[PATCH] x86/microcode/AMD: fix memleak in update_cache()

From: shuwang
Date: Wed Jul 19 2017 - 00:21:48 EST


From: Shu Wang <shuwang@xxxxxxxxxx>

Found this issue by kmemleak. The mem is allocated in
verify_and_add_patch(), passed to update_cache(patch),
and just dropped the reference without free
if (p->patch_id >= new_patch->patch_id)
return;

unreferenced object 0xffff88010e780b40 (size 32):
comm "bash", pid 860, jiffies 4294690939 (age 29.297s)
backtrace:
[<ffffffff8176166a>] kmemleak_alloc+0x4a/0xa0
[<ffffffff8121801a>] kmem_cache_alloc_trace+0xca/0x1d0
[<ffffffff81050d60>] load_microcode_amd.isra.0+0x1d0/0x400
[<ffffffff81051053>] request_microcode_amd+0xc3/0x160
[<ffffffff8104eef1>] reload_store+0xe1/0x170
[<ffffffff814dd228>] dev_attr_store+0x18/0x30
[<ffffffff812ce8ca>] sysfs_kf_write+0x3a/0x50
[<ffffffff812ce38f>] kernfs_fop_write+0xff/0x180
[<ffffffff81242307>] __vfs_write+0x37/0x170
[<ffffffff81243892>] vfs_write+0xb2/0x1b0
[<ffffffff81244ec5>] SyS_write+0x55/0xc0
[<ffffffff81003857>] do_syscall_64+0x67/0x150
[<ffffffff8176d3e7>] return_from_SYSCALL_64+0x0/0x6a
[<ffffffffffffffff>] 0xffffffffffffffff

(gdb) list *0xffffffff81050d60
0xffffffff81050d60 is in load_microcode_amd
(arch/x86/kernel/cpu/microcode/amd.c:616).

Signed-off-by: Shu Wang <shuwang@xxxxxxxxxx>
---
arch/x86/kernel/cpu/microcode/amd.c | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/arch/x86/kernel/cpu/microcode/amd.c b/arch/x86/kernel/cpu/microcode/amd.c
index 21b1857..c6daec4 100644
--- a/arch/x86/kernel/cpu/microcode/amd.c
+++ b/arch/x86/kernel/cpu/microcode/amd.c
@@ -400,9 +400,12 @@ static void update_cache(struct ucode_patch *new_patch)

list_for_each_entry(p, &microcode_cache, plist) {
if (p->equiv_cpu == new_patch->equiv_cpu) {
- if (p->patch_id >= new_patch->patch_id)
+ if (p->patch_id >= new_patch->patch_id) {
/* we already have the latest patch */
+ kfree(new_patch->data);
+ kfree(new_patch);
return;
+ }

list_replace(&p->plist, &new_patch->plist);
kfree(p->data);
--
2.5.0