[IA64-SGI] Silent data corruption caused by XPC V2.
Jack Steiner identified a problem where XPC can cause a silent
data corruption. On module load, the placement may cause the
xpc_remote_copy_buffer to span two physical pages. DMA transfers are
done to the start virtual address translated to physical.
This patch changes the buffer from a statically allocated buffer to a
kmalloc'd buffer. Dean Nelson reviewed this before posting. I have
tested it in the configuration that was showing the memory corruption
and verified it works. I also added a BUG_ON statement to help catch
this if a similar situation is encountered.
Signed-off-by: Robin Holt <holt@sgi.com>
Signed-off-by: Dean Nelson <dcn@sgi.com>
Signed-off-by: Jack Steiner <steiner@sgi.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
diff --git a/arch/ia64/sn/kernel/xpc_partition.c b/arch/ia64/sn/kernel/xpc_partition.c
index 2a89cfc..57c723f 100644
--- a/arch/ia64/sn/kernel/xpc_partition.c
+++ b/arch/ia64/sn/kernel/xpc_partition.c
@@ -71,19 +71,15 @@
* Generic buffer used to store a local copy of portions of a remote
* partition's reserved page (either its header and part_nasids mask,
* or its vars).
- *
- * xpc_discovery runs only once and is a seperate thread that is
- * very likely going to be processing in parallel with receiving
- * interrupts.
*/
-char ____cacheline_aligned xpc_remote_copy_buffer[XPC_RP_HEADER_SIZE +
- XP_NASID_MASK_BYTES];
+char *xpc_remote_copy_buffer;
+void *xpc_remote_copy_buffer_base;
/*
* Guarantee that the kmalloc'd memory is cacheline aligned.
*/
-static void *
+void *
xpc_kmalloc_cacheline_aligned(size_t size, gfp_t flags, void **base)
{
/* see if kmalloc will give us cachline aligned memory by default */
@@ -148,7 +144,7 @@
}
}
- bte_res = xp_bte_copy(rp_pa, ia64_tpa(buf), buf_len,
+ bte_res = xp_bte_copy(rp_pa, buf, buf_len,
(BTE_NOTIFY | BTE_WACQUIRE), NULL);
if (bte_res != BTE_SUCCESS) {
dev_dbg(xpc_part, "xp_bte_copy failed %i\n", bte_res);
@@ -447,7 +443,7 @@
/* pull the remote_hb cache line */
bres = xp_bte_copy(part->remote_vars_pa,
- ia64_tpa((u64) remote_vars),
+ (u64) remote_vars,
XPC_RP_VARS_SIZE,
(BTE_NOTIFY | BTE_WACQUIRE), NULL);
if (bres != BTE_SUCCESS) {
@@ -498,8 +494,7 @@
/* pull over the reserved page header and part_nasids mask */
-
- bres = xp_bte_copy(*remote_rp_pa, ia64_tpa((u64) remote_rp),
+ bres = xp_bte_copy(*remote_rp_pa, (u64) remote_rp,
XPC_RP_HEADER_SIZE + xp_nasid_mask_bytes,
(BTE_NOTIFY | BTE_WACQUIRE), NULL);
if (bres != BTE_SUCCESS) {
@@ -554,11 +549,8 @@
return xpcVarsNotSet;
}
-
/* pull over the cross partition variables */
-
- bres = xp_bte_copy(remote_vars_pa, ia64_tpa((u64) remote_vars),
- XPC_RP_VARS_SIZE,
+ bres = xp_bte_copy(remote_vars_pa, (u64) remote_vars, XPC_RP_VARS_SIZE,
(BTE_NOTIFY | BTE_WACQUIRE), NULL);
if (bres != BTE_SUCCESS) {
return xpc_map_bte_errors(bres);
@@ -1239,7 +1231,7 @@
part_nasid_pa = (u64) XPC_RP_PART_NASIDS(part->remote_rp_pa);
- bte_res = xp_bte_copy(part_nasid_pa, ia64_tpa((u64) nasid_mask),
+ bte_res = xp_bte_copy(part_nasid_pa, (u64) nasid_mask,
xp_nasid_mask_bytes, (BTE_NOTIFY | BTE_WACQUIRE), NULL);
return xpc_map_bte_errors(bte_res);