[PATCH] cpusets: mems generation deadlock fix

[Linux Kernel Mailing List]

ChangeSet 1.2231.1.152, 2005/03/28 19:58:44-08:00, [EMAIL PROTECTED]

        [PATCH] cpusets: mems generation deadlock fix
        
        The cpuset code to update mems_generation could (in theory) deadlock on
        cpuset_sem if it needed to allocate some memory while creating (mkdir) 
or
        removing (rmdir) a cpuset, so already held cpuset_sem.  Some other 
process
        would have to mess with this tasks cpuset memory placement at the same
        time.
        
        We avoid this possible deadlock by always updating mems_generation 
after we
        grab cpuset_sem on such operations, before we risk any operations that
        might require memory allocation.
        
        Thanks to Jack Steiner <[EMAIL PROTECTED]> for noticing this.
        
        Signed-off-by: Paul Jackson <[EMAIL PROTECTED]>
        Signed-off-by: Andrew Morton <[EMAIL PROTECTED]>
        Signed-off-by: Linus Torvalds <[EMAIL PROTECTED]>



 cpuset.c |   34 ++++++++++++++++++++++++++++++++--
 1 files changed, 32 insertions(+), 2 deletions(-)


diff -Nru a/kernel/cpuset.c b/kernel/cpuset.c
--- a/kernel/cpuset.c   2005-03-28 21:39:53 -08:00
+++ b/kernel/cpuset.c   2005-03-28 21:39:53 -08:00
@@ -505,6 +505,35 @@
 }
 
 /*
+ * Refresh current tasks mems_allowed and mems_generation from
+ * current tasks cpuset.  Call with cpuset_sem held.
+ *
+ * Be sure to call refresh_mems() on any cpuset operation which
+ * (1) holds cpuset_sem, and (2) might possibly alloc memory.
+ * Call after obtaining cpuset_sem lock, before any possible
+ * allocation.  Otherwise one risks trying to allocate memory
+ * while the task cpuset_mems_generation is not the same as
+ * the mems_generation in its cpuset, which would deadlock on
+ * cpuset_sem in cpuset_update_current_mems_allowed().
+ *
+ * Since we hold cpuset_sem, once refresh_mems() is called, the
+ * test (current->cpuset_mems_generation != cs->mems_generation)
+ * in cpuset_update_current_mems_allowed() will remain false,
+ * until we drop cpuset_sem.  Anyone else who would change our
+ * cpusets mems_generation needs to lock cpuset_sem first.
+ */
+
+static void refresh_mems(void)
+{
+       struct cpuset *cs = current->cpuset;
+
+       if (current->cpuset_mems_generation != cs->mems_generation) {
+               guarantee_online_mems(cs, &current->mems_allowed);
+               current->cpuset_mems_generation = cs->mems_generation;
+       }
+}
+
+/*
  * is_cpuset_subset(p, q) - Is cpuset p a subset of cpuset q?
  *
  * One cpuset is a subset of another if all its allowed CPUs and
@@ -1224,6 +1253,7 @@
                return -ENOMEM;
 
        down(&cpuset_sem);
+       refresh_mems();
        cs->flags = 0;
        if (notify_on_release(parent))
                set_bit(CS_NOTIFY_ON_RELEASE, &cs->flags);
@@ -1277,6 +1307,7 @@
        /* the vfs holds both inode->i_sem already */
 
        down(&cpuset_sem);
+       refresh_mems();
        if (atomic_read(&cs->count) > 0) {
                up(&cpuset_sem);
                return -EBUSY;
@@ -1433,8 +1464,7 @@
                return;         /* task is exiting */
        if (current->cpuset_mems_generation != cs->mems_generation) {
                down(&cpuset_sem);
-               guarantee_online_mems(cs, &current->mems_allowed);
-               current->cpuset_mems_generation = cs->mems_generation;
+               refresh_mems();
                up(&cpuset_sem);
        }
 }
-
To unsubscribe from this list: send the line "unsubscribe bk-commits-head" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at  http://vger.kernel.org/majordomo-info.html