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]
Message-ID: <ujs6kaisllqu3qzm76qkwpmdy2vnulp6z742ooclbsdz36zl5f@m7ujgar4pwqs>
Date:   Fri, 6 Oct 2023 17:01:34 +0200
From:   Maxime Ripard <mripard@...nel.org>
To:     Benedikt Spranger <b.spranger@...utronix.de>
Cc:     linux-kernel@...r.kernel.org, linux-clk@...r.kernel.org,
        Stephen Boyd <sboyd@...nel.org>,
        Michael Turquette <mturquette@...libre.com>,
        Dinh Nguyen <dinguyen@...nel.org>
Subject: Re: [PATCH 1/1] clk: socfpga: gate: Fix of by factor 2 for serial
 console

On Thu, Oct 05, 2023 at 08:32:23PM +0200, Benedikt Spranger wrote:
> On Thu, 5 Oct 2023 13:34:01 +0200
> Maxime Ripard <mripard@...nel.org> wrote:
> 
> > Hi,
> > 
> > On Thu, Oct 05, 2023 at 11:59:27AM +0200, Benedikt Spranger wrote:
> > > Commit 9607beb917df ("clk: socfpga: gate: Add a determine_rate
> > > hook") introduce a specific determine_rate hook. As a result the
> > > calculated clock for the serial IP is off by factor 2 after that
> > > i.e. if the system configures a baudrate of 115200 it is set
> > > physicaly to 57600.  
> > 
> > Where is that factor 2 coming from?
> In drivers/tty/serial/8250/8250_dw.c p->uartclk is set twice as high,
> as it should be: 
> 
> dw8250_set_termios() is called and rate is evaluated to 20000000 in the
> bad and 10000000 in the good case. As a result p->uartclk is set to
> 20000000 in the bad case.

Sure, sorry I worded that poorly. What I meant was what clock tree
decision is taken now that wasn't taken before that leads to that factor
2 difference.

> > > Change the determine_rate hook to the reparent variant
> > > __clk_mux_determine_rate() to fix the issue.  
> > 
> > It's also not clear to me why that would fix anything. This patch
> > should only make the old behaviour explicit, could you expand a bit
> > on what happens?
> Booting the kernel with console=ttyS0,115200 result in a corrupted
> character output. Setting the serial terminal application to 57600
> make the serial console working.
> 
> I dug deeper and added some debug output (see patch below):

Thanks for the traces, that's helpful. It looks like the culprit is:

Good:

            init-1       [001] .....     0.125643: clk_rate_request_start: l4_sp_clk min 0 max 4294967295, parent per_base_clk (200000000)
            init-1       [001] .....     0.125651: clk_rate_request_done: l4_sp_clk min 0 max 4294967295, parent per_base_clk (200000000)
            init-1       [001] .....     0.125657: dw8250_set_termios: dw8250_set_termios: rate = 200000000 newrate = 1843200

vs Bad:

            init-1       [001] .....     0.116063: clk_rate_request_start: l4_sp_clk min 0 max 4294967295, parent per_base_clk (200000000)
            init-1       [001] .....     0.116089: clk_rate_request_done: l4_sp_clk min 0 max 4294967295, parent per_base_clk (200000000)
            init-1       [001] .....     0.116096: dw8250_set_termios: dw8250_set_termios: rate = 4294967274 newrate = 1843200

The rate returned is super suspicious, as it's an -EINVAL casted into an
unsigned long. So I think something on that clock chain is returning an
error for some reason, which is then treated as a rate by the rest and
everybody's just confused.

What is the board that you're using?

Maxime

Download attachment "signature.asc" of type "application/pgp-signature" (229 bytes)

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