linux-kernel - Re: [PATCH 3/3] can: c_can: cache frames to operate as a true FIFO

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]

Date:   Mon, 10 May 2021 14:25:12 +0200
From:   Marc Kleine-Budde <mkl@...gutronix.de>
To:     Dario Binacchi <dariobin@...ero.it>
Cc:     linux-kernel@...r.kernel.org,
        "David S. Miller" <davem@...emloft.net>,
        Gianluca Falavigna <gianluca.falavigna@...ind.it>,
        Jakub Kicinski <kuba@...nel.org>,
        Oliver Hartkopp <socketcan@...tkopp.net>,
        Vincent Mailhol <mailhol.vincent@...adoo.fr>,
        Wolfgang Grandegger <wg@...ndegger.com>,
        linux-can@...r.kernel.org, netdev@...r.kernel.org
Subject: Re: [PATCH 3/3] can: c_can: cache frames to operate as a true FIFO

On 09.05.2021 14:43:09, Dario Binacchi wrote:
> As reported by a comment in the c_can_start_xmit() this was not a FIFO.
> C/D_CAN controller sends out the buffers prioritized so that the lowest
> buffer number wins.
> 
> What did c_can_start_xmit() do if it found tx_active = 0x80000000 ? It
> waited until the only frame of the FIFO was actually transmitted by the
> controller. Only one message in the FIFO but we had to wait for it to
> empty completely to ensure that the messages were transmitted in the
> order in which they were loaded.
> 
> By storing the frames in the FIFO without requiring its transmission, we
> will be able to use the full size of the FIFO even in cases such as the
> one described above. The transmission interrupt will trigger their
> transmission only when all the messages previously loaded but stored in
> less priority positions of the buffers have been transmitted.

The algorithm you implemented looks a bit too complicated to me. Let me
sketch the algorithm that's implemented by several other drivers.

- have a power of two number of TX objects
- add a number of objects to struct priv (tx_num)
  (or make it a define, if the number of tx objects is compile time fixed)
- add two "unsigned int" variables to your struct priv,
  one "tx_head", one "tx_tail"
- the hard_start_xmit() writes to priv->tx_head & (priv->tx_num - 1)
- increment tx_head
- stop the tx_queue if there is no space or if the object with the
  lowest prio has been written
- in TX complete IRQ, handle priv->tx_tail object
- increment tx_tail
- wake queue if there is space but don't wake if we wait for the lowest
  prio object to be TX completed.

Special care needs to be taken to implement that lock-less and race
free. I suggest to look the the mcp251xfd driver.

Marc

-- 
Pengutronix e.K.                 | Marc Kleine-Budde           |
Embedded Linux                   | https://www.pengutronix.de  |
Vertretung West/Dortmund         | Phone: +49-231-2826-924     |
Amtsgericht Hildesheim, HRA 2686 | Fax:   +49-5121-206917-5555 |

Download attachment "signature.asc" of type "application/pgp-signature" (489 bytes)