Re: CURRENT crashes at early boot on Lenovo T540p: rtsx to blame - 13.0-RELEASE crashes same way!

On Wed, 12 May 2021 16:44:59 +0300
Lev Serebryakov <lev_at_FreeBSD.org> wrote:

> On 12.05.2021 16:34, Gary Jennejohn wrote:
> 
> > Is sysctl debug.debugger_on_panic set to 1?  You should automatically
> > land in ddb if that is set.  I suppose it is, since you posted some
> > back trace in an earlier mail.  It also seems to be the default, at
> > least in my FreeBSD-14 kernel.  
> 
>   I see ddb report about panic, but it's all. Keyboard doesn't work, and panic occurs before any disks are detected, so there is no space for crash dump.
> 
> > AFAIK ddb has a command to generate a crash dump.  But I can't easily
> > check that :(  
> 
>   Yep, but console is dead & there are no disks for crash dump at this stage.
> 

A real challenge to debug.

Maybe you could break into boot and start it with the -d flag (see boot(8))
so that it will enter ddb as early as possible.  Don't know whether you use
the boot menu or not.  I always disable it.

> > It seems like there's a major bug when no SD card is inserted and the
> > driver is in the kernel.  And a timing problem when a card is in the
> > slot at boot time.
> > 
> > Good to know that the module still works.  
> 
>   Yep, smoke tests for module (without inversion!) pass: I could mount R/W SD card with release image, mount FS, read and write files, unmount FS and eject card without panics and errors. I didn't try any high-load tests, though.
> 
> > 
> > Difficult to debug without your laptop model in the hands of a developer.  
> 
>   It is "pretty common" (but old) Lenovo ThinkPad T540p. I could provide remote access to it, but it will not help for such low-level and early panic :-(
> 

I took a look and even refurbished T540p's are still pretty expensive.
On the order of 500 Euros.

-- 
Gary Jennejohn