- 17 Jun, 2021 26 commits
-
-
Chaitanya Kulkarni authored
NVMe TP 4053 – Zoned Namespaces (ZNS) allows host software to communicate with a non-volatile memory subsystem using zones for NVMe protocol-based controllers. NVMeOF already support the ZNS NVMe Protocol compliant devices on the target in the passthru mode. There are generic zoned block devices like Shingled Magnetic Recording (SMR) HDDs that are not based on the NVMe protocol. This patch adds ZNS backend support for non-ZNS zoned block devices as NVMeOF targets. This support includes implementing the new command set NVME_CSI_ZNS, adding different command handlers for ZNS command set such as NVMe Identify Controller, NVMe Identify Namespace, NVMe Zone Append, NVMe Zone Management Send and NVMe Zone Management Receive. With the new command set identifier, we also update the target command effects logs to reflect the ZNS compliant commands. Signed-off-by:
Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com> Reviewed-by:
Damien Le Moal <damien.lemoal@wdc.com> Signed-off-by:
Christoph Hellwig <hch@lst.de>
-
Chaitanya Kulkarni authored
NVMe TP 4056 allows controllers to support different command sets. NVMeoF target currently only supports namespaces that contain traditional logical blocks that may be randomly read and written. In some applications there is a value in exposing namespaces that contain logical blocks that have special access rules (e.g. sequentially write required namespace such as Zoned Namespace (ZNS)). In order to support the Zoned Block Devices (ZBD) backend, controllers need to have support for ZNS Command Set Identifier (CSI). In this preparation patch, we adjust the code such that it can now support the default command set identifier. We update the namespace data structure to store the CSI value which defaults to NVME_CSI_NVM that represents traditional logical blocks namespace type. The CSI support is required to implement the ZBD backend for NVMeOF with host side NVMe ZNS interface, since ZNS commands belong to the different command set than the default one. Signed-off-by:
-
Chaitanya Kulkarni authored
In current code there exists two backends which are using inline bio optimization, that adds a duplicate code for freeing the bio. For Zoned Block Device backend we also use the same optimzation and it will lead to having duplicate code in the three backends: generic bdev, passsthru, and generic zns. Add a helper function to avoid duplicate code and update the respective backends. Signed-off-by:
-
Chaitanya Kulkarni authored
We report error and complete the request when identify cns value is not handled in nvmet_execute_identify(). This error reporting is also needed for Zone Block Device backend for NVMeOF target. Add a helper nvmet_req_cns_error_compplete() to report an error and complete the request when idenitfy command cns not handled value. Signed-off-by:
-
Chaitanya Kulkarni authored
The block layer provides emulation of zone management operations targeting all zones of a zoned block device only for the zone reset operation (REQ_OP_ZONE_RESET). In order to correctly implement exporting of zoned block devices with NVMeOF, emulating zone management operations targeting all zones of a device is also necessary for the open, close and finish zone operations (REQ_OP_ZONE_OPEN, REQ_OP_ZONE_CLOSE and REQ_OP_ZONE_FINISH). Instead of duplicating the code, export the existing helper from block layer so we can use a bio chaining pattern that is present in the block layer for REQ_OP_ZONE RESET all emulation in the NVMeOF zoned block device backend. Reviewed-by:
-
Chaitanya Kulkarni authored
In function errno_to_nvme_status() we store the value of the NVMe status into the local variable and don't do anything useful with that but just return. Remove the local variable and return the value directly from switch. This also removed extra break statements. Signed-off-by:
-
Chaitanya Kulkarni authored
There is no point in keeping the status variable that is used only once in the function nvmet_async_events_failall(). Remove the variable and use the value directly. Signed-off-by:
-
Chaitanya Kulkarni authored
In function nvmet_max_nsid() we calculate the max nsid by iterating over the XArray and store it in the variable nsid that has type of unsigned long. Since the value of this function is stored into the subsys->max_nsid which is of type u32, change the local variable nsid type and the return type of the same function to u32. Signed-off-by:
