linux-kernel - Re: [PATCH v1] Device-mapper(LVM): Convert to use time_after

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [day] [month] [year] [list]

Message-ID: <9cdc4de9-82b8-87fe-4845-6b18ffe26d18@redhat.com>
Date: Mon, 26 Aug 2024 11:23:26 +0200 (CEST)
From: Mikulas Patocka <mpatocka@...hat.com>
To: 陈玉凡 <chenyufan@...o.com>
cc: Alasdair Kergon <agk@...hat.com>, Mike Snitzer <snitzer@...nel.org>, 
    "dm-devel@...ts.linux.dev" <dm-devel@...ts.linux.dev>, 
    "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>, 
    "opensource.kernel" <opensource.kernel@...o.com>
Subject: Re: [PATCH v1] Device-mapper(LVM): Convert to use time_after_eq
 macro



On Mon, 26 Aug 2024, 陈玉凡 wrote:

> 在 2024/8/24 3:01, Mikulas Patocka 写道:
> 
> 
> On Fri, 23 Aug 2024, 陈玉凡 wrote:
> 
> 在 2024/8/23 1:08, Mikulas Patocka 写道:
> 
> On Thu, 22 Aug 2024, Chen Yufan wrote:
> 
> Use time_after_eq macro instead of opening it for readability.
> 
> Signed-off-by: Chen Yufan <chenyufan@...o.com>
> ---
>   drivers/md/dm-writecache.c | 5 +++--
>   1 file changed, 3 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/md/dm-writecache.c b/drivers/md/dm-writecache.c
> index 7ce8847b3..548d4d37e 100644
> --- a/drivers/md/dm-writecache.c
> +++ b/drivers/md/dm-writecache.c
> @@ -16,6 +16,7 @@
>   #include <linux/pfn_t.h>
>   #include <linux/libnvdimm.h>
>   #include <linux/delay.h>
> +#include <linux/jiffies.h>
>   #include "dm-io-tracker.h"
>   
>   #define DM_MSG_PREFIX "writecache"
> @@ -1994,8 +1995,8 @@ static void writecache_writeback(struct work_struct *work)
>   	while (!list_empty(&wc->lru) &&
>   	       (wc->writeback_all ||
>   		wc->freelist_size + wc->writeback_size <= wc->freelist_low_watermark ||
> -		(jiffies - container_of(wc->lru.prev, struct wc_entry, lru)->age >=
> -		 wc->max_age - wc->max_age / MAX_AGE_DIV))) {
> +		(time_after_eq(jiffies, container_of(wc->lru.prev, struct wc_entry, lru)->age +
> +		 (wc->max_age - wc->max_age / MAX_AGE_DIV))))) {
>   
>   		n_walked++;
>   		if (unlikely(n_walked > WRITEBACK_LATENCY) &&
> -- 
> 2.39.0
> 
> I'm not sure about this. The old and new code is not really equivalent.
> 
> Mikulas
> 
> The code here is susceptible to overflow issues, and the time_*() macros 
> can handle this.
> 
> Chen
> 
> So, describe some case (i.e. the values of jiffies, 
> container_of(wc->lru.prev, struct wc_entry, lru)->age and wc->max_age) 
> where the old code misbehaves and the new code doesn't.
> 
> If we want to fix a bug, we need to know what the bug actually is.
> 
> Mikulas
> 
> When jiffies increased beyond the maximum value of unsigned long, it 
> wraps around to zero, and the value of jiffies would be smaller than the 
> container_of(wc->lru.prev, struct wc_entry, lru)->age value despite 
> logically being larger. Eventurally, because of the wraparound, the 
> result of the condition would be wrong.
> 
> Chen

For example, if "jiffies" is 0x10 (because it wrapped around) and 
"container_of(wc->lru.prev, struct wc_entry, lru)->age" is 0xfffffff0, 
then the expression "jiffies - container_of(wc->lru.prev, struct wc_entry, 
lru)->age" would be 0x20. That is the correct value, I don't see any 
problem with this.

Mikulas