[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <582ce2a4-072d-5744-5465-fbd07b68b53e@linux.intel.com>
Date: Tue, 17 Jul 2018 11:03:39 -0700
From: Dave Hansen <dave.hansen@...ux.intel.com>
To: Jiang Biao <jiang.biao2@....com.cn>, tglx@...utronix.de,
mingo@...hat.com
Cc: luto@...nel.org, hpa@...or.com, x86@...nel.org, albcamus@...il.com,
linux-kernel@...r.kernel.org, zhong.weidong@....com.cn
Subject: Re: [PATCH v2 2/2] x86/pti: check the return value of
pti_user_pagetable_walk_pmd
On 07/17/2018 12:23 AM, Jiang Biao wrote:
> Check the return value of pti_user_pagetable_walk_pmd() to avoid
> NULL pointer dereference. And add warning for fail allocation.
For both of these:
Acked-by: Dave Hansen <dave.hansen@...el.com>
It's minor, but if you redo these, I'd appreciate a slightly different
form. Instead of:
> @@ -239,8 +239,10 @@ static pmd_t *pti_user_pagetable_walk_pmd(unsigned long address)
> static __init pte_t *pti_user_pagetable_walk_pte(unsigned long address)
> {
> gfp_t gfp = (GFP_KERNEL | __GFP_NOTRACK | __GFP_ZERO);
> - pmd_t *pmd = pti_user_pagetable_walk_pmd(address);
> pte_t *pte;
> + pmd_t *pmd = pti_user_pagetable_walk_pmd(address);
> + if (!pmd)
> + return NULL;
I'd much rather see separation of code -- especially _important_ code
like an allocation -- from local variable definitions. Like this:
gfp_t gfp = (GFP_KERNEL | __GFP_NOTRACK | __GFP_ZERO);
pmd_t *pmd;
pte_t *pte;
pmd = pti_user_pagetable_walk_pmd(address);
if (!pmd)
return NULL;
That clearly separtes the variables from the _code_ and also nicely
pairs the action with the check for that action being successful.
Powered by blists - more mailing lists