-
Chaitanya Kulkarni authored
Use u32 type for the nsid_max member of the nvmet_subsys structure. This avoids the type confusion when updating the subsys->nax_nsid from ns->nsid. This also matches the nvmet_ns->nsid member. Signed-off-by:
-
Chaitanya Kulkarni authored
The function nvmet_file_parse_io_cmd() is called from the fast path. The local variable to that function cmd is only used once. Remove the local variable and use req->cmd directly. Signed-off-by:
-
Chaitanya Kulkarni authored
The function nvmet_bdev_parse_io_cmd() is called from the fast path. The local variable to that function cmd is only used once. Remove the local variable and use req->cmd directly. Signed-off-by:
-
Noam Gottlieb authored
Once some host has connected to the nvmf target, make sure that the version number is stable and cannot be changed. Signed-off-by:
Max Gurtovoy <mgurtovoy@nvidia.com> Signed-off-by:
Noam Gottlieb <ngottlieb@nvidia.com> Signed-off-by:
-
Noam Gottlieb authored
Currently, once the subsystem's model_number is set for the first time there is no way to change it. However, as long as no connection was established to nvmf target, there is no reason for such restriction and we should allow to change the subsystem's model_number as many times as needed. In addition, in order to simplfy the changes and make the model number flow more similar to the rest of the attributes in the Identify Controller data structure, we set a default value for the model number at the initiation of the subsystem. Reviewed-by:
-
Noam Gottlieb authored
Once some host has connected to the target, make sure that the serial number is stable and cannot be changed. Reviewed-by:
-
Noam Gottlieb authored
According to the NVM specification, the serial_number should be 20 bytes (bytes 23:04 of the Identify Controller data structure), and should contain only ASCII characters. In accordance, the serial_number size is changed to 20 bytes and before any attempt to store a new value in serial_number we check that the input is valid - i.e. contains only ASCII characters, is not empty and does not exceed 20 bytes. Signed-off-by:
-
Hannes Reinecke authored
When parsing a request in nvmet_fc_handle_fcp_rqst() we should not check for invalid target ports; if we do the command is aborted from the fcp layer, causing the host to assume a transport error. Rather we should still forward this request to the nvmet layer, which will then correctly fail the command with an appropriate error status. Signed-off-by:
Hannes Reinecke <hare@suse.de> Reviewed-by:
James Smart <jsmart2021@gmail.com> Signed-off-by:
-
Chaitanya Kulkarni authored
Declare and initialize structure variable to the zero values so that we can get rid of the zeroout memset call. Signed-off-by:
-
Chaitanya Kulkarni authored
Declare and initialize structure variable to the zero values so that we can get rid of the zeroout memset call. Signed-off-by:
-
Chaitanya Kulkarni authored
Declare and initialize structure variable to the zero values so that we can get rid of the zeroout memset call. Signed-off-by:
-
Chaitanya Kulkarni authored
Declare and initialize structure variable to the zero values so that we can get rid of the zeroout memset call. Signed-off-by:
-
Chaitanya Kulkarni authored
Use the helper to check NVMe controller's SGL support. Reviewed-by:
Sagi Grimberg <sagi@grimberg.me> Signed-off-by:
-
Chaitanya Kulkarni authored
Use the helper to check NVMe controller's SGL support. Signed-off-by:
-
Chaitanya Kulkarni authored
Use the helper to check NVMe controller's SGL support. Reviewed-by:
-
Chaitanya Kulkarni authored
For various transports such as fc/tcp/pci it is common to check if NVMe SGLs are supported or not by the controller. In this preparation patch we add a helper to avoid the open coding of such checks in the various transport. Signed-off-by:
-
Chaitanya Kulkarni authored
Remove the extra white line at the end of the functions. Signed-off-by:
-
JK Kim authored
nvmeq->cq_head is compared with nvmeq->q_depth and changed the value and cq_phase for handling the next cq db. but, nvmeq->q_depth's type is u32 and max. value is 0x10000 when CQP.MSQE is 0xffff and io_queue_depth is 0x10000. current temp. variable for comparing with nvmeq->q_depth is overflowed when previous nvmeq->cq_head is 0xffff. in this case, nvmeq->cq_phase is not updated. so, fix data type for temp. variable to u32. Signed-off-by:
