in an early stage). But *before* kill_and_confirm of sq->ref (i.e. the admin connect managed to get an sq live reference). In this case, sq->ctrl was allocated however after it was captured in a local variable in nvmet_sq_destroy. This prevented the final reference drop on the ctrl. Solve this by re-capturing the sq->ctrl after all inflight request has completed, where for sure sq->ctrl reference is final, and move forward based on that. This issue was observed in an environment with many hosts connecting multiple ctrls simoutanuosly, creating a delay in allocating a ctrl leading up to this race window. Reported-by: Alex Turin Signed-off-by: Sagi Grimberg Reviewed-by: Christoph Hellwig Signed-off-by: Keith Busch Signed-off-by: Sasha Levin

nvme: adjust multiples of NVME_CTRL_PAGE_SIZE in offset

2024-07-11T10:49:20+00:00

[ Upstream commit 1bd293fcf3af84674e82ed022c049491f3768840 ] bio_vec start offset may be relatively large particularly when large folio gets added to the bio. A bigger offset will result in avoiding the single-segment mapping optimization and end up using expensive mempool_alloc further. Rather than using absolute value, adjust bv_offset by NVME_CTRL_PAGE_SIZE while checking if segment can be fitted into one/two PRP entries. Suggested-by: Christoph Hellwig Signed-off-by: Kundan Kumar Reviewed-by: Sagi Grimberg Reviewed-by: Christoph Hellwig Signed-off-by: Keith Busch Signed-off-by: Sasha Levin

nvme-multipath: find NUMA path only for online numa-node

2024-07-11T10:49:20+00:00

[ Upstream commit d3a043733f25d743f3aa617c7f82dbcb5ee2211a ] In current native multipath design when a shared namespace is created, we loop through each possible numa-node, calculate the NUMA distance of that node from each nvme controller and then cache the optimal IO path for future reference while sending IO. The issue with this design is that we may refer to the NUMA distance table for an offline node which may not be populated at the time and so we may inadvertently end up finding and caching a non-optimal path for IO. Then latter when the corresponding numa-node becomes online and hence the NUMA distance table entry for that node is created, ideally we should re-calculate the multipath node distance for the newly added node however that doesn't happen unless we rescan/reset the controller. So essentially, we may keep using non-optimal IO path for a node which is made online after namespace is created. This patch helps fix this issue ensuring that when a shared namespace is created, we calculate the multipath node distance for each online numa-node instead of each possible numa-node. Then latter when a node becomes online and we receive any IO on that newly added node, we would calculate the multipath node distance for newly added node but this time NUMA distance table would have been already populated for newly added node. Hence we would be able to correctly calculate the multipath node distance and choose the optimal path for the IO. Signed-off-by: Nilay Shroff Reviewed-by: Christoph Hellwig Signed-off-by: Keith Busch Signed-off-by: Sasha Levin

nvmet-passthru: propagate status from id override functions

2024-06-21T12:38:35+00:00

[ Upstream commit d76584e53f4244dbc154bec447c3852600acc914 ] The id override functions return a status which is not propagated to the caller. Fixes: c1fef73f793b ("nvmet: add passthru code to process commands") Signed-off-by: Daniel Wagner Reviewed-by: Chaitanya Kulkarni Reviewed-by: Christoph Hellwig Signed-off-by: Keith Busch Signed-off-by: Sasha Levin

nvme: fix nvme_pr_* status code parsing

2024-06-21T12:38:29+00:00

[ Upstream commit b1a1fdd7096dd2d67911b07f8118ff113d815db4 ] Fix the parsing if extra status bits (e.g. MORE) is present. Fixes: 7fb42780d06c ("nvme: Convert NVMe errors to PR errors") Signed-off-by: Weiwen Hu Signed-off-by: Keith Busch Signed-off-by: Sasha Levin

