[PATCH][V2][Fix commit message] x86/RAS/mce_amd_inj: fix signed wrap around when decrementing index i

[PATCH][V2][Fix commit message] x86/RAS/mce_amd_inj: fix signed wrap around when decrementing index i

[Colin King]

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

Integer index i needs to be a signed int rather than unsigned to avoid
a wrap-around when decrementing in the while loop.  For example, if the
debugfs_create_file fails when i is zero, the current situation will
predecrement i in the while loop, wrapping i to the maximum signed
integer and cause multiple out of bounds reads on dfs_fls[i].d as
the loop interates to zero.

Also add (int) cast to fix warning that the original fix attempted
to fix.

Fixes: 7cc4ef8ed132 ("x86/RAS/mce_amd_inj: Fix some W= warnings")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
---
 arch/x86/ras/mce_amd_inj.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/arch/x86/ras/mce_amd_inj.c b/arch/x86/ras/mce_amd_inj.c
index cd318d9..cb9779a 100644
--- a/arch/x86/ras/mce_amd_inj.c
+++ b/arch/x86/ras/mce_amd_inj.c
@@ -440,7 +440,7 @@ static struct dfs_node {
 
 static int __init init_mce_inject(void)
 {
-	unsigned int i;
+	int i;
 	u64 cap;
 
 	rdmsrl(MSR_IA32_MCG_CAP, cap);
@@ -450,7 +450,7 @@ static int __init init_mce_inject(void)
 	if (!dfs_inj)
 		return -EINVAL;
 
-	for (i = 0; i < ARRAY_SIZE(dfs_fls); i++) {
+	for (i = 0; i < (int)ARRAY_SIZE(dfs_fls); i++) {
 		dfs_fls[i].d = debugfs_create_file(dfs_fls[i].name,
 						    dfs_fls[i].perm,
 						    dfs_inj,
-- 
2.9.3

Re: [PATCH][V2][Fix commit message] x86/RAS/mce_amd_inj: fix signed wrap around when decrementing index i

[Borislav Petkov]

(drop peterz from CC).

On Fri, Sep 16, 2016 at 12:22:23PM +0100, Colin King wrote:
> From: Colin Ian King <colin.king@canonical.com>
> 
> Integer index i needs to be a signed int rather than unsigned to avoid
> a wrap-around when decrementing in the while loop.  For example, if the

Damn corner cases :-\

> debugfs_create_file fails when i is zero, the current situation will
> predecrement i in the while loop, wrapping i to the maximum signed
> integer and cause multiple out of bounds reads on dfs_fls[i].d as
> the loop interates to zero.
> 
> Also add (int) cast to fix warning that the original fix attempted
> to fix.

Thanks, that's a good catch!

But, can we keep the unsigned-ness of i as it is an index and do this
instead?

(Oh, and make it return ENODEV - that ENOMEM is just plain wrong most of
the time).

---
diff --git a/arch/x86/ras/mce_amd_inj.c b/arch/x86/ras/mce_amd_inj.c
index cd318d93099e..a2dafca467e6 100644
--- a/arch/x86/ras/mce_amd_inj.c
+++ b/arch/x86/ras/mce_amd_inj.c
@@ -457,8 +457,12 @@ static int __init init_mce_inject(void)
 						    &i_mce,
 						    dfs_fls[i].fops);
 
-		if (!dfs_fls[i].d)
+		if (!dfs_fls[i].d) {
+			if (!i)
+				return -ENODEV;
+
 			goto err_dfs_add;
+		}
 	}
 
 	return 0;
@@ -470,7 +474,7 @@ err_dfs_add:
 	debugfs_remove(dfs_inj);
 	dfs_inj = NULL;
 
-	return -ENOMEM;
+	return -ENODEV;
 }
 
 static void __exit exit_mce_inject(void)

-- 
Regards/Gruss,
    Boris.

SUSE Linux GmbH, GF: Felix Imendörffer, Jane Smithard, Graham Norton, HRB 21284 (AG Nürnberg)
--