LinuxKPI: Add ksize() implementation
ClosedPublic
Actions

Authored by wulf on Aug 27 2020, 8:47 PM.

Details

Reviewers

• hselasky
manu
kib

Group Reviewers

manpages

Commits

rS365383: MFC r364964: LinuxKPI: Implement ksize() function.
rS364964: LinuxKPI: Implement ksize() function.

Summary

ksize() gets the actual amount of memory allocated for a given object and is used once in drm-kmod (drivers/dma-buf/reservation.c) to align C99 flexible array size with UMA zone item size.
This patch is borrowed from kern_malloc.c and allows to strip "#ifdef _ _FreeBSD_ _" from aforementioned drm-kmod's file. Quick test shows no positive effect from better array size alignment at least in my setup.

Test Plan

Test allocation sizes with GENERIC kernel config and with DEBUG_MEMGUARD and DEBUG_REDZONE options enabled.

Diff Detail

Repository

rS FreeBSD src repository - subversion

Lint

Lint Skipped

Unit

Tests Skipped

Event Timeline

wulf created this revision.Aug 27 2020, 8:47 PM

Herald added subscribers: emaste, imp. · View Herald TranscriptAug 27 2020, 8:47 PM

wulf requested review of this revision.Aug 27 2020, 8:47 PM

wulf edited the summary of this revision. (Show Details)Aug 27 2020, 8:57 PM

• hselasky added a reviewer: kib.Aug 27 2020, 8:59 PM

Looks good to me. Let Konstantin have a say too before it goes in.

This revision is now accepted and ready to land.Aug 27 2020, 9:01 PM

Make sure you have the word LinuxKPI in the commit message.

I suppose this was extracted from realloc(9) ?

I dislike the addition of code that knows so much about internals of malloc(9), into linuxkpi. I suggest to move this function to kern_malloc.c as is and name it e.g. malloc_usable_size(), since this is the usermode name for the same functionality. Then linuxkpi is just a renaming macro.

sys/compat/linuxkpi/common/include/linux/slab.h
193	Can we drop all that extern from function declarations in linuxkpi ?

Move main routine in to kern_malloc.c
Add pair of lines to malloc(9) manpage

This revision now requires review to proceed.Aug 28 2020, 7:14 PM

Herald added a reviewer: manpages. · View Herald TranscriptAug 28 2020, 7:14 PM

In D26215#582535, @kib wrote:

I suppose this was extracted from realloc(9) ?

Yes. it was extracted from realloc(9) and free(9)

• hselasky accepted this revision.Aug 28 2020, 7:18 PM

This revision is now accepted and ready to land.Aug 28 2020, 7:18 PM

kib accepted this revision.Aug 28 2020, 8:14 PM

kib added inline comments.

sys/sys/malloc.h
251	Move this line right after malloc_uninit() to keep declarations sorted by name.
sys/vm/memguard.h
49	Similarly move to after memguard_free, but this is too unordered already.