[PATCH] x86/microcode/AMD: fix read of uninitialized data

[PATCH] x86/microcode/AMD: fix read of uninitialized data

[Colin King]

From: Colin Ian King <colin.king@canonical.com>

Currently function __load_ucode_amd can assign an uninitialized cp to *ret
if get_builtin_microcode returns false. This could cause function
load_ucode_amd_bsp to attempt to apply the microcode if the garbage data in
cp is non-zero.  Fix this by moving the zero'ing of cp from function
load_ucode_amd_bsp to __load_ucode_amd.

Detected by CoverityScan, CID#1398332 ("Uninitialized scalar variable")

Fixes: 69f5f983001f ("x86/microcode/AMD: Remove AP scanning optimization")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
---
 arch/x86/kernel/cpu/microcode/amd.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/arch/x86/kernel/cpu/microcode/amd.c b/arch/x86/kernel/cpu/microcode/amd.c
index 0624957aa068..745135699233 100644
--- a/arch/x86/kernel/cpu/microcode/amd.c
+++ b/arch/x86/kernel/cpu/microcode/amd.c
@@ -254,7 +254,7 @@ static bool get_builtin_microcode(struct cpio_data *cp, unsigned int family)
 static void __load_ucode_amd(unsigned int cpuid_1_eax, struct cpio_data *ret)
 {
 	struct ucode_cpu_info *uci;
-	struct cpio_data cp;
+	struct cpio_data cp = { };
 	const char *path;
 	bool use_pa;
 
@@ -279,7 +279,7 @@ static void __load_ucode_amd(unsigned int cpuid_1_eax, struct cpio_data *ret)
 
 void __init load_ucode_amd_bsp(unsigned int cpuid_1_eax)
 {
-	struct cpio_data cp = { };
+	struct cpio_data cp;
 
 	__load_ucode_amd(cpuid_1_eax, &cp);
 	if (!(cp.data && cp.size))
-- 
2.15.1

Re: [PATCH] x86/microcode/AMD: fix read of uninitialized data

[Dan Carpenter]

On Tue, Apr 10, 2018 at 02:41:23PM +0100, Colin King wrote:
> From: Colin Ian King <colin.king@canonical.com>
> 
> Currently function __load_ucode_amd can assign an uninitialized cp to *ret
> if get_builtin_microcode returns false.

What are you looking at?  Here is how it looks in linux-next:

   270  
   271          if (!get_builtin_microcode(&cp, x86_family(cpuid_1_eax)))
   272                  cp = find_microcode_in_initrd(path, use_pa);
   273  


If get_builtin_microcode() returns false, then cp is initialized on the
next line.

Maybe the issue is that cp.name is uninitialized.  It doesn't cause an
issue but it would be understandable for a static checker to warn about
it when we do "*ret = cp;".

regards,
dan carpenter

Re: [PATCH] x86/microcode/AMD: fix read of uninitialized data

[Colin Ian King]

On 10/04/18 15:22, Dan Carpenter wrote:
> On Tue, Apr 10, 2018 at 02:41:23PM +0100, Colin King wrote:
>> From: Colin Ian King <colin.king@canonical.com>
>>
>> Currently function __load_ucode_amd can assign an uninitialized cp to *ret
>> if get_builtin_microcode returns false.
> 
> What are you looking at?  Here is how it looks in linux-next:
> 
>    270  
>    271          if (!get_builtin_microcode(&cp, x86_family(cpuid_1_eax)))
>    272                  cp = find_microcode_in_initrd(path, use_pa);
>    273  
> 
> 
> If get_builtin_microcode() returns false, then cp is initialized on the
> next line.
> 
> Maybe the issue is that cp.name is uninitialized.  It doesn't cause an
> issue but it would be understandable for a static checker to warn about
> it when we do "*ret = cp;".

Oh, oops, you are totally correct on that. My bad.


> 
> regards,
> dan carpenter
>