MIME-Version: 1.0
References: <Y/cBB+q0Ono9j2Jy@gondor.apana.org.au> <20230224231430.2948-1-kunyu@nfschina.com>
 <Y/cy5wUtk10OahpO@gondor.apana.org.au> <CACRpkdYyB=-UnE1bmdVszSSB5ReECZ0fUoWJX6XtYbKHEe52tA@mail.gmail.com>
 <Y/c7iVW67Xhhdu8e@gondor.apana.org.au> <Y/hQdzsKMYgkIfMY@gondor.apana.org.au>
In-Reply-To: <Y/hQdzsKMYgkIfMY@gondor.apana.org.au>
From:   Linus Walleij <linus.walleij@linaro.org>
Date:   Sat, 25 Feb 2023 01:01:46 +0100
Message-ID: <CACRpkdZe3cMMxJesD0mpqHTwvuWHjSGVHsiFUQQyuA+VWknMTQ@mail.gmail.com>
Subject: Re: [v2 PATCH] crypto: stm32 - Save and restore between each request
To:     Herbert Xu <herbert@gondor.apana.org.au>
Cc:     Lionel Debieve <lionel.debieve@foss.st.com>,
        Li kunyu <kunyu@nfschina.com>, davem@davemloft.net,
        linux-arm-kernel@lists.infradead.org, linux-crypto@vger.kernel.org,
        linux-kernel@vger.kernel.org,
        linux-stm32@st-md-mailman.stormreply.com, mcoquelin.stm32@gmail.com
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

Hi Herbert,

I tested this on the Ux500 and sadly this happens
already in probe():

[    2.802539] stm32-hash a03c2000.hash: dma-maxburst not specified, using 0
[    2.809342] stm32-hash a03c2000.hash: No IRQ, use polling mode
[    2.815226] stm32-hash a03c2000.hash: DMA mode not available
[    2.821140] stm32-hash a03c2000.hash: will run requests pump with
realtime priority
[    2.828815] stm32-hash a03c2000.hash: Algo 0 : 0 failed
[    2.834144] stm32-hash: probe of a03c2000.hash failed with error -22

I don't quite understand why, we never get to the tests...

On Fri, Feb 24, 2023 at 6:52 AM Herbert Xu <herbert@gondor.apana.org.au> wrote:

> v2 fixes potential state clobbering from the disconnect between
> hdev->flags and rctx->flags.
>
> ---8<---
> The Crypto API hashing paradigm requires the hardware state to
> be exported between *each* request because multiple unrelated
> hashes may be processed concurrently.
>
> The stm32 hardware is capable of producing the hardware hashing
> state but it was only doing it in the export function.  This is
> not only broken for export as you can't export a kernel pointer
> and reimport it, but it also means that concurrent hashing was
> fundamentally broken.
>
> Fix this by moving the saving and restoring of hardware hash
> state between each and every hashing request.
>
> Fixes: 8a1012d3f2ab ("crypto: stm32 - Support for STM32 HASH module")
> Reported-by: Li kunyu <kunyu@nfschina.com>
> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

I think I understand the direction of the patch but it seems
the pm_runtime_* calls get unbalanced since now this is taken
in
stm32_hash_one_request
 -> stm32_hash_hw_init()
    -> pm_runtime_get_sync()

But no corresponding
pm_runtime_mark_last_busy(hdev->dev);
pm_runtime_put_autosuspend(hdev->dev);

Exist anymore? You know the semantics better than me,
I'm not sure where to put this, I guess where you save
the HW state in stm32_hash_update_cpu()?

I don't know about the DMA case then though.

Yours,
Linus Walleij