Commits · 0day-ci/linux

We can't split bio with more than BIO_MAX_VECS sectors, otherwise the below call trace was triggered because we could allocate oversized write behind bio later. [ 8.097936] bvec_alloc+0x90/0xc0 [ 8.098934] bio_alloc_bioset+0x1b3/0x260 [ 8.099959] raid1_make_request+0x9ce/0xc50 [raid1] [ 8.100988] ? __bio_clone_fast+0xa8/0xe0 [ 8.102008] md_handle_request+0x158/0x1d0 [md_mod] [ 8.103050] md_submit_bio+0xcd/0x110 [md_mod] [ 8.104084] submit_bio_noacct+0x139/0x530 [ 8.105127] submit_bio+0x78/0x1d0 [ 8.106163] ext4_io_submit+0x48/0x60 [ext4] [ 8.107242] ext4_writepages+0x652/0x1170 [ext4] [ 8.108300] ? do_writepages+0x41/0x100 [ 8.109338] ? __ext4_mark_inode_dirty+0x240/0x240 [ext4] [ 8.110406] do_writepages+0x41/0x100 [ 8.111450] __filemap_fdatawrite_range+0xc5/0x100 [ 8.112513] file_write_and_wait_range+0x61/0xb0 [ 8.113564] ext4_sync_file+0x73/0x370 [ext4] [ 8.114607] __x64_sys_fsync+0x33/0x60 [ 8.115635] do_syscall_64+0x33/0x40 [ 8.116670] entry_SYSCALL_64_after_hwframe+0x44/0xae [1]. https://bugs.archlinux.org/task/70992 Reported-by: Jens Stutte <jens@chianterastutte.eu> Tested-by: Jens Stutte <jens@chianterastutte.eu> Signed-off-by: Guoqing Jiang <jiangguoqing@kylinos.cn>

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>

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: Song Liu <song@kernel.org>

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: Christoph Hellwig <hch@lst.de> Signed-off-by: Guoqing Jiang <jiangguoqing@kylinos.cn> Signed-off-by: Song Liu <song@kernel.org>

The attribute_group structs are never modified, they're only passed to sysfs_create_group() and sysfs_remove_group(). Make them const to allow the compiler to put them in read-only memory. Signed-off-by: Rikard Falkeborn <rikard.falkeborn@gmail.com> Signed-off-by: Song Liu <song@kernel.org>

Mark the three personalities (linear, fault and multipath) as deprecated because: 1. people can use dm multipath or nvme multipath. 2. linear is already deprecated in MODULE_ALIAS. 3. no one actively using fault. Signed-off-by: Guoqing Jiang <jiangguoqing@kylinos.cn> Signed-off-by: Song Liu <song@kernel.org>

For raid10, we record the start time between split bio and clone bio, and finish the accounting in the final endio. Also introduce start_time in r10bio accordingly. Signed-off-by: Guoqing Jiang <jiangguoqing@kylinos.cn> Signed-off-by: Song Liu <song@kernel.org>

For raid1, we record the start time between split bio and clone bio, and finish the accounting in the final endio. Also introduce start_time in r1bio accordingly. Signed-off-by: Guoqing Jiang <jiangguoqing@kylinos.cn> Signed-off-by: Song Liu <song@kernel.org>

The caller of raid1_read_request could pass NULL or a valid pointer for "struct r1bio *r1_bio", so it actually means whether r1_bio is existed or not. Signed-off-by: Guoqing Jiang <jiangguoqing@kylinos.cn> Signed-off-by: Song Liu <song@kernel.org>

After enable io accounting, chunk read bio could be cloned twice which is not good. To avoid such inefficiency, let's clone align_bio from io_acct_set too, then we need only call md_account_bio in make_request unconditionally. Signed-off-by: Guoqing Jiang <jiangguoqing@kylinos.cn> Signed-off-by: Song Liu <song@kernel.org>

…hunk We don't need to clone bio if the relevant region has badblock. Signed-off-by: Guoqing Jiang <jiangguoqing@kylinos.cn> Signed-off-by: Song Liu <song@kernel.org>

We introduce a new bioset (io_acct_set) for raid0 and raid5 since they don't own clone infrastructure to accounting io. And the bioset is added to mddev instead of to raid0 and raid5 layer, because with this way, we can put common functions to md.h and reuse them in raid0 and raid5. Also struct md_io_acct is added accordingly which includes io start_time, the origin bio and cloned bio. Then we can call bio_{start,end}_io_acct to get related io status. Signed-off-by: Guoqing Jiang <jiangguoqing@kylinos.cn> Signed-off-by: Song Liu <song@kernel.org>

