
Am 09.07.2012 19:35, schrieb Corey Bryant:
On 07/09/2012 11:46 AM, Kevin Wolf wrote:
Am 09.07.2012 17:05, schrieb Corey Bryant:
I'm not sure this is an issue with current design. I know things have changed a bit as the email threads evolved, so I'll paste the current design that I am working from. Please let me know if you still see any issues.
FD passing: ----------- New monitor commands enable adding/removing an fd to/from a set. New monitor command query-fdsets enables querying of current monitor fdsets. The set of fds should all refer to the same file, with each fd having different access flags (ie. O_RDWR, O_RDONLY). qemu_open can then dup the fd that has the matching access mode flags.
Design points: -------------- 1. add-fd -> fd is passed via SCM rights and qemu adds fd to first unused fdset (e.g. /dev/fdset/1) -> add-fd monitor function initializes the monitor inuse flag for the fdset to true -> add-fd monitor function initializes the remove flag for the fd to false -> add-fd returns fdset number and received fd number (e.g fd=3) to caller
2. drive_add file=/dev/fdset/1 -> qemu_open uses the first fd in fdset1 that has access flags matching the qemu_open action flags and has remove flag set to false -> qemu_open increments refcount for the fdset -> Need to make sure that if a command like 'device-add' fails that refcount is not incremented
3. add-fd fdset=1 -> fd is passed via SCM rights -> add-fd monitor function adds the received fd to the specified fdset (or fails if fdset doesn't exist) -> add-fd monitor function initializes the remove flag for the fd to false -> add-fd returns fdset number and received fd number (e.g fd=4) to caller
4. block-commit -> qemu_open performs "reopen" by using the first fd from the fdset that has access flags matching the qemu_open action flags and has remove flag set to false -> qemu_open increments refcount for the fdset -> Need to make sure that if a command like 'block-commit' fails that refcount is not incremented
5. remove-fd fdset=1 fd=4 -> remove-fd monitor function fails if fdset doesn't exist -> remove-fd monitor function turns on remove flag for fd=4
What was again the reason why we keep removed fds in the fdset at all?
Because if refcount is > 0 for the fd set, then the fd could be in use by a block device. So we keep it around until refcount is decremented to zero, at which point it is safe to close.
The removed flag would make sense for a fdset after a hypothetical close-fdset call because the fdset needs to be kept around until the last user closes it, but I think removed fds can be deleted immediately.
fds in an fd set really need to be kept around until zero block devices reference them. At that point, if '(refcount == 0 && (!inuse || remove))' is true, then we'll officially close the fd.
Block devices don't reference an fd in the fdset. There are two references in a block device. The first one is obviously the file descriptor they are using; it is a fd dup()ed from an fd in the fdset, but it's now independent of it. The other reference is the file name that is kept in the BlockDriverState, and it always points to "/dev/fdset/X", that is, the whole fdset instead of a single fd. What happens if you remove a file descriptor from an fdset that is in use, is that you can't reopen the fdset with the flags of the removed file descriptor any more. Which I believe is exactly the expected behaviour. libvirt would use this to revoke r/w access, for example (and which behaviour you already provide by checking removed in qemu_open). Are there any other use cases where it makes a difference whether a file descriptor is kept in the fdset with removed=1 or whether it's actually removed from the fdset?
I think I might have confused remove-fd and close-fdset in earlier emails in this thread, so I hope this isn't inconsistent with what I said before.
Ok no problem.
6. qemu_close (need to replace all close calls in block layer with qemu_close) -> qemu_close decrements refcount for fdset -> qemu_close closes all fds that have (refcount == 0 && (!inuse || remove)) -> qemu_close frees the fdset if no fds remain in it
7. disconnecting the QMP monitor -> monitor disconnect visits all fdsets on monitor and turns off monitor in-use flag for fdset
And close all fds with refcount == 0.
Yes, this makes sense.
It also makes sense to close removed fds with refcount == 0 in the remove-fd function. Basically this will be the same thing we do in qemu_close. We'll close any fds that evaulate the following as true:
(refcount == 0 && (!inuse || remove))
Yes, whatever condition we'll come up with, but it should be the same and checked in all places where its value might change. Kevin