Skip to contentSkip to conversation

reiser4 merging action list

DISCUSSION Andi Kleen 3 participants 4 messages Jun 28, 2005 7 messages not in the archive

lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): 200506240334.j5O3YowB008100@laptop11.inf.utfsm.cl.suse.lists.linux.kernel

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

Andi Kleen

author
#7 lore raw
Quoted 10 lines · Hans Reiser

* metafiles should be disabled until we can present code that works
right. Half the list thinks we cannot solve the cycles problem ever. Disable metafiles and postpone problem until working code, or the failure to produce it, makes it possible to do more than rant at each other. This is currently already done in the -mm patches, but is mentioned lest someone think it forgotten.

* update the locking documentation

Probably I forget something.

These are all big picture issues, but I think some low level attention to the individual code is still needed.

Some stuff that stood out from a very quick look:

I would like for the custom spin lock debugging (spin_macros.h) and profiling code to be removed (prof.[ch], spinprof.[ch]). Such code shouldn't be in specific subsystems.

The division functions in lib.h are useless IMHO, both callers seem to use divide by a power of two. And gcc supports shift in 64bit fine in the kernel. Can you remove that please?

statcnt.h: This is completely useless because you don't align the individual fields for cache lines - so you will still have false sharing everywhere. Also using NR_CPUS is nasty because it can be very big - num_possible_cpus() is better. It should use the new dynamic per cpu allocator.

Best you just remove it for now and use atomic_t and readd properly when you do real SMP tuning with measurements.

debug.[ch]: A lot of these functions like "schedulable" are name space space polluting.
reiser4_kmalloc() such wrappers are deprecated. Please remove. xmemset et.al should be replaced with the normal functions everywhere

Best would be probably to remove most of these files for submission.

What is reiser4_internal? Can't you just use static like everybody else?

status_flags.c: Please remove that CONFIG_FRAME_POINTER code. In general i think it would be better if you removed that "private mini crashdumping".

Is there any reason you can't just use wait queues like everybody else instead of these reimplemented condition variables in kcond.[ch]?

In general it would be good if someone experienced not from the reiser team would read the whole source and looks for obvious problems
(I didn't, just mentioning stuff I from a quick look at some support files)

-Andi

#8 lore raw

Hello

Quoted 13 lines · Andi Kleen
Quoted 10 lines · Hans Reiser

* metafiles should be disabled until we can present code that works
right. Half the list thinks we cannot solve the cycles problem ever. Disable metafiles and postpone problem until working code, or the failure to produce it, makes it possible to do more than rant at each other. This is currently already done in the -mm patches, but is mentioned lest someone think it forgotten.

* update the locking documentation

Probably I forget something.

Quoted 9 lines

These are all big picture issues, but I think some low level attention to the individual code is still needed.

Some stuff that stood out from a very quick look:

I would like for the custom spin lock debugging (spin_macros.h) and profiling code to be removed (prof.[ch], spinprof.[ch]). Such code shouldn't be in specific subsystems.

sorry, Andi, I guess you are looking at something old. Reiser4 does not have neither prof.[ch], nor spinprof.[ch] and we removed already some debugging code from spin_macros.h.

Quoted 4 lines

The division functions in lib.h are useless IMHO, both callers seem to use divide by a power of two. And gcc supports shift in 64bit fine in the kernel. Can you remove that please?

ok

Quoted 6 lines

statcnt.h: This is completely useless because you don't align the individual fields for cache lines - so you will still have false sharing everywhere. Also using NR_CPUS is nasty because it can be very big - num_possible_cpus() is better. It should use the new dynamic per cpu allocator.

statcnt.h is already removed.

Quoted 6 lines

Best you just remove it for now and use atomic_t and readd properly when you do real SMP tuning with measurements.

debug.[ch]: A lot of these functions like "schedulable" are name space space polluting.
reiser4_kmalloc() such wrappers are deprecated. Please remove.

ok

Quoted 2 lines

xmemset et.al should be replaced with the normal functions everywhere

done.

Quoted 4 lines

Best would be probably to remove most of these files for submission.

What is reiser4_internal? Can't you just use static like everybody else?

ok

Quoted 3 lines

status_flags.c: Please remove that CONFIG_FRAME_POINTER code. In general i think it would be better if you removed that "private mini crashdumping".

ok

Quoted 4 lines

Is there any reason you can't just use wait queues like everybody else instead of these reimplemented condition variables in kcond.[ch]?

I will investigate that

Quoted 8 lines

In general it would be good if someone experienced not from the reiser team would read the whole source and looks for obvious problems
(I didn't, just mentioning stuff I from a quick look at some support files)

-Andi

author
#9 lore raw
Quoted 2 lines · Vladimir Saveliev

have neither prof.[ch], nor spinprof.[ch] and we removed already some debugging code from spin_macros.h.

Yes, i was looking at some older tree with reiser4. Sorry, just ignore what is already done.

But still spin_macros.h should be completely removed imho. Such custom lock wrappers are strongly discouraged because it makes it hard for others to read your code.

Quoted 7 lines
Quoted 6 lines

statcnt.h: This is completely useless because you don't align the individual fields for cache lines - so you will still have false sharing everywhere. Also using NR_CPUS is nasty because it can be very big - num_possible_cpus() is better. It should use the new dynamic per cpu allocator.

statcnt.h is already removed.

Great.

-Andi

#10 lore raw
Quoted 6 lines · Andi Kleen

Yes, i was looking at some older tree with reiser4. Sorry, just ignore what is already done.

But still spin_macros.h should be completely removed imho. Such
custom lock wrappers are strongly discouraged because it makes it hard for others to read your code.

I may comfirm that this makes Reiser4 very difficult to investigate. Ctags doesn't work with that too..

Signature · 3 lines
Best Regards,
Artem B. Bityuckiy,
St.-Petersburg, Russia.