Re: 2.5.73-mm1 nbd: boot hang in add_disk at first call from nbd_init

[Michael Frank <mflt1@micrologica.com.hk>]

On Sunday 29 June 2003 01:20, Randy.Dunlap wrote:
> > On Saturday 28 June 2003 12:55, Michael Frank wrote:
> >> On Saturday 28 June 2003 10:41, Andrew Morton wrote:
> >> > Michael Frank <mflt1@micrologica.com.hk> wrote:
> >> > > Changes were recently made to the nbd.c in 2.5.73-mm1
> >> >
> >> > And tons more will be in -mm2, which I shall prepare right now. Please
> >>
> >> retest on that and if it still hangs, capture the output from pressing
> >> alt-sysrq-T.
> >>
> >> Legacy free, no serial port.
> >>
> >>
> >>
> >> Sorry, -mm2 hang at booting kernel on 2 machines.
> >
> > Oh Murphy! Bug: 250K log buffer causes a hang on boot.
> >
> > Sorry for the shock. I configured the log buffer bigger - 250K and it
> > hangs on boot.
> >
> > Default 14K log buffer all OK, the NBD hang is fixed too.
>
> Default value of 14 is a shift count (2 << 14), which gives a

No sh**,

> 16 KB buffer.
> Did you enter '250' for the shift value?

Yes, I meant 250K bytes. 

> Yes, that wouldn't boot.
> Maybe consult the help text??

I'll put it on a CD under my pillow tonight....

>
> > This was my only config change besides that driver which didn't compile
> > ;)
> >
> > I want a bigger log buffer in preparation for testing swsusp on 2.5. On
> > 2.4, the test io load prevent the big swsusp logs from making it to
> > disk...
>
> Andrew, do you want a min/max limit on the LOG_BUF_SHIFT value,
> now that Roman has added that feature for Kconfig?
>

Making this a shift count is a brilliant trap designed to humble buffalos 
who do not bother to read the documentation. To put a check there would 
just spoil the fun ;)

-- 
Powered by linux-2.5.73-mm2, compiled with gcc-2.95-3 - not fancy but rock solid

My current linux related activities:
- Test development and testing of swsusp
- Everyday usage of 2.5 kernel

More info on the 2.5 kernel: http://www.codemonkey.org.uk/post-halloween-2.5.txt