Date: Sun, 26 Jan 2003 17:11:49 -0500 From: Jake Burkholder <jake@locore.ca> To: David Xu <davidxu@FreeBSD.org> Cc: cvs-committers@FreeBSD.org, cvs-all@FreeBSD.org Subject: Re: cvs commit: src/sys/alpha/alpha trap.c vm_machdep.c src/sys/ddb db_ps.c src/sys/i386/i386 critical.c exception.s genassym.c trap.c vm_machdep.c src/sys/ia64/ia64 trap.c vm_machdep.c src/sys/kern init_main.c kern_clock.c kern_exec.c kern_exit.c kern_fork.c ... Message-ID: <20030126171149.B79685@locore.ca> In-Reply-To: <200301261141.h0QBfZxZ001813@repoman.freebsd.org>; from davidxu@FreeBSD.org on Sun, Jan 26, 2003 at 03:41:35AM -0800 References: <200301261141.h0QBfZxZ001813@repoman.freebsd.org>
next in thread | previous in thread | raw e-mail | index | archive | help
Apparently, On Sun, Jan 26, 2003 at 03:41:35AM -0800, David Xu said words to the effect of; > davidxu 2003/01/26 03:41:35 PST > > Modified files: > sys/alpha/alpha trap.c vm_machdep.c > sys/ddb db_ps.c > sys/i386/i386 critical.c exception.s genassym.c trap.c > vm_machdep.c > sys/ia64/ia64 trap.c vm_machdep.c > sys/kern init_main.c kern_clock.c kern_exec.c > kern_exit.c kern_fork.c kern_lock.c > kern_resource.c kern_sig.c kern_switch.c > kern_thread.c subr_prof.c subr_trap.c > subr_witness.c > sys/powerpc/powerpc vm_machdep.c > sys/sparc64/sparc64 trap.c vm_machdep.c > sys/sys buf.h lockmgr.h proc.h resourcevar.h > systm.h > Log: > Move UPCALL related data structure out of kse, introduce a new > data structure called kse_upcall to manage UPCALL. All KSE binding > and loaning code are gone. > Ok, this change should have had much more significant review and testing. A lot of things were broken by this commit, and a lot of changes were made which are not mentioned in the commit message, should have been committed separately, and should have had more review than just julian. The part that I care about is the profiling changes. The TDF_ASTPENDING flag is not tested on anyhting but x86, which breaks userland profiling on all other architectures. I don't see how this change can have had significant testing given that you don't have sufficient infrastructure to test profiling of non-trivial threaded programs. In addition, this blocks getting addupc_intr out of sched_lock for non-threaded processes. Please back this part out and re-implement it in a way that does not break profiling for other architectures, and that does not require sched_lock to be held accross addupc_intr for threaded processes. The only way I can see to do that is with per-thread profiling buffers. Jake To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe cvs-all" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20030126171149.B79685>