Received-SPF: pass (google.com: domain of linux-kernel+bounces-175934-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45e3:2400::1 as permitted sender) client-ip=2604:1380:45e3:2400::1;
Date: Fri, 10 May 2024 17:15:08 +0200
From: Jan Kara <jack@suse.cz>
To: Justin Stitt <justinstitt@google.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>,
	Christian Brauner <brauner@kernel.org>, Jan Kara <jack@suse.cz>,
	Nathan Chancellor <nathan@kernel.org>,
	Bill Wendling <morbo@google.com>, linux-fsdevel@vger.kernel.org,
	linux-kernel@vger.kernel.org, llvm@lists.linux.dev,
	linux-hardening@vger.kernel.org
Subject: Re: [PATCH] fs: fix unintentional arithmetic wraparound in offset
 calculation
Message-ID: <20240510151508.hajqjxsn7rghk3dj@quack3>
References: <20240509-b4-sio-read_write-v1-1-06bec2022697@google.com>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20240509-b4-sio-read_write-v1-1-06bec2022697@google.com>

On Thu 09-05-24 21:34:58, Justin Stitt wrote:
> When running syzkaller with the newly reintroduced signed integer
> overflow sanitizer we encounter this report:
> 
> [   67.991989] ------------[ cut here ]------------
> [   67.995501] UBSAN: signed-integer-overflow in ../fs/read_write.c:91:10
> [   68.000067] 9223372036854775807 + 4096 cannot be represented in type 'loff_t' (aka 'long long')
> [   68.006266] CPU: 4 PID: 10851 Comm: syz-executor.5 Not tainted 6.8.0-rc2-00035-gb3ef86b5a957 #1
> [   68.012353] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.3-debian-1.16.3-2 04/01/2014
> [   68.018983] Call Trace:
> [   68.020803]  <TASK>
> [   68.022540]  dump_stack_lvl+0x93/0xd0
> [   68.025222]  handle_overflow+0x171/0x1b0
> [   68.028053]  generic_file_llseek_size+0x35b/0x380
> ...
> 
> Historically, the signed integer overflow sanitizer did not work in the
> kernel due to its interaction with `-fwrapv` but this has since been
> changed [1] in the newest version of Clang. It was re-enabled in the
> kernel with Commit 557f8c582a9ba8ab ("ubsan: Reintroduce signed overflow
> sanitizer").
> 
> Since @offset is later limited by @maxsize, we can proactively safeguard
> against exceeding that value and also dodge some accidental overflow
> (which may cause bad file access):
> 
> 	loff_t vfs_setpos(struct file *file, loff_t offset, loff_t maxsize)
> 	{
> 		if (offset < 0 && !unsigned_offsets(file))
> 			return -EINVAL;
> 		if (offset > maxsize)
> 			return -EINVAL;
> 		...
> 
> Link: https://github.com/llvm/llvm-project/pull/82432 [1]
> Closes: https://github.com/KSPP/linux/issues/358
> Cc: linux-hardening@vger.kernel.org
> Signed-off-by: Justin Stitt <justinstitt@google.com>
> ---
> Here's the syzkaller reproducer:
> | # {Threaded:false Repeat:false RepeatTimes:0 Procs:1 Slowdown:1 Sandbox:
> | # SandboxArg:0 Leak:false NetInjection:false NetDevices:false
> | # NetReset:false Cgroups:false BinfmtMisc:false CloseFDs:false KCSAN:false
> | # DevlinkPCI:false NicVF:false USB:false VhciInjection:false Wifi:false
> | # IEEE802154:false Sysctl:false Swap:false UseTmpDir:false
> | # HandleSegv:false Repro:false Trace:false LegacyOptions:{Collide:false
> | # Fault:false FaultCall:0 FaultNth:0}}
> | r0 = openat$sysfs(0xffffffffffffff9c, &(0x7f0000000000)='/sys/kernel/address_bits', 0x0, 0x98)
> | lseek(r0, 0x7fffffffffffffff, 0x2)
> 
> ... which was used against Kees' tree here (v6.8rc2):
> https://git.kernel.org/pub/scm/linux/kernel/git/kees/linux.git/log/?h=wip/v6.9-rc2/unsigned-overflow-sanitizer
> 
> ... with this config:
> https://gist.github.com/JustinStitt/824976568b0f228ccbcbe49f3dee9bf4
> ---
>  fs/read_write.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/fs/read_write.c b/fs/read_write.c
> index d4c036e82b6c..10c3eaa5ef55 100644
> --- a/fs/read_write.c
> +++ b/fs/read_write.c
> @@ -88,7 +88,7 @@ generic_file_llseek_size(struct file *file, loff_t offset, int whence,
>  {
>  	switch (whence) {
>  	case SEEK_END:
> -		offset += eof;
> +		offset = min_t(loff_t, offset, maxsize - eof) + eof;

Well, but by this you change the behavior of seek(2) for huge offsets.
Previously we'd return -EINVAL (from following vfs_setpos()), now we set
position to maxsize. I don't think that is desirable?

Also the addition in SEEK_CUR could overflow in the same way AFAICT so we
could treat that in one patch so that the whole function is fixed at once?

								Honza
-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR