My machines silently die with this patch :(
The seems to be a rock-solid deadlock somewhere.
Is that inode leak fix in the 2.2.2-pre2 the same as the one I've got from
Andrea?
Andrea Arcangeli wrote:
> On Thu, 4 Feb 1999, Tomasz Przygoda wrote:
>
> > Hi,
> >
> > Andrea Arcangeli wrote:
> >
> > > On Wed, 3 Feb 1999, Tomasz Przygoda wrote:
> > > > /proc/sys/fs/inode-nr : 1268624 3
> > > > /proc/sys/fs/inode-state : 1268624 9 1 0 0 0 0
> > >
> > > Please send me your .config and show me the output of `mount`. What the
> > > system is used for? How can you reproduce? How can I reproduce here?
> >
> > Well, I'm able to get same sort of numbers with any of the 2.2.x; where x =
> > {0-pre7, 0, 1}.
>
> Just fixed, apply this patch against 2.2.1 and be happy:
>
> Index: inode.c
> ===================================================================
> RCS file: /var/cvs/linux/fs/inode.c,v
> retrieving revision 1.1.1.2
> diff -u -r1.1.1.2 inode.c
> --- inode.c 1999/01/26 18:30:24 1.1.1.2
> +++ inode.c 1999/02/04 15:35:46
> @@ -4,6 +4,10 @@
> * (C) 1997 Linus Torvalds
> */
>
> +/*
> + * Fixed inode leakage in 2.2.1. Copyright (C) 1999 Andrea Arcangeli
> + */
> +
> #include <linux/fs.h>
> #include <linux/string.h>
> #include <linux/mm.h>
> @@ -62,9 +66,8 @@
> struct {
> int nr_inodes;
> int nr_free_inodes;
> - int preshrink; /* pre-shrink dcache? */
> - int dummy[4];
> -} inodes_stat = {0, 0, 0,};
> + int dummy[5];
> +} inodes_stat = {0, 0, };
>
> int max_inodes;
>
> @@ -329,12 +332,14 @@
>
> static int free_inodes(int goal)
> {
> - struct list_head *tmp, *head = &inode_in_use;
> + struct list_head * head = &inode_in_use, * tmp, * this = head->prev;
> LIST_HEAD(freeable);
> - int found = 0, depth = goal << 1;
> + int found = 0;
>
> - while ((tmp = head->prev) != head && depth--) {
> + while ((tmp = this) != head)
> + {
> struct inode * inode = list_entry(tmp, struct inode, i_list);
> + this = this->prev;
> list_del(tmp);
> if (CAN_UNUSE(inode)) {
> list_del(&inode->i_hash);
> @@ -403,7 +408,7 @@
> /*
> * Check whether to restock the unused list.
> */
> - if (inodes_stat.preshrink) {
> + if (inodes_stat.nr_inodes > max_inodes) {
> struct list_head *tmp;
> try_to_free_inodes(8);
> tmp = inode_unused.next;
> @@ -436,9 +441,6 @@
> */
> inodes_stat.nr_inodes += INODES_PER_PAGE;
> inodes_stat.nr_free_inodes += INODES_PER_PAGE - 1;
> - inodes_stat.preshrink = 0;
> - if (inodes_stat.nr_inodes > max_inodes)
> - inodes_stat.preshrink = 1;
> return inode;
> }
>
> @@ -447,7 +449,6 @@
> * the dcache and then try again to free some inodes.
> */
> prune_dcache(inodes_stat.nr_inodes >> 2);
> - inodes_stat.preshrink = 1;
>
> spin_lock(&inode_lock);
> free_inodes(inodes_stat.nr_inodes >> 2);
>
> Andrea Arcangeli
-- Tomek,
"Office'97, Windows'98 - every Microsoft product has an expiration date."
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/