The commit 41d2d84 ("md: improve io stats accounting") could cause double fault problem per the report [1], and also it is not correct to change ->bi_end_io if md don't own it, so let's revert it. And io stats accounting will be replemented in later commits. [1]. https://lore.kernel.org/linux-raid/3bf04253-3fad-434a-63a7-20214e38cf26@gmail.com/T/#t Fixes: 41d2d84 ("md: improve io stats accounting") Signed-off-by: Guoqing Jiang <jiangguoqing@kylinos.cn> Signed-off-by: Song Liu <song@kernel.org>

pstore-blk just pokes directly into the pagecache for the block device without going through the file operations for that by faking up it's own file operations that do not match the block device ones. As this breaks the control of the block layer of it's page cache, and even now just works by accident only the best thing is to just disable this driver. Fixes: 17639f6 ("pstore/blk: Introduce backend for block devices") Signed-off-by: Christoph Hellwig <hch@lst.de> Link: https://lore.kernel.org/r/20210608161327.1537919-1-hch@lst.de Signed-off-by: Jens Axboe <axboe@kernel.dk>

Fixes scripts/checkpatch.pl warning: WARNING: Possible unnecessary 'out of memory' message Remove it can help us save a bit of memory. Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>

… for-5.14/drivers Pull NVMe updates from Christoph: "nvme updates for Linux 5.14 - improve the APST configuration algorithm (Alexey Bogoslavsky) - look for StorageD3Enable on companion ACPI device (Mario Limonciello) - allow selecting the network interface for TCP connections (Martin Belanger) - misc cleanups (Amit Engel, Chaitanya Kulkarni, Colin Ian King, me)" * tag 'nvme-5.14-2021-06-08' of git://git.infradead.org/nvme: nvmet: remove a superfluous variable nvmet: move ka_work initialization to nvmet_alloc_ctrl nvme: remove nvme_{get,put}_ns_from_disk nvme: split nvme_report_zones nvme: move the CSI sanity check into nvme_ns_report_zones nvme: add a sparse annotation to nvme_ns_head_ctrl_ioctl nvme: open code nvme_put_ns_from_disk in nvme_ns_head_ctrl_ioctl nvme: open code nvme_{get,put}_ns_from_disk in nvme_ns_head_ioctl nvme: open code nvme_put_ns_from_disk in nvme_ns_head_chr_ioctl nvme-fabrics: remove extra braces nvme-fabrics: remove an extra comment nvme-fabrics: remove extra new lines in the switch nvme-fabrics: fix the kerneldco comment for nvmf_log_connect_error() nvme-tcp: allow selecting the network interface for connections nvme-pci: look for StorageD3Enable on companion ACPI device instead nvme: extend and modify the APST configuration algorithm nvme: remove redundant initialization of variable ret

Remove the superfluous variable "bdev" that is only used once in the nvmet_bdev_alloc_bip() and use req->ns->bdev that is used everywhere in the code to access the nvmet request's bdev. Signed-off-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com> Signed-off-by: Christoph Hellwig <hch@lst.de>

Initialize keep-alive work only once, as part of alloc_ctrl and not each time that nvmet_start_keep_alive_timer is being called Signed-off-by: Amit Engel <amit.engel@dell.com> Reviewed-by: Hou Pu <houpu.main@gmail.com>

Now that only one caller is left remove the helpers by restructuring nvme_pr_command so that it has two helpers for sending a command of to a given nsid using either the ns_head for multipath, or the namespace stored in the gendisk. Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Keith Busch <kbusch@kernel.org> Reviewed-by: Sagi Grimberg <sagi@grimberg.me>

Split multipath support out of nvme_report_zones into a separate helper and simplify the non-multipath version as a result. Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Keith Busch <kbusch@kernel.org> Reviewed-by: Sagi Grimberg <sagi@grimberg.me>

Move the CSI check into nvme_ns_report_zones to clean up the code a little bit and prepare for further refactoring. Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Keith Busch <kbusch@kernel.org> Reviewed-by: Sagi Grimberg <sagi@grimberg.me>

Add the __releases annotation to tell sparse that nvme_ns_head_ctrl_ioctl is expected to unlock head->srcu. Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Keith Busch <kbusch@kernel.org> Reviewed-by: Sagi Grimberg <sagi@grimberg.me>

nvme_ns_head_ctrl_ioctl is always used on multipath nodes, so just call srcu_read_unlock directly. Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Keith Busch <kbusch@kernel.org> Reviewed-by: Sagi Grimberg <sagi@grimberg.me>

nvme_ns_head_ioctl is always used on multipath nodes, no need to deal with the de-multiplexers. Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Keith Busch <kbusch@kernel.org> Reviewed-by: Sagi Grimberg <sagi@grimberg.me>

nvme_ns_head_chr_ioctl is always used on multipath nodes, so just call srcu_read_unlock and consolidate the two unlock paths. Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Keith Busch <kbusch@kernel.org> Reviewed-by: Sagi Grimberg <sagi@grimberg.me>

No need to use the braces around ~ operator. No functionality change in this patch. Signed-off-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com> Signed-off-by: Christoph Hellwig <hch@lst.de>

Remove the comment at the end of the switch that is not needed as function is small enough. No functionality change in this patch. Signed-off-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com> Signed-off-by: Christoph Hellwig <hch@lst.de>

Remove the extra lines in the switch block that is not common practice in the kernel code. No functionality change in this patch. Signed-off-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com> Signed-off-by: Christoph Hellwig <hch@lst.de>

Fix the comment style that matches existing code. No functionality change in this patch. Signed-off-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com> Signed-off-by: Christoph Hellwig <hch@lst.de>

In our application, we need a way to force TCP connections to go out a specific IP interface instead of letting Linux select the interface based on the routing tables. Add the 'host-iface' option to allow specifying the interface to use. When the option host-iface is specified, the driver uses the specified interface to set the option SO_BINDTODEVICE on the TCP socket before connecting. This new option is needed in addtion to the existing host-traddr for the following reasons: Specifying an IP interface by its associated IP address is less intuitive than specifying the actual interface name and, in some cases, simply doesn't work. That's because the association between interfaces and IP addresses is not predictable. IP addresses can be changed or can change by themselves over time (e.g. DHCP). Interface names are predictable [1] and will persist over time. Consider the following configuration. 1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state ... link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00 inet 100.0.0.100/24 scope global lo valid_lft forever preferred_lft forever 2: enp0s3: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc ... link/ether 08:00:27:21:65:ec brd ff:ff:ff:ff:ff:ff inet 100.0.0.100/24 scope global enp0s3 valid_lft forever preferred_lft forever 3: enp0s8: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc ... link/ether 08:00:27:4f:95:5c brd ff:ff:ff:ff:ff:ff inet 100.0.0.100/24 scope global enp0s8 valid_lft forever preferred_lft forever The above is a VM that I configured with the same IP address (100.0.0.100) on all interfaces. Doing a reverse lookup to identify the unique interface associated with 100.0.0.100 does not work here. And this is why the option host_iface is required. I understand that the above config does not represent a standard host system, but I'm using this to prove a point: "We can never know how users will configure their systems". By te way, The above configuration is perfectly fine by Linux. The current TCP implementation for host_traddr performs a bind()-before-connect(). This is a common construct to set the source IP address on a TCP socket before connecting. This has no effect on how Linux selects the interface for the connection. That's because Linux uses the Weak End System model as described in RFC1122 [2]. On the other hand, setting the Source IP Address has benefits and should be supported by linux-nvme. In fact, setting the Source IP Address is a mandatory FedGov requirement (e.g. connection to a RADIUS/TACACS+ server). Consider the following configuration. $ ip addr list dev enp0s8 3: enp0s8: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc ... link/ether 08:00:27:4f:95:5c brd ff:ff:ff:ff:ff:ff inet 192.168.56.101/24 brd 192.168.56.255 scope global enp0s8 valid_lft 426sec preferred_lft 426sec inet 192.168.56.102/24 scope global secondary enp0s8 valid_lft forever preferred_lft forever inet 192.168.56.103/24 scope global secondary enp0s8 valid_lft forever preferred_lft forever inet 192.168.56.104/24 scope global secondary enp0s8 valid_lft forever preferred_lft forever Here we can see that several addresses are associated with interface enp0s8. By default, Linux always selects the default IP address, 192.168.56.101, as the source address when connecting over interface enp0s8. Some users, however, want the ability to specify a different source address (e.g., 192.168.56.102, 192.168.56.103, ...). The option host_traddr can be used as-is to perform this function. In conclusion, I believe that we need 2 options for TCP connections. One that can be used to specify an interface (host-iface). And one that can be used to set the source address (host-traddr). Users should be allowed to use one or the other, or both, or none. Of course, the documentation for host_traddr will need some clarification. It should state that when used for TCP connection, this option only sets the source address. And the documentation for host_iface should say that this option is only available for TCP connections. References: [1] https://www.freedesktop.org/wiki/Software/systemd/PredictableNetworkInterfaceNames/ [2] https://tools.ietf.org/html/rfc1122 Tested both IPv4 and IPv6 connections. Signed-off-by: Martin Belanger <martin.belanger@dell.com> Reviewed-by: Sagi Grimberg <sagi@grimberg.me> Reviewed-by: Hannes Reinecke <hare@suse.de> Signed-off-by: Christoph Hellwig <hch@lst.de>

Commits on Aug 13, 2021

Commits on Jun 15, 2021

Commits on Jun 14, 2021

Commits on Jun 9, 2021

Commits on Jun 8, 2021

Commits on Jun 3, 2021