Netdev List
 help / color / mirror / Atom feed
From: Vladimir Oltean <olteanv@gmail.com>
To: "Russell King (Oracle)" <linux@armlinux.org.uk>
Cc: Marek Vasut <marex@denx.de>,
	netdev@vger.kernel.org, "David S. Miller" <davem@davemloft.net>,
	Andrew Lunn <andrew@lunn.ch>,
	Arun Ramadoss <arun.ramadoss@microchip.com>,
	Eric Dumazet <edumazet@google.com>,
	Florian Fainelli <f.fainelli@gmail.com>,
	Jakub Kicinski <kuba@kernel.org>, Paolo Abeni <pabeni@redhat.com>,
	UNGLinuxDriver@microchip.com,
	Woojung Huh <woojung.huh@microchip.com>,
	stable@vger.kernel.org
Subject: Re: [PATCH] net: dsa: microchip: Fix gigabit set and get function for KSZ87xx
Date: Wed, 22 Feb 2023 23:25:26 +0200	[thread overview]
Message-ID: <20230222212526.bfcef4dsg6crwlmh@skbuf> (raw)
In-Reply-To: <Y/YPfxg8Ackb8zmW@shell.armlinux.org.uk>

On Wed, Feb 22, 2023 at 12:50:07PM +0000, Russell King (Oracle) wrote:
> Looking at this driver, I have to say that it looks utterly vile
> from the point of view of being sure that it is correct, and I
> think this patch illustrates why.
...
> Now, what about other KSZ devices - I've analysed this for the KSZ8795,
> but what about any of the others which use this register table? It
> looks to me like those that use ksz8795_regs[] all use ksz8_dev_ops
> and the same masks and bitvals, so they should be the same.
> 
> That is a hell of a lot of work to prove that setting both
> P_XMII_CTRL_0 and P_XMII_CTRL_1 to point at the same register is
> in fact safe. Given the number of registers, the masks, and bitval
> arrays, doing this to prove every combination and then analysing
> the code is utterly impractical - and thus why I label this driver
> as "vile". Is there really no better option to these register
> arrays, bitval arrays and mask arrays - something that makes it
> easier to review and prove correctness?

Only my 2 cents. What is utterly vile is the decision of hardware design
to break software compatibility in such a deliberate and gratuitous way
across switch generations. A driver can only do so much when fed with
such hardware as input.

The ksz driver could use struct reg_field from regmap to mitigate that
to a certain extent (like the ocelot driver does), but certain quirks
will still remain present in the ksz driver. For example, the "bitval"
array. The value "1" written to the P_GMII_1GBIT reg_field indicates
gigabit for ksz8795, but !gigabit on ksz9477. I am not aware of any
abstraction to mask that away in common code other than the bitvals.

Even with struct reg_field, it would still not address the fundamental
problem which is simply that the register fields responsible for a
certain function have hopped so much from generation to generation,
that getting all offsets and bits right for each generation is a
challenge in itself.

  parent reply	other threads:[~2023-02-22 21:25 UTC|newest]

Thread overview: 22+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-02-22  3:17 [PATCH] net: dsa: microchip: Fix gigabit set and get function for KSZ87xx Marek Vasut
2023-02-22  3:52 ` Arun.Ramadoss
2023-02-22 12:50 ` Russell King (Oracle)
2023-02-22 13:03   ` Russell King (Oracle)
2023-02-22 15:10     ` Marek Vasut
2023-02-22 15:56       ` Russell King (Oracle)
2023-02-22 16:30         ` Marek Vasut
2023-02-22 16:39           ` Russell King (Oracle)
2023-02-22 18:43             ` Marek Vasut
2023-02-22 19:12               ` Russell King (Oracle)
2023-02-22 21:25   ` Vladimir Oltean [this message]
2023-02-22 21:08 ` Vladimir Oltean
2023-02-22 22:05   ` Marek Vasut
2023-02-22 22:31     ` Vladimir Oltean
2023-02-22 22:58       ` Marek Vasut
2023-02-22 23:05         ` Florian Fainelli
2023-02-22 23:07         ` Russell King (Oracle)
2023-02-22 23:21         ` Vladimir Oltean
2023-02-22 23:55           ` Marek Vasut
2023-02-23  0:22             ` Vladimir Oltean
2023-02-23  5:17               ` Marek Vasut
2023-02-23 14:20                 ` Vladimir Oltean

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20230222212526.bfcef4dsg6crwlmh@skbuf \
    --to=olteanv@gmail.com \
    --cc=UNGLinuxDriver@microchip.com \
    --cc=andrew@lunn.ch \
    --cc=arun.ramadoss@microchip.com \
    --cc=davem@davemloft.net \
    --cc=edumazet@google.com \
    --cc=f.fainelli@gmail.com \
    --cc=kuba@kernel.org \
    --cc=linux@armlinux.org.uk \
    --cc=marex@denx.de \
    --cc=netdev@vger.kernel.org \
    --cc=pabeni@redhat.com \
    --cc=stable@vger.kernel.org \
    --cc=woojung.huh@microchip.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox