[PATCH] mm: fix wrong vmap address calculations with odd NR_CPUSvalues

From: Stefan Richter
Date: Sun Aug 14 2011 - 09:13:03 EST

From: Clemens Ladisch <clemens@xxxxxxxxxx>
Date: Tue, 21 Jun 2011 22:09:50 +0200

Commit db64fe02258f (mm: rewrite vmap layer) introduced code that does
address calculations under the assumption that VMAP_BLOCK_SIZE is
a power of two. However, this might not be true if CONFIG_NR_CPUS is
not set to a power of two.

Wrong vmap_block index/offset values could lead to memory corruption.
However, this has never been observed in practice (or never been
diagnosed correctly); what caught this was the BUG_ON in vb_alloc() that
checks for inconsistent vmap_block indices.

To fix this, ensure that VMAP_BLOCK_SIZE always is a power of two.

BugLink: https://bugzilla.kernel.org/show_bug.cgi?id=31572
Reported-by: Pavel Kysilka <goldenfish@xxxxxxxxxxxx>
Reported-by: Matias A. Fonzo <selk@xxxxxxxxxxx>
Signed-off-by: Clemens Ladisch <clemens@xxxxxxxxxx>
Signed-off-by: Stefan Richter <stefanr@xxxxxxxxxxxxxxxxx>
Cc: 2.6.28+ <stable@xxxxxxxxxx>
Resend with corrected Cc list; sorry.

This fixes instant and fully repeatable crashes if NR_CPUS is not a power
of two and vm_map_ram() or something like that is executed, for example
firewire-ohci probe in its 2.6.38+ incarnation.

mm/vmalloc.c | 7 ++++---
1 files changed, 4 insertions(+), 3 deletions(-)

diff --git a/mm/vmalloc.c b/mm/vmalloc.c
index 1d34d75..d3d451b 100644
--- a/mm/vmalloc.c
+++ b/mm/vmalloc.c
@@ -732,9 +732,10 @@ static void free_unmap_vmap_area_addr(unsigned long addr)
#define VMAP_MIN(x, y) ((x) < (y) ? (x) : (y)) /* can't use min() */
#define VMAP_MAX(x, y) ((x) > (y) ? (x) : (y)) /* can't use max() */
+#define VMAP_BBMAP_BITS \
+ VMALLOC_PAGES / roundup_pow_of_two(NR_CPUS) / 16))


To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/