[lkml]   [2008]   [May]   [29]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: [PATCH] [genirq] Expose default irq affinity mask (take 2)
Review comments ... just nits ... patch looks good overall.

1. Trailing whitespace on Doc lines:
+irq subdir is one subdir for each IRQ, and two files; default_smp_affinity and
+smp_affinity is a bitmask, in which you can specify which CPUs can handle the
+default_smp_affinity mask applies to all non-active IRQs. In other words IRQs

(Granted, there are already about 300 hundred trailing whitespaces in that file,
so this is like trying not to spit into the Pacific ocean ;).

2. I think that the following line should be removed from kernel/irq/proc.c:

+extern cpumask_t irq_default_affinity;

(You already pick up that extern from <linux/irq.h>, as it should be.)

3. The trailing period was missing after the line:

irq subdir is one subdir for each IRQ, and one file; prof_cpu_mask

Might as well add it, after "prof_cpu_mask", while you're there anyway.

4. In the doc:

+ > echo 1 > /proc/irq/5/smp_affinity
+This means that only the first CPU will handle the IRQ, but you can also echo
+5 which means that only the first and fourth CPU can handle the IRQ.

it might be a good idea to not use "5" for both the example CPU
and for the example mask. It increases the chance of a reader
getting confused as to which 5 is which. Perhaps try a different
example CPU, such as:

echo 1 > /proc/irq/2/smp_affinity

5. The original Doc text, before your patch, stated that the
"The contents of the prof_cpu_mask file ... by default."

After your patch, this statement that the default contents
of the prof_cpu_mask file was ffffffff got lost.

6. The grammar of the following got mangled a little:

+default_smp_affinity mask applies to all non-active IRQs. In other words IRQs
+that have not been allocated/activated yet (for which /proc/irq/[0-9]* directory
+does not exist).

Perhaps the following is better:

+The default_smp_affinity mask applies to all non-active IRQs, which are the
+IRQs which have not yet been allocated/activated, and hence which lack a
+/proc/irq/[0-9]* directory.

7. The lines of code (with their definition of cpumask_t tmp):
+ cpus_and(tmp, new_value, cpu_online_map);
+ if (cpus_empty(tmp))
+ return -EINVAL;

can be better written, I'm pretty sure, as:
+ if (!cpus_intersects(new_value, cpu_online_map))
+ return -EINVAL;

Minimizing cpumask_t's as local stack variables is a good thing,
especially on some of the high (thousands) count CPU systems that
SGI is developing.

Nice work. Thanks.

I won't rest till it's the best ...
Programmer, Linux Scalability
Paul Jackson <> 1.940.382.4214

 \ /
  Last update: 2008-05-29 06:55    [W:0.056 / U:23.288 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site