linux-kernel - Re: [PATCH] zram: unsafe zram_get_element call in zram_read

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] [day] [month] [year] [list]

Message-ID: <20231108014229.GE11577@google.com>
Date:   Wed, 8 Nov 2023 10:42:29 +0900
From:   Sergey Senozhatsky <senozhatsky@...omium.org>
To:     Vasily Averin <vasily.averin@...ux.dev>
Cc:     Sergey Senozhatsky <senozhatsky@...omium.org>,
        Minchan Kim <minchan@...nel.org>, linux-kernel@...r.kernel.org,
        linux-block@...r.kernel.org
Subject: Re: [PATCH] zram: unsafe zram_get_element call in zram_read_page()

On (23/11/07 21:19), Vasily Averin wrote:
> On 11/7/23 13:40, Sergey Senozhatsky wrote:
> > On (23/11/07 16:39), Sergey Senozhatsky wrote:
> >> Hmmm,
> >> We may want to do more here. Basically, we probably need to re-confirm
> >> after read_from_bdev() that the entry at index still has ZRAM_WB set
> >> and, if so, that it points to the same blk_idx. IOW, check that it has
> >> not been free-ed and re-used under us.
> > --- a/drivers/block/zram/zram_drv.c
> > +++ b/drivers/block/zram/zram_drv.c
> > @@ -1364,14 +1364,21 @@ static int zram_read_page(struct zram *zram, struct page *page, u32 index,
> >  		ret = zram_read_from_zspool(zram, page, index);
> >  		zram_slot_unlock(zram, index);
> >  	} else {
> > +		unsigned long idx = zram_get_element(zram, index);
> >  		/*
> >  		 * The slot should be unlocked before reading from the backing
> >  		 * device.
> >  		 */
> >  		zram_slot_unlock(zram, index);
> >  
> > -		ret = read_from_bdev(zram, page, zram_get_element(zram, index),
> > -				     parent);
> > +		ret = read_from_bdev(zram, page, idx, parent);
> > +		if (ret == 0) {
> > +			zram_slot_lock(zram, index);
> > +			if (!zram_test_flag(zram, index, ZRAM_WB) ||
> > +			    idx != zram_get_element(zram, index))
> > +				ret = -EINVAL;
> > +			zram_slot_unlock(zram, index);
> > +		}
> 
> Why overwritten page can not be pushed to WB to the same blk_idx? 

Yeah, so I thought about it too but didn't want to go too deep into it.
We probably can only address it if we synchronize free_page (?), read_page()
and writeback(), so that we never have concurrent bitmap modifications when
one of the operations is in progress.