JK Kim <jongkang.kim2@gmail.com> Signed-off-by:
-
- 16 Jun, 2021 7 commits
-
-
Dan Carpenter authored
These error paths currently return success but they should return -EOPNOTSUPP. Fixes: 73ffcefc ("nvme-tcp: check sgl supported by target") Fixes: 3f2304f8 ("nvme-tcp: add NVMe over TCP host driver") Signed-off-by:
Dan Carpenter <dan.carpenter@oracle.com> Reviewed-by:
-
Chaitanya Kulkarni authored
Add a helper nvme_validate_passthru_nsid() to validate the nsid that removes the nsid validation and error message print code from nvme_user_cmd() and nvme_user_cmd64(). Signed-off-by:
-
Geert Uytterhoeven authored
Fix a singular/plural mismatch in the CONFIG_NVME_MULTIPATH help text. Signed-off-by:
Geert Uytterhoeven <geert+renesas@glider.be> Signed-off-by:
-
Daniel Wagner authored
Commit ce86dad2 ("nvme-multipath: reset bdev to ns head when failover") moved the reset code where the bio is added to the requeue_list for the failover path. But it left the original bio_set_dev in nvme_requeue_work. There is a second path to nvme_requee_work. It is via nvme_ns_head_submit_bio. Though we don't have to set bio->bi_bdev for this path either, as it points to the correct bdev already. Let's remove the bio_set_dev. It's updating the bio->bi_bdev with the same pointer and thus it's unnecessary. Signed-off-by:
Daniel Wagner <dwagner@suse.de> Signed-off-by:
-
Daniel Wagner authored
The controller is required to have a non-zero MNAN value if it supports ANA: If the controller supports Asymmetric Namespace Access Reporting, then this field shall be set to a non-zero value that is less than or equal to the NN value. Reviewed-by:
-
Mario Limonciello authored
AMD systems from Renoir and Lucienne require that the NVME controller is put into D3 over a Modern Standby / suspend-to-idle cycle. This is "typically" accomplished using the `StorageD3Enable` property in the _DSD, but this property was introduced after many of these systems launched and most OEM systems don't have it in their BIOS. On AMD Renoir without these drives going into D3 over suspend-to-idle the resume will fail with the NVME controller being reset and a trace like this in the kernel logs: ``` [ 83.556118] nvme nvme0: I/O 161 QID 2 timeout, aborting [ 83.556178] nvme nvme0: I/O 162 QID 2 timeout, aborting [ 83.556187] nvme nvme0: I/O 163 QID 2 timeout, aborting [ 83.556196] nvme nvme0: I/O 164 QID 2 timeout, aborting [ 95.332114] nvme nvme0: I/O 25 QID 0 timeout, reset controller [ 95.332843] nvme nvme0: Abort status: 0x371 [ 95.332852] nvme nvme0: Abort status: 0x371 [ 95.332856] nvme nvme0: Abort status: 0x371 [ 95.332859] nvme nvme0: Abort status: 0x371 [ 95.332909] PM: dpm_run_callback(): pci_pm_resume+0x0/0xe0 returns -16 [ 95.332936] nvme 0000:03:00.0: PM: failed to resume async: error -16 ``` The Microsoft documentation for StorageD3Enable mentioned that Windows has a hardcoded allowlist for D3 support, which was used for these platforms. Introduce quirks to hardcode them for Linux as well. As this property is now "standardized", OEM systems using AMD Cezanne and newer APU's have adopted this property, and quirks like this should not be necessary. CC: Shyam-sundar S-k <Shyam-sundar.S-k@amd.com> CC: Alexander Deucher <Alexander.Deucher@amd.com> CC: Prike Liang <prike.liang@amd.com> Link: https://docs.microsoft.com/en-us/windows-hardware/design/component-guidelines/power-management-for-storage-hardware-devices-introSigned-off-by:
Mario Limonciello <mario.limonciello@amd.com> Acked-by:
Rafael J. Wysocki <rafael.j.wysocki@intel.com> Tested-by:
Julian Sikorski <belegdol@gmail.com> Signed-off-by:
-
Mario Limonciello authored
Although first implemented for NVME, this check may be usable by other drivers as well. Microsoft's specification explicitly mentions that is may be usable by SATA and AHCI devices. Google also indicates that they have used this with SDHCI in a downstream kernel tree that a user can plug a storage device into. Link: https://docs.microsoft.com/en-us/windows-hardware/design/component-guidelines/power-management-for-storage-hardware-devices-introSuggested-by:
Keith Busch <kbusch@kernel.org> CC: Shyam-sundar S-k <Shyam-sundar.S-k@amd.com> CC: Alexander Deucher <Alexander.Deucher@amd.com> CC: Rafael J. Wysocki <rjw@rjwysocki.net> CC: Prike Liang <prike.liang@amd.com> Signed-off-by:
-
- 15 Jun, 2021 7 commits
-
-
Jens Axboe authored
