From: James Bottomley <James.Bottomley@steeleye.com>
To: Andrew Morton <akpm@osdl.org>, johnstultz@us.ibm.com
Cc: Linux Kernel <linux-kernel@vger.kernel.org>
Subject: [PATCH] fix get_jiffies_64 to work on voyager
Date: 06 Jan 2004 10:04:07 -0600 [thread overview]
Message-ID: <1073405053.2047.28.camel@mulgrave> (raw)
This patch
ChangeSet@1.1534.5.2, 2003-12-30 15:40:23-08:00, akpm@osdl.org
[PATCH] ia32 jiffy wrapping fixes
Causes the voyager boot to hang. The problem is this change:
--- a/arch/i386/kernel/timers/timer_tsc.c Tue Jan 6 09:57:34 2004
+++ b/arch/i386/kernel/timers/timer_tsc.c Tue Jan 6 09:57:34 2004
@@ -141,7 +140,7 @@
#ifndef CONFIG_NUMA
if (!use_tsc)
#endif
- return (unsigned long long)jiffies * (1000000000 / HZ);
+ return (unsigned long long)get_jiffies_64() *
(1000000000 / HZ);
Apart from the fact (that I've whined about before) that this
sched_clock() function should be one of the timer function pointers, so
there isn't this CONFIG_NUMA dependence (unless I can also add a
CONFIG_X86_VOYAGER dependence to it as well), the problem seems to be
some type of bad resonance between the jiffies_64 update and the
xtime_lock in get_jiffies_64(). I think this may indicate that HZ needs
to be reduced to 100 on voyager; however, there is also no need to get
the xtime sequence lock every time we do a jiffies_64 read, since the
only unstable time is when we may be updating both halves of it
non-atomically. Thus, we only need the sequence lock when the bottom
half is zero. This should improve the fast path of get_jiffies_64() for
all x86 arch's.
James
===== kernel/time.c 1.18 vs edited =====
--- 1.18/kernel/time.c Wed Oct 22 00:09:54 2003
+++ edited/kernel/time.c Tue Jan 6 09:20:38 2004
@@ -422,13 +422,20 @@
#if (BITS_PER_LONG < 64)
u64 get_jiffies_64(void)
{
- unsigned long seq;
- u64 ret;
+ u64 ret = jiffies_64;
- do {
- seq = read_seqbegin(&xtime_lock);
+ /* We only have read problems when the lower 32 bits are zero
+ * indicating that we may be in the process of updating the upper
+ * 32 bits */
+ while (unlikely((jiffies_64 & 0xffffffffULL) == 0)) {
+ unsigned long seq = read_seqbegin(&xtime_lock);
+
+ rmb();
ret = jiffies_64;
- } while (read_seqretry(&xtime_lock, seq));
+ rmb();
+ if(!read_seqretry(&xtime_lock, seq))
+ break;
+ }
return ret;
}
next reply other threads:[~2004-01-06 16:04 UTC|newest]
Thread overview: 16+ messages / expand[flat|nested] mbox.gz Atom feed top
2004-01-06 16:04 James Bottomley [this message]
2004-01-06 16:19 ` [PATCH] fix get_jiffies_64 to work on voyager Andrew Morton
2004-01-06 16:26 ` James Bottomley
2004-01-06 17:05 ` Andrew Morton
2004-01-06 18:53 ` James Bottomley
2004-01-06 16:29 ` Linus Torvalds
2004-01-06 17:05 ` Andrew Morton
2004-01-06 17:11 ` Tim Schmielau
2004-01-06 17:22 ` Andrew Morton
2004-01-06 17:37 ` Tim Schmielau
2004-01-06 18:02 ` Linus Torvalds
2004-01-06 16:30 ` Tim Schmielau
2004-01-06 16:26 ` Linus Torvalds
2004-01-06 18:29 ` Paulo Marques
2004-01-06 18:46 ` Linus Torvalds
2004-01-06 19:04 ` Paulo Marques
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1073405053.2047.28.camel@mulgrave \
--to=james.bottomley@steeleye.com \
--cc=akpm@osdl.org \
--cc=johnstultz@us.ibm.com \
--cc=linux-kernel@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox