mainline inclusion from mainline-v6.10 commit a8f62f50b4e4ea92a938fca2ec1bd108d7f210e9 category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/IB81UH

Reference: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?i...

-------------------------------------------------

The virtiofs filesystem is mounted using a "tag" which is exported by the virtiofs device:

# mount -t virtiofs <tag> /mnt

The virtiofs driver knows about all the available tags but these are currently not exported to user space.

People have asked for these tags to be exported to user space. Most recently Lennart Poettering has asked for it as he wants to scan the tags and mount virtiofs automatically in certain cases.

https://gitlab.com/virtio-fs/virtiofsd/-/issues/128

This patch exports tags at /sys/fs/virtiofs/<N>/tag where N is the id of the virtiofs device. The filesystem tag can be obtained by reading this "tag" file.

There is also a symlink at /sys/fs/virtiofs/<N>/device that points to the virtiofs device that exports this tag.

This patch converts the existing struct virtio_fs into a full kobject. It already had a refcount so it's an easy change. The virtio_fs objects can then be exposed in a kset at /sys/fs/virtiofs/. Note that virtio_fs objects may live slightly longer than we wish for them to be exposed to userspace, so kobject_del() is called explicitly when the underlying virtio_device is removed. The virtio_fs object is freed when all references are dropped (e.g. active mounts) but disappears as soon as the virtiofs device is gone.

Originally-by: Vivek Goyal vgoyal@redhat.com Signed-off-by: Stefan Hajnoczi stefanha@redhat.com Reviewed-by: Vivek Goyal vgoyal@redhat.com Signed-off-by: Miklos Szeredi mszeredi@redhat.com Signed-off-by: Shuai Wu wushuai51@huawei.com --- Documentation/ABI/testing/sysfs-fs-virtiofs | 11 ++ fs/fuse/virtio_fs.c | 112 ++++++++++++++++---- 2 files changed, 101 insertions(+), 22 deletions(-) create mode 100644 Documentation/ABI/testing/sysfs-fs-virtiofs

diff --git a/Documentation/ABI/testing/sysfs-fs-virtiofs b/Documentation/ABI/testing/sysfs-fs-virtiofs new file mode 100644 index 000000000..4839dbce9 --- /dev/null +++ b/Documentation/ABI/testing/sysfs-fs-virtiofs @@ -0,0 +1,11 @@ +What: /sys/fs/virtiofs/<n>/tag +Date: Feb 2024 +Contact: virtio-fs@lists.linux.dev +Description: + [RO] The mount "tag" that can be used to mount this filesystem. + +What: /sys/fs/virtiofs/<n>/device +Date: Feb 2024 +Contact: virtio-fs@lists.linux.dev +Description: + Symlink to the virtio device that exports this filesystem. diff --git a/fs/fuse/virtio_fs.c b/fs/fuse/virtio_fs.c index d84dacbdc..99b6113bb 100644 --- a/fs/fuse/virtio_fs.c +++ b/fs/fuse/virtio_fs.c @@ -31,6 +31,9 @@ static DEFINE_MUTEX(virtio_fs_mutex); static LIST_HEAD(virtio_fs_instances);

+/* The /sys/fs/virtio_fs/ kset */ +static struct kset *virtio_fs_kset; + enum { VQ_HIPRIO, VQ_REQUEST @@ -55,7 +58,7 @@ struct virtio_fs_vq {

/* A virtio-fs device instance */ struct virtio_fs { - struct kref refcount; + struct kobject kobj; struct list_head list; /* on virtio_fs_instances */ char *tag; struct virtio_fs_vq *vqs; @@ -161,18 +164,40 @@ static inline void dec_in_flight_req(struct virtio_fs_vq *fsvq) complete(&fsvq->in_flight_zero); }

-static void release_virtio_fs_obj(struct kref *ref) +static ssize_t tag_show(struct kobject *kobj, + struct kobj_attribute *attr, char *buf) +{ + struct virtio_fs *fs = container_of(kobj, struct virtio_fs, kobj); + + return sysfs_emit(buf, fs->tag); +} + +static struct kobj_attribute virtio_fs_tag_attr = __ATTR_RO(tag); + +static struct attribute *virtio_fs_attrs[] = { + &virtio_fs_tag_attr.attr, + NULL +}; +ATTRIBUTE_GROUPS(virtio_fs); + +static void virtio_fs_ktype_release(struct kobject *kobj) { - struct virtio_fs *vfs = container_of(ref, struct virtio_fs, refcount); + struct virtio_fs *vfs = container_of(kobj, struct virtio_fs, kobj);

kfree(vfs->vqs); kfree(vfs); }

+static const struct kobj_type virtio_fs_ktype = { + .release = virtio_fs_ktype_release, + .sysfs_ops = &kobj_sysfs_ops, + .default_groups = virtio_fs_groups, +}; + /* Make sure virtiofs_mutex is held */ static void virtio_fs_put(struct virtio_fs *fs) { - kref_put(&fs->refcount, release_virtio_fs_obj); + kobject_put(&fs->kobj); }

static void virtio_fs_fiq_release(struct fuse_iqueue *fiq) @@ -243,25 +268,44 @@ static void virtio_fs_start_all_queues(struct virtio_fs *fs) }

/* Add a new instance to the list or return -EEXIST if tag name exists*/ -static int virtio_fs_add_instance(struct virtio_fs *fs) +static int virtio_fs_add_instance(struct virtio_device *vdev, + struct virtio_fs *fs) { struct virtio_fs *fs2; - bool duplicate = false; + int ret;

mutex_lock(&virtio_fs_mutex);

list_for_each_entry(fs2, &virtio_fs_instances, list) { - if (strcmp(fs->tag, fs2->tag) == 0) - duplicate = true; + if (strcmp(fs->tag, fs2->tag) == 0) { + mutex_unlock(&virtio_fs_mutex); + return -EEXIST; + } + } + + /* Use the virtio_device's index as a unique identifier, there is no + * need to allocate our own identifiers because the virtio_fs instance + * is only visible to userspace as long as the underlying virtio_device + * exists. + */ + fs->kobj.kset = virtio_fs_kset; + ret = kobject_add(&fs->kobj, NULL, "%d", vdev->index); + if (ret < 0) { + mutex_unlock(&virtio_fs_mutex); + return ret; + } + + ret = sysfs_create_link(&fs->kobj, &vdev->dev.kobj, "device"); + if (ret < 0) { + kobject_del(&fs->kobj); + mutex_unlock(&virtio_fs_mutex); + return ret; }

- if (!duplicate) - list_add_tail(&fs->list, &virtio_fs_instances); + list_add_tail(&fs->list, &virtio_fs_instances);

mutex_unlock(&virtio_fs_mutex);

- if (duplicate) - return -EEXIST; return 0; }

@@ -274,7 +318,7 @@ static struct virtio_fs *virtio_fs_find_instance(const char *tag)

list_for_each_entry(fs, &virtio_fs_instances, list) { if (strcmp(fs->tag, tag) == 0) { - kref_get(&fs->refcount); + kobject_get(&fs->kobj); goto found; } } @@ -875,7 +919,7 @@ static int virtio_fs_probe(struct virtio_device *vdev) fs = kzalloc(sizeof(*fs), GFP_KERNEL); if (!fs) return -ENOMEM; - kref_init(&fs->refcount); + kobject_init(&fs->kobj, &virtio_fs_ktype); vdev->priv = fs;

ret = virtio_fs_read_tag(vdev, fs); @@ -897,7 +941,7 @@ static int virtio_fs_probe(struct virtio_device *vdev) */ virtio_device_ready(vdev);

- ret = virtio_fs_add_instance(fs); + ret = virtio_fs_add_instance(vdev, fs); if (ret < 0) goto out_vqs;

@@ -906,11 +950,10 @@ static int virtio_fs_probe(struct virtio_device *vdev) out_vqs: virtio_reset_device(vdev); virtio_fs_cleanup_vqs(vdev); - kfree(fs->vqs);

out: vdev->priv = NULL; - kfree(fs); + kobject_put(&fs->kobj); return ret; }

@@ -934,6 +977,8 @@ static void virtio_fs_remove(struct virtio_device *vdev) mutex_lock(&virtio_fs_mutex); /* This device is going away. No one should get new reference */ list_del_init(&fs->list); + sysfs_remove_link(&fs->kobj, "device"); + kobject_del(&fs->kobj); virtio_fs_stop_all_queues(fs); virtio_fs_drain_all_queues_locked(fs); virtio_reset_device(vdev); @@ -1520,21 +1565,43 @@ static struct file_system_type virtio_fs_type = { .kill_sb = virtio_kill_sb, };

+static int __init virtio_fs_sysfs_init(void) +{ + virtio_fs_kset = kset_create_and_add("virtiofs", NULL, fs_kobj); + if (!virtio_fs_kset) + return -ENOMEM; + return 0; +} + +static void __exit virtio_fs_sysfs_exit(void) +{ + kset_unregister(virtio_fs_kset); + virtio_fs_kset = NULL; +} + static int __init virtio_fs_init(void) { int ret;

- ret = register_virtio_driver(&virtio_fs_driver); + ret = virtio_fs_sysfs_init(); if (ret < 0) return ret;

+ ret = register_virtio_driver(&virtio_fs_driver); + if (ret < 0) + goto sysfs_exit; + ret = register_filesystem(&virtio_fs_type); - if (ret < 0) { - unregister_virtio_driver(&virtio_fs_driver); - return ret; - } + if (ret < 0) + goto unregister_virtio_driver;

return 0; + +unregister_virtio_driver: + unregister_virtio_driver(&virtio_fs_driver); +sysfs_exit: + virtio_fs_sysfs_exit(); + return ret; } module_init(virtio_fs_init);

@@ -1542,6 +1609,7 @@ static void __exit virtio_fs_exit(void) { unregister_filesystem(&virtio_fs_type); unregister_virtio_driver(&virtio_fs_driver); + virtio_fs_sysfs_exit(); } module_exit(virtio_fs_exit);

mainline inclusion from mainline-v6.10 commit 96d88f65adfbcaca153afd7d3e20d74ba379c599 category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/IB81UH

Reference: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?i...

-------------------------------------------------

The internal tag string doesn't contain a newline. Append one when emitting the tag via sysfs.

[Stefan] Orthogonal to the newline issue, sysfs_emit(buf, "%s", fs->tag) is needed to prevent format string injection.

Signed-off-by: Brian Foster bfoster@redhat.com Fixes: a8f62f50b4e4 ("virtiofs: export filesystem tags through sysfs") Signed-off-by: Miklos Szeredi mszeredi@redhat.com Signed-off-by: Shuai Wu wushuai51@huawei.com --- fs/fuse/virtio_fs.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/fuse/virtio_fs.c b/fs/fuse/virtio_fs.c index 62a446037..5083e3f8f 100644 --- a/fs/fuse/virtio_fs.c +++ b/fs/fuse/virtio_fs.c @@ -169,7 +169,7 @@ static ssize_t tag_show(struct kobject *kobj, { struct virtio_fs *fs = container_of(kobj, struct virtio_fs, kobj);

- return sysfs_emit(buf, fs->tag); + return sysfs_emit(buf, "%s\n", fs->tag); }

static struct kobj_attribute virtio_fs_tag_attr = __ATTR_RO(tag);

mainline inclusion from mainline-v6.10 commit 529395d2ae6456c556405016ea0c43081fe607f3 category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/IB81UH

Reference: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?i...

-------------------------------------------------

This commit creates a multi-queue mapping at device bring-up. The driver first attempts to use the existing MSI-X interrupt affinities (previously disabled), and if not present, will distribute the request queues evenly over the CPUs. If the latter fails as well, all CPUs are mapped to request queue zero.

