From c28cc36469554dc55540f059fbdc7fa22a2c31fc Mon Sep 17 00:00:00 2001 From: Nick Piggin Date: Fri, 7 Jan 2011 17:49:53 +1100 Subject: fs: fs_struct use seqlock Use a seqlock in the fs_struct to enable us to take an atomic copy of the complete cwd and root paths. Use this in the RCU lookup path to avoid a thread-shared spinlock in RCU lookup operations. Multi-threaded apps may now perform path lookups with scalability matching multi-process apps. Operations such as stat(2) become very scalable for multi-threaded workload. Signed-off-by: Nick Piggin --- fs/fs_struct.c | 10 ++++++++++ 1 file changed, 10 insertions(+) (limited to 'fs/fs_struct.c') diff --git a/fs/fs_struct.c b/fs/fs_struct.c index ed45a9cf5f3d..60b8531f41c5 100644 --- a/fs/fs_struct.c +++ b/fs/fs_struct.c @@ -14,9 +14,11 @@ void set_fs_root(struct fs_struct *fs, struct path *path) struct path old_root; spin_lock(&fs->lock); + write_seqcount_begin(&fs->seq); old_root = fs->root; fs->root = *path; path_get(path); + write_seqcount_end(&fs->seq); spin_unlock(&fs->lock); if (old_root.dentry) path_put(&old_root); @@ -31,9 +33,11 @@ void set_fs_pwd(struct fs_struct *fs, struct path *path) struct path old_pwd; spin_lock(&fs->lock); + write_seqcount_begin(&fs->seq); old_pwd = fs->pwd; fs->pwd = *path; path_get(path); + write_seqcount_end(&fs->seq); spin_unlock(&fs->lock); if (old_pwd.dentry) @@ -52,6 +56,7 @@ void chroot_fs_refs(struct path *old_root, struct path *new_root) fs = p->fs; if (fs) { spin_lock(&fs->lock); + write_seqcount_begin(&fs->seq); if (fs->root.dentry == old_root->dentry && fs->root.mnt == old_root->mnt) { path_get(new_root); @@ -64,6 +69,7 @@ void chroot_fs_refs(struct path *old_root, struct path *new_root) fs->pwd = *new_root; count++; } + write_seqcount_end(&fs->seq); spin_unlock(&fs->lock); } task_unlock(p); @@ -88,8 +94,10 @@ void exit_fs(struct task_struct *tsk) int kill; task_lock(tsk); spin_lock(&fs->lock); + write_seqcount_begin(&fs->seq); tsk->fs = NULL; kill = !--fs->users; + write_seqcount_end(&fs->seq); spin_unlock(&fs->lock); task_unlock(tsk); if (kill) @@ -105,6 +113,7 @@ struct fs_struct *copy_fs_struct(struct fs_struct *old) fs->users = 1; fs->in_exec = 0; spin_lock_init(&fs->lock); + seqcount_init(&fs->seq); fs->umask = old->umask; get_fs_root_and_pwd(old, &fs->root, &fs->pwd); } @@ -144,6 +153,7 @@ EXPORT_SYMBOL(current_umask); struct fs_struct init_fs = { .users = 1, .lock = __SPIN_LOCK_UNLOCKED(init_fs.lock), + .seq = SEQCNT_ZERO, .umask = 0022, }; -- cgit v1.2.3 From b3e19d924b6eaf2ca7d22cba99a517c5171007b6 Mon Sep 17 00:00:00 2001 From: Nick Piggin Date: Fri, 7 Jan 2011 17:50:11 +1100 Subject: fs: scale mntget/mntput The problem that this patch aims to fix is vfsmount refcounting scalability. We need to take a reference on the vfsmount for every successful path lookup, which often go to the same mount point. The fundamental difficulty is that a "simple" reference count can never be made scalable, because any time a reference is dropped, we must check whether that was the last reference. To do that requires communication with all other CPUs that may have taken a reference count. We can make refcounts more scalable in a couple of ways, involving keeping distributed counters, and checking for the global-zero condition less frequently. - check the global sum once every interval (this will delay zero detection for some interval, so it's probably a showstopper for vfsmounts). - keep a local count and only taking the global sum when local reaches 0 (this is difficult for vfsmounts, because we can't hold preempt off for the life of a reference, so a counter would need to be