Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Mon, 10 May 2021 10:24:27 +0000
From:   Quentin Perret <qperret@google.com>
To:     gregkh@linuxfoundation.org, masahiroy@kernel.org
Cc:     linux-kernel@vger.kernel.org, maennich@google.com,
        gprocida@google.com, kernel-team@android.com
Subject: Re: [PATCH] export: Make CRCs robust to symbol trimming
Message-ID: <YJkJ21n71SIkUppu@google.com>
References: <20210408180105.2496212-1-qperret@google.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20210408180105.2496212-1-qperret@google.com>
Precedence: bulk

Hi,

On Thursday 08 Apr 2021 at 18:01:05 (+0000), Quentin Perret wrote:
> The CRC calculation done by genksyms is triggered when the parser hits
> EXPORT_SYMBOL*() macros. At this point, genksyms recursively expands the
> types, and uses that as the input for the CRC calculation. In the case
> of forward-declared structs, the type expands to 'UNKNOWN'. Next, the
> result of the expansion of each type is cached, and is re-used when/if
> the same type is seen again for another exported symbol in the file.
> 
> Unfortunately, this can cause CRC 'stability' issues when a struct
> definition becomes visible in the middle of a C file. For example, let's
> assume code with the following pattern:
> 
>     struct foo;
> 
>     int bar(struct foo *arg)
>     {
> 	/* Do work ... */
>     }
>     EXPORT_SYMBOL_GPL(bar);
> 
>     /* This contains struct foo's definition */
>     #include "foo.h"
> 
>     int baz(struct foo *arg)
>     {
> 	/* Do more work ... */
>     }
>     EXPORT_SYMBOL_GPL(baz);
> 
> Here, baz's CRC will be computed using the expansion of struct foo that
> was cached after bar's CRC calculation ('UNKOWN' here). But if
> EXPORT_SYMBOL_GPL(bar) is removed from the file (because of e.g. symbol
> trimming using CONFIG_TRIM_UNUSED_KSYMS), struct foo will be expanded
> late, during baz's CRC calculation, which now has visibility over the
> full struct definition, hence resulting in a different CRC for baz.
> 
> This can cause annoying issues for distro kernel (such as the Android
> Generic Kernel Image) which use CONFIG_UNUSED_KSYMS_WHITELIST. Indeed,
> as per the above, adding a symbol to the whitelist can change the CRC of
> symbols that are already kept exported. As such, modules built against a
> kernel with a trimmed ABI may not load against the same kernel built
> with an extended whitelist, even though they are still strictly binary
> compatible. While rebuilding the modules would obviously solve the
> issue, I believe this classifies as an odd genksyms corner case, and it
> gets in the way of kernel updates in the GKI context.
> 
> To work around the issue, make sure to keep issuing the
> __GENKSYMS_EXPORT_SYMBOL macros for all trimmed symbols, hence making
> the genksyms parsing insensitive to symbol trimming.
> 
> Signed-off-by: Quentin Perret <qperret@google.com>

Gentle ping.

Is there anything else I should do in this one?

Thanks,
Quentin