When a request is handed from FUSE to the virtio-fs device driver, the driver will use the current CPU to index into the multi-queue mapping and determine the optimal request queue to use.

We measured the performance of this patch with the fio benchmarking tool, increasing the number of queues results in a significant speedup for both read and write operations, demonstrating the effectiveness of multi-queue support.

Host: - Dell PowerEdge R760 - CPU: Intel(R) Xeon(R) Gold 6438M, 128 cores - VM: KVM with 32 cores Virtio-fs device: - BlueField-3 DPU - CPU: ARM Cortex-A78AE, 16 cores - One thread per queue, each busy polling on one request queue - Each queue is 1024 descriptors deep Workload: - fio, sequential read or write, ioengine=libaio, numjobs=32, 4GiB file per job, iodepth=8, bs=256KiB, runtime=30s Performance Results: +===========================+==========+===========+ | Number of queues | Fio read | Fio write | +===========================+==========+===========+ | 1 request queue (GiB/s) | 6.1 | 4.6 | +---------------------------+----------+-----------+ | 8 request queues (GiB/s) | 25.8 | 10.3 | +---------------------------+----------+-----------+ | 16 request queues (GiB/s) | 30.9 | 19.5 | +---------------------------+----------+-----------+ | 32 request queue (GiB/s) | 33.2 | 22.6 | +---------------------------+----------+-----------+ | Speedup | 5.5x | 5x | +---------------=-----------+----------+-----------+

Signed-off-by: Peter-Jan Gootzen pgootzen@nvidia.com Signed-off-by: Yoray Zack yorayz@nvidia.com Signed-off-by: Max Gurtovoy mgurtovoy@nvidia.com Reviewed-by: Stefan Hajnoczi stefanha@redhat.com Acked-by: Michael S. Tsirkin mst@redhat.com Signed-off-by: Miklos Szeredi mszeredi@redhat.com Signed-off-by: Shuai Wu wushuai51@huawei.com --- fs/fuse/virtio_fs.c | 70 +++++++++++++++++++++++++++++++++++++++------ 1 file changed, 62 insertions(+), 8 deletions(-)

diff --git a/fs/fuse/virtio_fs.c b/fs/fuse/virtio_fs.c index eb5581325..18e2965aa 100644 --- a/fs/fuse/virtio_fs.c +++ b/fs/fuse/virtio_fs.c @@ -7,6 +7,8 @@ #include <linux/fs.h> #include <linux/dax.h> #include <linux/pci.h> +#include <linux/interrupt.h> +#include <linux/group_cpus.h> #include <linux/pfn_t.h> #include <linux/memremap.h> #include <linux/module.h> @@ -66,6 +68,8 @@ struct virtio_fs { unsigned int num_request_queues; /* number of request queues */ struct dax_device *dax_dev;

+ unsigned int *mq_map; /* index = cpu id, value = request vq id */ + /* DAX memory window where file contents are mapped */ void *window_kaddr; phys_addr_t window_phys_addr; @@ -184,6 +188,7 @@ static void virtio_fs_ktype_release(struct kobject *kobj) { struct virtio_fs *vfs = container_of(kobj, struct virtio_fs, kobj);

+ kfree(vfs->mq_map); kfree(vfs->vqs); kfree(vfs); } @@ -705,6 +710,44 @@ static void virtio_fs_requests_done_work(struct work_struct *work) } }