per-thread or tied strongly to a particular CPU which requires more locking). - keep a local difference of increments and decrements, which allows us to sum the total difference and hence find the refcount when summing all CPUs. Then, keep a single integer "long" refcount for slow and long lasting references, and only take the global sum of local counters when the long refcount is 0. This last scheme is what I implemented here. Attached mounts and process root and working directory references are "long" references, and everything else is a short reference. This allows scalable vfsmount references during path walking over mounted subtrees and unattached (lazy umounted) mounts with processes still running in them. This results in one fewer atomic op in the fastpath: mntget is now just a per-CPU inc, rather than an atomic inc; and mntput just requires a spinlock and non-atomic decrement in the common case. However code is otherwise bigger and heavier, so single threaded performance is basically a wash. Signed-off-by: Nick Piggin --- fs/fs_struct.c | 26 ++++++++++++++++---------- 1 file changed, 16 insertions(+), 10 deletions(-) (limited to 'fs/fs_struct.c') diff --git a/fs/fs_struct.c b/fs/fs_struct.c index 60b8531f41c5..68ca487bedb1 100644 --- a/fs/fs_struct.c +++ b/fs/fs_struct.c @@ -17,11 +17,11 @@ void set_fs_root(struct fs_struct *fs, struct path *path) write_seqcount_begin(&fs->seq); old_root = fs->root; fs->root = *path; - path_get(path); + path_get_long(path); write_seqcount_end(&fs->seq); spin_unlock(&fs->lock); if (old_root.dentry) - path_put(&old_root); + path_put_long(&old_root); } /* @@ -36,12 +36,12 @@ void set_fs_pwd(struct fs_struct *fs, struct path *path) write_seqcount_begin(&fs->seq); old_pwd = fs->pwd; fs->pwd = *path; - path_get(path); + path_get_long(path); write_seqcount_end(&fs->seq); spin_unlock(&fs->lock); if (old_pwd.dentry) - path_put(&old_pwd); + path_put_long(&old_pwd); } void chroot_fs_refs(struct path *old_root, struct path *new_root) @@ -59,13 +59,13 @@ void chroot_fs_refs(struct path *old_root, struct path *new_root) write_seqcount_begin(&fs->seq); if (fs->root.dentry == old_root->dentry && fs->root.mnt == old_root->mnt) { - path_get(new_root); + path_get_long(new_root); fs->root = *new_root; count++; } if (fs->pwd.dentry == old_root->dentry && fs->pwd.mnt == old_root->mnt) { - path_get(new_root); + path_get_long(new_root); fs->pwd = *new_root; count++; } @@ -76,13 +76,13 @@ void chroot_fs_refs(struct path *old_root, struct path *new_root) } while_each_thread(g, p); read_unlock(&tasklist_lock); while (count--) - path_put(old_root); + path_put_long(old_root); } void free_fs_struct(struct fs_struct *fs) { - path_put(&fs->root); - path_put(&fs->pwd); + path_put_long(&fs->root); + path_put_long(&fs->pwd); kmem_cache_free(fs_cachep, fs); } @@ -115,7 +115,13 @@ struct fs_struct *copy_fs_struct(struct fs_struct *old) spin_lock_init(&fs->lock); seqcount_init(&fs->seq); fs->umask = old->umask; - get_fs_root_and_pwd(old, &fs->root, &fs->pwd); + + spin_lock(&old->lock); + fs->root = old->root; + path_get_long(&fs->root); + fs->pwd = old->pwd; + path_get_long(&fs->pwd); + spin_unlock(&old->lock); } return fs; } -- cgit v1.2.3 From f03c65993b98eeb909a4012ce7833c5857d74755 Mon Sep 17 00:00:00 2001 From: Al Viro Date: Fri, 14 Jan 2011 22:30:21 -0500 Subject: sanitize vfsmount refcounting changes Instead of splitting refcount between (per-cpu) mnt_count and (SMP-only) mnt_longrefs, make all references contribute to mnt_count again and keep track of how many are longterm ones. Accounting rules for longterm count: * 1 for each fs_struct.root.mnt * 1 