nvmet: fix ns enable/disable possible hang

2024-06-12T09:12:53+00:00

[ Upstream commit f97914e35fd98b2b18fb8a092e0a0799f73afdfe ] When disabling an nvmet namespace, there is a period where the subsys->lock is released, as the ns disable waits for backend IO to complete, and the ns percpu ref to be properly killed. The original intent was to avoid taking the subsystem lock for a prolong period as other processes may need to acquire it (for example new incoming connections). However, it opens up a window where another process may come in and enable the ns, (re)intiailizing the ns percpu_ref, causing the disable sequence to hang. Solve this by taking the global nvmet_config_sem over the entire configfs enable/disable sequence. Fixes: a07b4970f464 ("nvmet: add a generic NVMe target") Signed-off-by: Sagi Grimberg Reviewed-by: Christoph Hellwig Reviewed-by: Chaitanya Kulkarni Signed-off-by: Keith Busch Signed-off-by: Sasha Levin

nvme-multipath: fix io accounting on failover

2024-06-12T09:12:53+00:00

[ Upstream commit a2e4c5f5f68dbd206f132bc709b98dea64afc3b8 ] There are io stats accounting that needs to be handled, so don't call blk_mq_end_request() directly. Use the existing nvme_end_req() helper that already handles everything. Fixes: d4d957b53d91ee ("nvme-multipath: support io stats on the mpath device") Reviewed-by: Christoph Hellwig Reviewed-by: Sagi Grimberg Signed-off-by: Keith Busch Signed-off-by: Sasha Levin

nvmet: prevent sprintf() overflow in nvmet_subsys_nsid_exists()

2024-06-12T09:11:30+00:00

[ Upstream commit d15dcd0f1a4753b57e66c64c8dc2a9779ff96aab ] The nsid value is a u32 that comes from nvmet_req_find_ns(). It's endian data and we're on an error path and both of those raise red flags. So let's make this safer. 1) Make the buffer large enough for any u32. 2) Remove the unnecessary initialization. 3) Use snprintf() instead of sprintf() for even more safety. 4) The sprintf() function returns the number of bytes printed, not counting the NUL terminator. It is impossible for the return value to be <= 0 so delete that. Fixes: 505363957fad ("nvmet: fix nvme status code when namespace is disabled") Signed-off-by: Dan Carpenter Reviewed-by: Sagi Grimberg Signed-off-by: Keith Busch Signed-off-by: Sasha Levin

nvmet: fix nvme status code when namespace is disabled

2024-06-12T09:11:30+00:00

[ Upstream commit 505363957fad35f7aed9a2b0d8dad73451a80fb5 ] If the user disabled a nvmet namespace, it is removed from the subsystem namespaces list. When nvmet processes a command directed to an nsid that was disabled, it cannot differentiate between a nsid that is disabled vs. a non-existent namespace, and resorts to return NVME_SC_INVALID_NS with the dnr bit set. This translates to a non-retryable status for the host, which translates to a user error. We should expect disabled namespaces to not cause an I/O error in a multipath environment. Address this by searching a configfs item for the namespace nvmet failed to find, and if we found one, conclude that the namespace is disabled (perhaps temporarily). Return NVME_SC_INTERNAL_PATH_ERROR in this case and keep DNR bit cleared. Reported-by: Jirong Feng Tested-by: Jirong Feng Signed-off-by: Sagi Grimberg Reviewed-by: Christoph Hellwig Signed-off-by: Keith Busch Signed-off-by: Sasha Levin

nvmet-tcp: fix possible memory leak when tearing down a controller

2024-06-12T09:11:30+00:00

[ Upstream commit 6825bdde44340c5a9121f6d6fa25cc885bd9e821 ] When we teardown the controller, we wait for pending I/Os to complete (sq->ref on all queues to drop to zero) and then we go over the commands, and free their command buffers in case they are still fetching data from the host (e.g. processing nvme writes) and have yet to take a reference on the sq. However, we may miss the case where commands have failed before executing and are queued for sending a response, but will never occur because the queue socket is already down. In this case we may miss deallocating command buffers. Solve this by freeing all commands buffers as nvmet_tcp_free_cmd_buffers is idempotent anyways. Reported-by: Yi Zhang Tested-by: Yi Zhang Signed-off-by: Sagi Grimberg Reviewed-by: Christoph Hellwig Signed-off-by: Keith Busch Signed-off-by: Sasha Levin