+static void virtio_fs_map_queues(struct virtio_device *vdev, struct virtio_fs *fs) +{ + const struct cpumask *mask, *masks; + unsigned int q, cpu; + + /* First attempt to map using existing transport layer affinities + * e.g. PCIe MSI-X + */ + if (!vdev->config->get_vq_affinity) + goto fallback; + + for (q = 0; q < fs->num_request_queues; q++) { + mask = vdev->config->get_vq_affinity(vdev, VQ_REQUEST + q); + if (!mask) + goto fallback; + + for_each_cpu(cpu, mask) + fs->mq_map[cpu] = q; + } + + return; +fallback: + /* Attempt to map evenly in groups over the CPUs */ + masks = group_cpus_evenly(fs->num_request_queues); + /* If even this fails we default to all CPUs use queue zero */ + if (!masks) { + for_each_possible_cpu(cpu) + fs->mq_map[cpu] = 0; + return; + } + + for (q = 0; q < fs->num_request_queues; q++) { + for_each_cpu(cpu, &masks[q]) + fs->mq_map[cpu] = q; + } + kfree(masks); +} + /* Virtqueue interrupt handler */ static void virtio_fs_vq_done(struct virtqueue *vq) { @@ -741,6 +784,11 @@ static int virtio_fs_setup_vqs(struct virtio_device *vdev, { struct virtqueue **vqs; vq_callback_t **callbacks; + /* Specify pre_vectors to ensure that the queues before the + * request queues (e.g. hiprio) don't claim any of the CPUs in + * the multi-queue mapping and interrupt affinities + */ + struct irq_affinity desc = { .pre_vectors = VQ_REQUEST }; const char **names; unsigned int i; int ret = 0; @@ -762,7 +810,9 @@ static int virtio_fs_setup_vqs(struct virtio_device *vdev, callbacks = kmalloc_array(fs->nvqs, sizeof(callbacks[VQ_HIPRIO]), GFP_KERNEL); names = kmalloc_array(fs->nvqs, sizeof(names[VQ_HIPRIO]), GFP_KERNEL); - if (!vqs || !callbacks || !names) { + fs->mq_map = kcalloc_node(nr_cpu_ids, sizeof(*fs->mq_map), GFP_KERNEL, + dev_to_node(&vdev->dev)); + if (!vqs || !callbacks || !names || !fs->mq_map) { ret = -ENOMEM; goto out; } @@ -782,7 +832,7 @@ static int virtio_fs_setup_vqs(struct virtio_device *vdev, names[i] = fs->vqs[i].name; }

- ret = virtio_find_vqs(vdev, fs->nvqs, vqs, callbacks, names, NULL); + ret = virtio_find_vqs(vdev, fs->nvqs, vqs, callbacks, names, &desc); if (ret < 0) goto out;

@@ -794,8 +844,10 @@ static int virtio_fs_setup_vqs(struct virtio_device *vdev, kfree(names); kfree(callbacks); kfree(vqs); - if (ret) + if (ret) { kfree(fs->vqs); + kfree(fs->mq_map); + } return ret; }

@@ -935,7 +987,7 @@ static int virtio_fs_probe(struct virtio_device *vdev) if (ret < 0) goto out;

- /* TODO vq affinity */ + virtio_fs_map_queues(vdev, fs);

ret = virtio_fs_setup_dax(vdev, fs); if (ret < 0) @@ -1284,7 +1336,7 @@ static int virtio_fs_enqueue_req(struct virtio_fs_vq *fsvq, static void virtio_fs_wake_pending_and_unlock(struct fuse_iqueue *fiq) __releases(fiq->lock) { - unsigned int queue_id = VQ_REQUEST; /* TODO multiqueue */ + unsigned int queue_id; struct virtio_fs *fs; struct fuse_req *req; struct virtio_fs_vq *fsvq; @@ -1298,11 +1350,13 @@ __releases(fiq->lock) spin_unlock(&fiq->lock);

fs = fiq->priv; + queue_id = VQ_REQUEST + fs->mq_map[raw_smp_processor_id()];

- pr_debug("%s: opcode %u unique %#llx nodeid %#llx in.len %u out.len %u\n", - __func__, req->in.h.opcode, req->in.h.unique, + pr_debug("%s: opcode %u unique %#llx nodeid %#llx in.len %u out.len %u queue_id %u\n", + __func__, req->in.h.opcode, req->in.h.unique, req->in.h.nodeid, req->in.h.len, - fuse_len_args(req->args->out_numargs, req->args->out_args)); + fuse_len_args(req->args->out_numargs, req->args->out_args), + queue_id);

fsvq = &fs->vqs[queue_id]; ret = virtio_fs_enqueue_req(fsvq, req, false);