From 91762cd331d955f1378dd90af08b56aa15861ab4 Mon Sep 17 00:00:00 2001 From: Miklos Szeredi Date: Thu, 29 Jun 2006 14:38:35 +0000 Subject: *** empty log message *** --- doc/kernel.txt | 118 ++++++++++++++++++++++++++++++++++----------------------- 1 file changed, 71 insertions(+), 47 deletions(-) (limited to 'doc') diff --git a/doc/kernel.txt b/doc/kernel.txt index 33f7431..a584f05 100644 --- a/doc/kernel.txt +++ b/doc/kernel.txt @@ -18,6 +18,14 @@ Non-privileged mount (or user mount): user. NOTE: this is not the same as mounts allowed with the "user" option in /etc/fstab, which is not discussed here. +Filesystem connection: + + A connection between the filesystem daemon and the kernel. The + connection exists until either the daemon dies, or the filesystem is + umounted. Note that detaching (or lazy umounting) the filesystem + does _not_ break the connection, in this case it will exist until + the last reference to the filesystem is released. + Mount owner: The user who does the mounting. @@ -86,16 +94,20 @@ Mount options The default is infinite. Note that the size of read requests is limited anyway to 32 pages (which is 128kbyte on i386). -Sysfs -~~~~~ +Control filesystem +~~~~~~~~~~~~~~~~~~ + +There's a control filesystem for FUSE, which can be mounted by: -FUSE sets up the following hierarchy in sysfs: + mount -t fusectl none /sys/fs/fuse/connections - /sys/fs/fuse/connections/N/ +Mounting it under the '/sys/fs/fuse/connections' directory makes it +backwards compatible with earlier versions. -where N is an increasing number allocated to each new connection. +Under the fuse control filesystem each connection has a directory +named by a unique number. -For each connection the following attributes are defined: +For each connection the following files exist within this directory: 'waiting' @@ -110,7 +122,47 @@ For each connection the following attributes are defined: connection. This means that all waiting requests will be aborted an error returned for all aborted and new requests. -Only a privileged user may read or write these attributes. +Only the owner of the mount may read or write these files. + +Interrupting filesystem operations +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +If a process issuing a FUSE filesystem request is interrupted, the +following will happen: + + 1) If the request is not yet sent to userspace AND the signal is + fatal (SIGKILL or unhandled fatal signal), then the request is + dequeued and returns immediately. + + 2) If the request is not yet sent to userspace AND the signal is not + fatal, then an 'interrupted' flag is set for the request. When + the request has been successfully transfered to userspace and + this flag is set, an INTERRUPT request is queued. + + 3) If the request is already sent to userspace, then an INTERRUPT + request is queued. + +INTERRUPT requests take precedence over other requests, so the +userspace filesystem will receive queued INTERRUPTs before any others. + +The userspace filesystem may ignore the INTERRUPT requests entirely, +or may honor them by sending a reply to the _original_ request, with +the error set to EINTR. + +It is also possible that there's a race between processing the +original request and it's INTERRUPT request. There are two possibilities: + + 1) The INTERRUPT request is processed before the original request is + processed + + 2) The INTERRUPT request is processed after the original request has + been answered + +If the filesystem cannot find the original request, it should wait for +some timeout and/or a number of new requests to arrive, after which it +should reply to the INTERRUPT request with an EAGAIN error. In case +1) the INTERRUPT request will be requeued. In case 2) the INTERRUPT +reply will be ignored. Aborting a filesystem connection ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ @@ -139,8 +191,8 @@ the filesystem. There are several ways to do this: - Use forced umount (umount -f). Works in all cases but only if filesystem is still attached (it hasn't been lazy unmounted) - - Abort filesystem through the sysfs interface. Most powerful - method, always works. + - Abort filesystem through the FUSE control filesystem. Most + powerful method, always works. How do non-privileged mounts work? ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ @@ -304,25 +356,7 @@ Scenario 1 - Simple deadlock | | for "file"] | | *DEADLOCK* -The solution for this is to allow requests to be interrupted while -they are in userspace: - - | [interrupted by signal] | - | fuse_unlink() - | | [queue req on fc->pending] - | | [wake up fc->waitq] - | | [sleep on req->waitq] - -If the filesystem daemon was single threaded, this will stop here, -since there's no other thread to dequeue and execute the request. -In this case the solution is to kill the FUSE daemon as well. If -there are multiple serving threads, you just have to kill them as -long as any remain. - -Moral: a filesystem which deadlocks, can soon find itself dead. +The solution for this is to allow the filesystem to be aborted. Scenario 2 - Tricky deadlock ---------------------------- @@ -355,24 +389,14 @@ but is caused by a pagefault. | | [lock page] | | * DEADLOCK * -Solution is again to let the the request be interrupted (not -elaborated further). - -An additional problem is that while the write buffer is being -copied to the request, the request must not be interrupted. This -is because the destination address of the copy may not be valid -after the request is interrupted. - -This is solved with doing the copy atomically, and allowing -interruption while the page(s) belonging to the write buffer are -faulted with get_user_pages(). The 'req->locked' flag indicates -when the copy is taking place, and interruption is delayed until -this flag is unset. +Solution is basically the same as above. -Scenario 3 - Tricky deadlock with asynchronous read ---------------------------------------------------- +An additional problem is that while the write buffer is being copied +to the request, the request must not be interrupted/aborted. This is +because the destination address of the copy may not be valid after the +request has returned. -The same situation as above, except thread-1 will wait on page lock -and hence it will be uninterruptible as well. The solution is to -abort the connection with forced umount (if mount is attached) or -through the abort attribute in sysfs. +This is solved with doing the copy atomically, and allowing abort +while the page(s) belonging to the write buffer are faulted with +get_user_pages(). The 'req->locked' flag indicates when the copy is +taking place, and abort is delayed until this flag is unset. -- cgit v1.2.3