linux-kernel - Re: [PATCH v2] smackfs: restrict bytes count in smackfs write functions

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: <8d66b6fd-81d3-38bd-703f-522a2e2d6fca@i-love.sakura.ne.jp>
Date:   Thu, 28 Jan 2021 23:24:00 +0900
From:   Tetsuo Handa <penguin-kernel@...ove.sakura.ne.jp>
To:     Sabyrzhan Tasbolatov <snovitoll@...il.com>
Cc:     andreyknvl@...gle.com, casey@...aufler-ca.com, jmorris@...ei.org,
        linux-kernel@...r.kernel.org,
        linux-security-module@...r.kernel.org, mhocko@...e.com,
        serge@...lyn.com,
        syzbot+a71a442385a0b2815497@...kaller.appspotmail.com
Subject: Re: [PATCH v2] smackfs: restrict bytes count in smackfs write
 functions

On 2021/01/28 22:27, Sabyrzhan Tasbolatov wrote:
>> Doesn't this change break legitimate requests like
>>
>>   char buffer[20000];
>>
>>   memset(buffer, ' ', sizeof(buffer));
>>   memcpy(buffer + sizeof(buffer) - 10, "foo", 3);
>>   write(fd, buffer, sizeof(buffer));
>>
>> ?
> 
> It does, in this case. Then I need to patch another version with
> whitespace stripping before, after label. I just followed the same thing
> that I see in security/selinux/selinuxfs.c sel_write_enforce() etc.
> 
> It has the same memdup_user_nul() and count >= PAGE_SIZE check prior to that.

Since sel_write_enforce() accepts string representation of an integer value, PAGE_SIZE is sufficient.
But since smk_write_onlycap() and smk_write_relabel_self() accept list of space-delimited words,
you need to prove why PAGE_SIZE does not break userspace in your patch.

Also, due to the "too small to fail" memory-allocation rule, memdup_user_nul() for
count < PAGE_SIZE * 8 bytes is "never fails with -ENOMEM unless SIGKILLed by the OOM
killer". Also, memdup_user_nul() for count >= PAGE_SIZE * (1 << MAX_ORDER) - 1 bytes is
"never succeeds". Thus, you can safely add

	if (count >= PAGE_SIZE * (1 << MAX_ORDER) - 1)
		return -EINVAL; // or -ENOMEM if you want compatibility

to smackfs write functions. But it is a strange requirement that the caller of
memdup_user_nul() has to be aware of upper limit in a way that we won't hit

	/*
	 * There are several places where we assume that the order value is sane
	 * so bail out early if the request is out of bound.
	 */
	if (unlikely(order >= MAX_ORDER)) {
		WARN_ON_ONCE(!(gfp_mask & __GFP_NOWARN));
		return NULL;
	}

path. memdup_user_nul() side should do

	if (count >= PAGE_SIZE * (1 << MAX_ORDER) - 1)
		return -ENOMEM;

check and return -ENOMEM if memdup_user_nul() does not want to use __GFP_NOWARN.
I still believe that memdup_user_nul() side should be fixed.