for each fs_struct.pwd.mnt * 1 for having non-NULL ->mnt_ns * decrement to 0 happens only under vfsmount lock exclusive That allows nice common case for mntput() - since we can't drop the final reference until after mnt_longterm has reached 0 due to the rules above, mntput() can grab vfsmount lock shared and check mnt_longterm. If it turns out to be non-zero (which is the common case), we know that this is not the final mntput() and can just blindly decrement percpu mnt_count. Otherwise we grab vfsmount lock exclusive and do usual decrement-and-check of percpu mnt_count. For fs_struct.c we have mnt_make_longterm() and mnt_make_shortterm(); namespace.c uses the latter in places where we don't already hold vfsmount lock exclusive and opencodes a few remaining spots where we need to manipulate mnt_longterm. Note that we mostly revert the code outside of fs/namespace.c back to what we used to have; in particular, normal code doesn't need to care about two kinds of references, etc. And we get to keep the optimization Nick's variant had bought us... Signed-off-by: Al Viro --- fs/fs_struct.c | 35 ++++++++++++++++++++++++----------- 1 file changed, 24 insertions(+), 11 deletions(-) (limited to 'fs/fs_struct.c') diff --git a/fs/fs_struct.c b/fs/fs_struct.c index 68ca487bedb1..78b519c13536 100644 --- a/fs/fs_struct.c +++ b/fs/fs_struct.c @@ -4,6 +4,19 @@ #include #include #include +#include "internal.h" + +static inline void path_get_longterm(struct path *path) +{ + path_get(path); + mnt_make_longterm(path->mnt); +} + +static inline void path_put_longterm(struct path *path) +{ + mnt_make_shortterm(path->mnt); + path_put(path); +} /* * Replace the fs->{rootmnt,root} with {mnt,dentry}. Put the old values. @@ -17,11 +30,11 @@ void set_fs_root(struct fs_struct *fs, struct path *path) write_seqcount_begin(&fs->seq); old_root = fs->root; fs->root = *path; - path_get_long(path); + path_get_longterm(path); write_seqcount_end(&fs->seq); spin_unlock(&fs->lock); if (old_root.dentry) - path_put_long(&old_root); + path_put_longterm(&old_root); } /* @@ -36,12 +49,12 @@ void set_fs_pwd(struct fs_struct *fs, struct path *path) write_seqcount_begin(&fs->seq); old_pwd = fs->pwd; fs->pwd = *path; - path_get_long(path); + path_get_longterm(path); write_seqcount_end(&fs->seq); spin_unlock(&fs->lock); if (old_pwd.dentry) - path_put_long(&old_pwd); + path_put_longterm(&old_pwd); } void chroot_fs_refs(struct path *old_root, struct path *new_root) @@ -59,13 +72,13 @@ void chroot_fs_refs(struct path *old_root, struct path *new_root) write_seqcount_begin(&fs->seq); if (fs->root.dentry == old_root->dentry && fs->root.mnt == old_root->mnt) { - path_get_long(new_root); + path_get_longterm(new_root); fs->root = *new_root; count++; } if (fs->pwd.dentry == old_root->dentry && fs->pwd.mnt == old_root->mnt) { - path_get_long(new_root); + path_get_longterm(new_root); fs->pwd = *new_root; count++; } @@ -76,13 +89,13 @@ void chroot_fs_refs(struct path *old_root, struct path *new_root) } while_each_thread(g, p); read_unlock(&tasklist_lock); while (count--) - path_put_long(old_root); + path_put_longterm(old_root); } void free_fs_struct(struct fs_struct *fs) { - path_put_long(&fs->root); - path_put_long(&fs->pwd); + path_put_longterm(&fs->root); + path_put_longterm(&fs->pwd); kmem_cache_free(fs_cachep, fs); } @@ -118,9 +131,9 @@ struct fs_struct *copy_fs_struct(struct fs_struct *old) spin_lock(&old->lock); fs->root = old->root; - path_get_long(&fs->root); + path_get_longterm(&fs->root); fs->pwd = old->pwd; - path_get_long(&fs->pwd); + path_get_longterm(&fs->pwd); spin_unlock(&old->lock); } return fs; -- cgit v1.2.3