enhanced spinlock debugging for intel - take 2
Hi -
I've revised and improved the intel spinlock debugging I posted two weeks ago.
Attached is the diff (against 2.4.8). It moves the debugging routines into a seperate C file (this is how sparc does it), fixes the race condition noted by Tachino Nobuhiro, and adds debugging routines for read-write spinlocks.
It compiles and runs fine on my single processor machine, and continues to detect problems with the USB SCSI code. Anybody want to try it on an SMP box and make sure it's OK?
Make sure you do a "make dep" after applying the patch, since it exports symbols...
Signature · 10 lines
-bwb
Brent Baccala
[email protected]
==============================================================================
For news from freesoft.org, subscribe to [email protected]:
mailto:[email protected]?subject=subscribe&body=subscribe
==============================================================================arch/i386/lib/Makefile +3 -1
@@ -7,9 +7,11 @@ L_TARGET = lib.a +export-objs = debuglocks.o + obj-y = checksum.o old-checksum.o delay.o \ usercopy.o getuser.o \ - memcpy.o strstr.o + memcpy.o strstr.o debuglocks.o obj-$(CONFIG_X86_USE_3DNOW) += mmx.o obj-$(CONFIG_HAVE_DEC_LOCK) += dec_and_lock.o
arch/i386/lib/debuglocks.c +183 -0
183 lines not on this page · raw message
include/asm-i386/spinlock.h +44 -38
157 lines not on this page · raw message
Quoted 4 lines · Brent Baccala
I've revised and improved the intel spinlock debugging I posted two weeks ago.
Make sure you do a "make dep" after applying the patch, since it exports symbols...
make mrproper; make dep is the only way to guarantee that module symbol versions are correct after a patch.