Merge branch 'md-next' of https://git.kernel.org/pub/scm/linux/kernel/git/song/md into for-5.14/drivers Pull MD changes from Song: "1) iostats rewrite by Guoqing Jiang; 2) raid5 lock contention optimization by Gal Ofri." * 'md-next' of https://git.kernel.org/pub/scm/linux/kernel/git/song/md: md/raid5: avoid device_lock in read_one_chunk() md: add comments in md_integrity_register md: check level before create and exit io_acct_set md: Constify attribute_group structs md: mark some personalities as deprecated md/raid10: enable io accounting md/raid1: enable io accounting md/raid1: rename print_msg with r1bio_existed md/raid5: avoid redundant bio clone in raid5_read_one_chunk md/raid5: move checking badblock before clone bio in raid5_read_one_chunk md: add io accounting for raid0 and raid5 md: revert io stats accounting
-
https://github.com/evdenis/linux-floppyJens Axboe authored
Pull floppy fixes from Denis: "Floppy patches for 5.14 Two oneliners to fix clang warnings: - -Wimplicit-fallthrough warning fix from Gustavo A. R. Silva. - Redundant assignment warning fix from Jiapeng Chong. No semantic and behavioural changes." * tag 'floppy-for-5.14' of https://github.com/evdenis/linux-floppy: floppy: Fix fall-through warning for Clang floppy: cleanup: remove redundant assignment to nr_sectors
-
Gustavo A. R. Silva authored
In preparation to enable -Wimplicit-fallthrough for Clang, fix a warning by explicitly adding a break statement instead of letting the code fall through to the next case. Link: https://github.com/KSPP/linux/issues/115 Link: https://lore.kernel.org/linux-hardening/47bcd36a-6524-348b-e802-0691d1b3c429@kernel.dk/Suggested-by:
Jens Axboe <axboe@kernel.dk> Signed-off-by:
Gustavo A. R. Silva <gustavoars@kernel.org> Signed-off-by:
Denis Efremov <efremov@linux.com>
-
Jiapeng Chong authored
Variable nr_sectors is set to zero but this value is never read as it is overwritten later on, hence it is a redundant assignment and can be removed. Clean up the following clang-analyzer warning: drivers/block/floppy.c:2333:2: warning: Value stored to 'nr_sectors' is never read [clang-analyzer-deadcode.DeadStores]. Link: https://lore.kernel.org/r/1619774805-121562-1-git-send-email-jiapeng.chong@linux.alibaba.comReported-by:
Abaci Robot <abaci@linux.alibaba.com> Signed-off-by:
Jiapeng Chong <jiapeng.chong@linux.alibaba.com> Signed-off-by:
-
Gal Ofri authored
There is a lock contention on device_lock in read_one_chunk(). device_lock is taken to sync conf->active_aligned_reads and conf->quiesce. read_one_chunk() takes the lock, then waits for quiesce=0 (resumed) before incrementing active_aligned_reads. raid5_quiesce() takes the lock, sets quiesce=2 (in-progress), then waits for active_aligned_reads to be zero before setting quiesce=1 (suspended). Introduce a fast (lockless) path in read_one_chunk(): activate aligned read without taking device_lock. In case quiesce starts while activating the aligned-read in fast path, deactivate it and revert to old behavior (take device_lock and wait for quiesce to finish). Add smp store/load in raid5_quiesce()/read_one_chunk() respectively to gaurantee that read_one_chunk() does not miss an ongoing quiesce. My setups: 1. 8 local nvme drives (each up to 250k iops). 2. 8 ram disks (brd). Each setup with raid6 (6+2), 1024 io threads on a 96 cpu-cores (48 per socket) system. Record both iops and cpu spent on this contention with rand-read-4k. Record bw with sequential-read-128k. Note: in most cases cpu is still busy but due to "new" bottlenecks. nvme: | iops | cpu | bw ----------------------------------------------- without patch | 1.6M | ~50% | 5.5GB/s with patch | 2M (throttled) | 0% | 16GB/s (throttled) ram (brd): | iops | cpu | bw ----------------------------------------------- without patch | 2M | ~80% | 24GB/s with patch | 4M | 0% | 55GB/s CC: Song Liu <song@kernel.org> CC: Neil Brown <neilb@suse.de> Reviewed-by:NeilBrown <neilb@suse.de> Signed-off-by:
Gal Ofri <gal.ofri@storing.io> Signed-off-by:
Song Liu <song@kernel.org>
-
Guoqing Jiang authored
Given it is not obvious for the error handling, let's try to add some comments here to make it clear. Signed-off-by:
Guoqing Jiang <jiangguoqing@kylinos.cn> Signed-off-by:
-
Guoqing Jiang authored
The bio_set (io_acct_set) is used by personalities to clone bio and trace the timestamp of bio. Some personalities such as raid1/10 don't need the bio_set, so add check to not create it unconditionally. Also update the comment for md_account_bio to make it more clear. Suggested-by:
-
