Important per-cpu fix.
Quoted 8 lines · In article <[email protected]> Rusty Russell wrote:
Frankly, I'm amazed the kernel worked for long without this.
Every linker script thinks the section is called .data.percpu. Without this patch, every CPU ends up sharing the same "per-cpu" variable.
This might explain the wierd per-cpu problem reports from Andrew and Dave, and also that nagging feeling that I'm an idiot...
Not only does this fix the tasklet BUG with 2.5.32 but it also fixes a serial console hang with my 2.5.32 version of Ingo/Davem/Alexey's scalable timers code that I have been debugging for the last two days. I use
a per-cpu tasklet to run the timers, so it was probably killing me there.
Thanks
Signature · 2 lines
Dipankar Sarma <[email protected]> http://lse.sourceforge.net Linux Technology Center, IBM Software Lab, Bangalore, India.
Quoted 3 lines
Quoted 2 lines · In article <[email protected]> Rusty Russell wrote:
This might explain the wierd per-cpu problem reports from Andrew and Dave, and also that nagging feeling that I'm an idiot...
Quoted 5 lines · Dipankar Sarma
Not only does this fix the tasklet BUG with 2.5.32 but it also fixes a serial console hang with my 2.5.32 version of Ingo/Davem/Alexey's scalable timers code that I have been debugging for the last two days. I use
a per-cpu tasklet to run the timers, so it was probably killing me there.
Are you running on NUMA-Q? Do you also see the tty deadlock?
Thanks,
Bill
Quoted 2 lines · William Lee Irwin III
Are you running on NUMA-Q? Do you also see the tty deadlock?
No, a 4CPU PIII box. What tty deadlock ?
Thanks
Signature · 2 lines
Dipankar Sarma <[email protected]> http://lse.sourceforge.net Linux Technology Center, IBM Software Lab, Bangalore, India.
Quoted 1 line · William Lee Irwin III
Quoted 1 line
Are you running on NUMA-Q? Do you also see the tty deadlock?
Quoted 2 lines · Dipankar Sarma
No, a 4CPU PIII box. What tty deadlock ?
Thanks
The one reported in the "mysterious tty deadlock" thread. Basically, some kind of badness happens where it has to be sent a signal-generating character before generating output a good chunk of the time, where extra carriage returns sometimes help.
Cheers,
Bill