linux-kernel - Re: [PATCH v1 1/1] serial: 8250_pci1xxxx: Drop quirk from 8250

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, 19 Feb 2024 18:19:42 +0200
From: Andy Shevchenko <andriy.shevchenko@...ux.intel.com>
To: Rengarajan.S@...rochip.com
Cc: linux-serial@...r.kernel.org, gregkh@...uxfoundation.org,
	linux-kernel@...r.kernel.org, jirislaby@...nel.org,
	Kumaravel.Thiagarajan@...rochip.com,
	Tharunkumar.Pasumarthi@...rochip.com
Subject: Re: [PATCH v1 1/1] serial: 8250_pci1xxxx: Drop quirk from 8250_port

On Thu, Feb 15, 2024 at 09:26:21AM +0000, Rengarajan.S@...rochip.com wrote:
> On Wed, 2024-02-14 at 15:50 +0200, Andy Shevchenko wrote:

..

> > +       /*
> > +        * 8250 core considers prescaller value to be always 16.
> > +        * The MCHP ports support downscaled mode and hence the
> > +        * functional UART clock can be lower, i.e. 62.5MHz, than
> > +        * software expects in order to support higher baud rates.
> > +        * Assign here 64MHz to support 4Mbps.
> > +        *
> > +        * The value itself is not really used anywhere except baud
> > +        * rate calculations, so we can mangle it as we wish.
> > +        */
> > +       port->port.uartclk = 64 * HZ_PER_MHZ;
> 
> As per internal MCHP DOS, PCI1XXXX driver uses a simple method of
> converting "legacy 16 bit baud rate generator" to a "32 bit fractional
> baud rate generator" which enables generation of an acceptable baud
> rate from any valuable frequency.
> 
> This is applicable only when the baud clock selected is 62.5 MHz, so
> when we configure the baud clock to 64 MHz(as above) will it be
> downscaled to 62.5 MHz, thus supporting the above feature? 

I specifically added the above comment. If you look closer, your driver does
not use this value at all, the 8250 port code uses it in several places:

- 8250_rsa case (not applicable to your driver)

- probe_baud() call (applicable iff the kernel command line misses the
  baudrate, but even without this patch it's broken for your driver)

- serial8250_update_uartclk() call (not applicable to your driver)

- serial8250_get_baud_rate() call (only to get max and min range;
  my change will have an effect on min (max is exactly what your
  quirk is doing right no), so 62500000/16/65535 ~= 59.6, while
  with my change 64000000/16/65535 ~= 61.0, but standard baudrate
  here is 50 and 75, the former isn't supported by the existing
  code either

- serial8250_do_get_divisor() call when magic_multiplier supplied
  (not applicable to your driver)

- autoconfig_16550a() call (not applicable to your driver)

Hope this clarifies the case.

Of course if you able to test, will be even better.
But wait for v2 where I update what Greg caught.

..

> Acked-by: Rengarajan S <rengarajan.s@...rochip.com>

Thank you!

-- 
With Best Regards,
Andy Shevchenko