From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S932112AbVHOHfz (ORCPT ); Mon, 15 Aug 2005 03:35:55 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S932140AbVHOHfz (ORCPT ); Mon, 15 Aug 2005 03:35:55 -0400 Received: from zproxy.gmail.com ([64.233.162.197]:23570 "EHLO zproxy.gmail.com") by vger.kernel.org with ESMTP id S932112AbVHOHfz convert rfc822-to-8bit (ORCPT ); Mon, 15 Aug 2005 03:35:55 -0400 DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:to:subject:cc:in-reply-to:mime-version:content-type:content-transfer-encoding:content-disposition:references; b=Ehlvwbez+2QWFPeKDQCqNQQJeFWER6KII7lOVREsCXtFKm2CLGiTkkHjW9PKFVeqrKkAjOH22qU8NHBsYF8xLnBfCZyHB7mUXuoUAAvzTwSLELHdY2uKuPiVlSPGBCpUaL8lccZqJAtw+9ccpwMYAhvpqa5/3ddaWbrY4bPDTCA= Message-ID: <398d69300508150035246230af@mail.gmail.com> Date: Mon, 15 Aug 2005 15:35:49 +0800 From: Xuekun Hu To: linux-kernel@vger.kernel.org Subject: Re: lockmeter: fix lock counter roll over issue Cc: raybry@sgi.com, akpm@osdl.org In-Reply-To: <398d69300508132100327e3712@mail.gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7BIT Content-Disposition: inline References: <398d69300508132100327e3712@mail.gmail.com> Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Does anyone have inputs? On 8/14/05, Xuekun Hu wrote: > When I collected lockmeter data for a longer duration, sometimes the > locks counter could roll over. I'm sure someone else maybe meet the > same situation. So I wrote the below patch, could you have a look? > > diff -Nraup linux/include/linux/lockmeter.h > linux.lockmeter/include/linux/lockmeter.h > --- linux/include/linux/lockmeter.h 2005-08-09 10:46:53.302083832 +0800 > +++ linux.lockmeter/include/linux/lockmeter.h 2005-08-05 > 17:22:37.000000000 +0800 > @@ -150,7 +150,7 @@ typedef struct { > uint32_t max_wait_ww_ticks; /* max wait time writer vs writer */ > /* prev 2 only used for write locks*/ > uint32_t acquire_time; /* time lock acquired this CPU */ > - uint32_t count[LSTAT_ACT_MAX_VALUES]; > + uint64_t count[LSTAT_ACT_MAX_VALUES]; > } lstat_lock_counts_t; > > typedef lstat_lock_counts_t lstat_cpu_counts_t[LSTAT_MAX_STAT_INDEX]; > @@ -167,7 +167,7 @@ typedef lstat_lock_counts_t lstat_cpu_co > #define LSTAT_MAX_READ_LOCK_INDEX 1000 > typedef struct { > POINTER lock_ptr; /* address of lock for output stats */ > - uint32_t read_lock_count; > + uint64_t read_lock_count; > int64_t cum_hold_ticks; /* sum of read lock hold times over */ > /* all callers. ....................*/ > uint32_t write_index; /* last write lock hash table index */ > > I also modified lockstat to compliant with this modification against > lockstat 1.4.11 version. > > diff -Nraup lockstat/include/linux/lockmeter.h > lockstat.fix/include/linux/lockmeter.h > --- lockstat/include/linux/lockmeter.h 2004-09-14 07:02:05.000000000 +0800 > +++ lockstat.fix/include/linux/lockmeter.h 2005-08-05 17:38:46.000000000 +0800 > @@ -150,7 +150,7 @@ typedef struct { > uint32_t max_wait_ww_ticks; /* max wait time writer vs writer */ > /* prev 2 only used for write locks*/ > uint32_t acquire_time; /* time lock acquired this CPU */ > - uint32_t count[LSTAT_ACT_MAX_VALUES]; > + uint64_t count[LSTAT_ACT_MAX_VALUES]; > } lstat_lock_counts_t; > > typedef lstat_lock_counts_t lstat_cpu_counts_t[LSTAT_MAX_STAT_INDEX]; > @@ -167,7 +167,7 @@ typedef lstat_lock_counts_t lstat_cpu_co > #define LSTAT_MAX_READ_LOCK_INDEX 1000 > typedef struct { > POINTER lock_ptr; /* address of lock for output stats */ > - uint32_t read_lock_count; > + uint64_t read_lock_count; > int64_t cum_hold_ticks; /* sum of read lock hold times over */ > /* all callers. ....................*/ > uint32_t write_index; /* last write lock hash table index */ > diff -Nraup lockstat/lockstat.c lockstat.fix/lockstat.c > --- lockstat/lockstat.c 2004-09-14 07:01:41.000000000 +0800 > +++ lockstat.fix/lockstat.c 2005-08-09 10:35:18.305356776 +0800 > @@ -36,7 +36,7 @@ static char lockstat_version[] = "1.4.1 > #include > #include > #include > -#include > +#include > #include > > extern void closeFiles(void); > @@ -267,7 +267,7 @@ typedef enum {Null_Entry, Spin_Entry, RL > > typedef struct { > lstat_lock_counts_t counts; > - uint32_t total; > + uint64_t total; > double contention; > double persec; > double utilization; > @@ -999,7 +999,8 @@ add_counts(lock_summary_t *sump, lstat_l > int > sum_counts(lock_summary_t *sump, entry_type_enum entry_type) > { > - int total, i, j; > + int i, j; > + uint64_t total; > double contention; > double utilization; > > @@ -1171,7 +1172,7 @@ print_lock(char *name, lock_summary_t *s > float mean_hold_time, max_hold_time; > float wait_time, mean_wait_time, max_wait_time, mean_wait_ww_time, > max_wait_ww_time; > double temp_time; > - uint32_t total_spin_count = 0; > + uint64_t total_spin_count = 0; > > print_header(); > > @@ -1317,7 +1318,7 @@ print_lock(char *name, lock_summary_t *s > } > > /* following done for ALL types of locks, except as indicated */ > - printf(" %9d", sump->total); > + printf(" %20ld", sump->total); > print_percentage(' > ',100.0*(double)sump->counts.count[LSTAT_ACT_NO_WAIT]/(double)sump->total,0); > print_percentage(' > ',100.0*(double)sump->counts.count[LSTAT_ACT_SPIN]/(double)sump->total,0); > if (entry_type == WLspin_Entry) { > @@ -1325,7 +1326,7 @@ print_lock(char *name, lock_summary_t *s > ')'); > } > #ifdef notyet > - printf("%9d", sump->counts.count[LSTAT_ACT_SLEPT]); > + printf("%20ld", sump->counts.count[LSTAT_ACT_SLEPT]); > #endif > if (entry_type == Spin_Entry) { > /* We only see these for spinlock_t, and even then it's rare. */ > > Any comments? >