linux-kernel - Re: [PATCH 1/3] zram: use u32 for entry ac

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

Message-ID: <20260114124522.1326519-1-clm@meta.com>
Date: Wed, 14 Jan 2026 04:45:07 -0800
From: Chris Mason <clm@...a.com>
To: Sergey Senozhatsky <senozhatsky@...omium.org>
CC: Chris Mason <clm@...a.com>, Andrew Morton <akpm@...ux-foundation.org>,
        Minchan Kim <minchan@...nel.org>, Brian Geffon <bgeffon@...gle.com>,
        "David
 Stevens" <stevensd@...gle.com>, <linux-kernel@...r.kernel.org>,
        <linux-mm@...ck.org>, <linux-block@...r.kernel.org>
Subject: Re: [PATCH 1/3] zram: use u32 for entry ac_time tracking

On Mon, 15 Dec 2025 14:47:11 +0900 Sergey Senozhatsky <senozhatsky@...omium.org> wrote:

> We can reduce sizeof(zram_table_entry) on 64-bit systems
> by converting flags and ac_time to u32.  Entry flags fit
> into u32, and for ac_time u32 gives us over a century of
> entry lifespan (approx 136 years) which is plenty (zram
> uses system boot time (seconds)).
> 
> In struct zram_table_entry we use bytes aliasing, because
> bit-wait API (for slot lock) requires a whole unsigned
> long word.

Hi everyone,

I ran mm-new through the AI patch review prompts, and this one was flagged:

> diff --git a/drivers/block/zram/zram_drv.c b/drivers/block/zram/zram_drv.c
> --- a/drivers/block/zram/zram_drv.c
> +++ b/drivers/block/zram/zram_drv.c

[ ... ]

> @@ -222,7 +221,7 @@ static void zram_accessed(struct zram *zram, u32 index)
>  	zram_clear_flag(zram, index, ZRAM_IDLE);
>  	zram_clear_flag(zram, index, ZRAM_PP_SLOT);
>  #ifdef CONFIG_ZRAM_TRACK_ENTRY_ACTIME
> -	zram->table[index].ac_time = ktime_get_boottime();
> +	zram->table[index].attr.ac_time = (u32)ktime_get_boottime_seconds();
>  #endif
>  }

This stores seconds in ac_time.

[ ... ]

> @@ -1588,7 +1588,7 @@ static ssize_t read_block_state(struct file *file, char __user *buf,
>  		if (!zram_allocated(zram, index))
>  			goto next;
>
> -		ts = ktime_to_timespec64(zram->table[index].ac_time);
> +		ts = ktime_to_timespec64(zram->table[index].attr.ac_time);
>  		copied = snprintf(kbuf + written, count,
>  			"%12zd %12lld.%06lu %c%c%c%c%c%c\n",
>  			index, (s64)ts.tv_sec,
>  			ts.tv_nsec / NSEC_PER_USEC,

ktime_to_timespec64() is defined as ns_to_timespec64(), which expects
nanoseconds.  Since ac_time now stores seconds, will this produce
incorrect output?

For example, if ac_time is 3600 (representing 1 hour of uptime),
ns_to_timespec64(3600) would compute ts.tv_sec = 3600 / 1000000000 = 0
and ts.tv_nsec = 3600, resulting in "0.000003" instead of "3600.000000".