vfs: fully lockless v_writecount adjustment

[mjg]

• upload correct patch

[kib]

sys/kern/vfs_default.c

1322

Isn't this unsafe? You should ref the vnode before adding 1 to execcount if you ever saw n == 0, and rollback before returning if actually used cmpset was for n < 0.

[mjg]

sys/kern/vfs_default.c

1322

this can only be a problem if someone bogusly adjusts writecount back to 0, but i can make the change if you insist