linux-kernel - Re: [PATCH] mm/zsmalloc: include sparsemem.h for MAX_PHYSMEM

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: <20201109170406.GC449970@google.com>
Date:   Mon, 9 Nov 2020 09:04:06 -0800
From:   Minchan Kim <minchan@...nel.org>
To:     Stefan Agner <stefan@...er.ch>
Cc:     Andrew Morton <akpm@...ux-foundation.org>, ngupta@...are.org,
        sergey.senozhatsky.work@...il.com, sjenning@...ux.vnet.ibm.com,
        gregkh@...uxfoundation.org, arnd@...db.de, linux-mm@...ck.org,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH] mm/zsmalloc: include sparsemem.h for MAX_PHYSMEM_BITS

On Sun, Nov 08, 2020 at 02:16:37AM +0100, Stefan Agner wrote:
> On 2020-11-08 01:56, Andrew Morton wrote:
> > On Sat,  7 Nov 2020 16:22:06 +0100 Stefan Agner <stefan@...er.ch> wrote:
> > 
> >> Most architectures define MAX_PHYSMEM_BITS in asm/sparsemem.h and don't
> >> include it in asm/pgtable.h. Include asm/sparsemem.h directly to get
> >> the MAX_PHYSMEM_BITS define on all architectures.
> >>
> >> This fixes a crash when accessing zram on 32-bit ARM platform with LPAE and
> >> more than 4GB of memory:
> >>   Unable to handle kernel NULL pointer dereference at virtual address 00000000
> > 
> > Mysterious.  Presumably without this include, some compilation unit is
> > picking up the wrong value of MAX_PHYSMEM_BITS?  But I couldn't
> > actually see where/how this occurs.  Can you please explain further?
> 
> Not sure if I got that right, but from what I understand if
> MAX_PHYSMEM_BITS is not set in mm/zsmalloc.c it will set
> MAX_PHYSMEM_BITS to BITS_PER_LONG. And this is 32-bit, too short when
> LPAE is in use...

True. It's headache in the zsmalloc.
Somedays I'd really like to fix it via redesigning of metadata
management.

Thanks for the fixing the ancient bug, Stefan.