linux-kernel - Re: [PATCH] x86, UV: Fix RTC latency bug by reading replicated cachelines

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20100122154140.GB4975@sgi.com>
Date:	Fri, 22 Jan 2010 09:41:40 -0600
From:	Dimitri Sivanich <sivanich@....com>
To:	Ingo Molnar <mingo@...e.hu>
Cc:	linux-kernel@...r.kernel.org, Jack Steiner <steiner@....com>,
	"H. Peter Anvin" <hpa@...or.com>, tglx@...utronix.de
Subject: Re: [PATCH] x86, UV: Fix RTC latency bug by reading replicated
	cachelines

For SGI UV node controllers (HUB) rev 2.0 or greater, use replicated
cachelines to read the RTC timer.  This optimization allows faster
simulataneous reads from a given socket.

Signed-off-by: Dimitri Sivanich <sivanich@....com>

---

 arch/x86/kernel/uv_time.c |   13 ++++++++++++-
 drivers/char/uv_mmtimer.c |   18 +++++++++++-------
 2 files changed, 23 insertions(+), 8 deletions(-)

Index: linux/arch/x86/kernel/uv_time.c
===================================================================
--- linux.orig/arch/x86/kernel/uv_time.c	2010-01-22 08:09:37.000000000 -0600
+++ linux/arch/x86/kernel/uv_time.c	2010-01-22 09:14:45.000000000 -0600
@@ -282,10 +282,21 @@ static int uv_rtc_unset_timer(int cpu, i
 
 /*
  * Read the RTC.
+ *
+ * Starting with HUB rev 2.0, the UV RTC register is replicated across all
+ * cachelines of it's own page.  This allows faster simultaneous reads
+ * from a given socket.
  */
 static cycle_t uv_read_rtc(struct clocksource *cs)
 {
-	return (cycle_t)uv_read_local_mmr(UVH_RTC);
+	unsigned long offset;
+
+	if (uv_get_min_hub_revision_id() == 1)
+		offset = 0;
+	else
+		offset = (uv_blade_processor_id() * L1_CACHE_BYTES) % PAGE_SIZE;
+
+	return (cycle_t)uv_read_local_mmr(UVH_RTC | offset);
 }
 
 /*
Index: linux/drivers/char/uv_mmtimer.c
===================================================================
--- linux.orig/drivers/char/uv_mmtimer.c	2010-01-22 08:09:38.000000000 -0600
+++ linux/drivers/char/uv_mmtimer.c	2010-01-22 09:23:34.000000000 -0600
@@ -89,13 +89,17 @@ static long uv_mmtimer_ioctl(struct file
 	switch (cmd) {
 	case MMTIMER_GETOFFSET:	/* offset of the counter */
 		/*
-		 * UV RTC register is on its own page
+		 * Starting with HUB rev 2.0, the UV RTC register is
+		 * replicated across all cachelines of it's own page.
+		 * This allows faster simultaneous reads from a given socket.
+		 *
+		 * The offset returned is in 64 bit units.
 		 */
-		if (PAGE_SIZE <= (1 << 16))
-			ret = ((UV_LOCAL_MMR_BASE | UVH_RTC) & (PAGE_SIZE-1))
-				/ 8;
+		if (uv_get_min_hub_revision_id() == 1)
+			ret = 0;
 		else
-			ret = -ENOSYS;
+			ret = ((uv_blade_processor_id() * L1_CACHE_BYTES) %
+					PAGE_SIZE) / 8;
 		break;
 
 	case MMTIMER_GETRES: /* resolution of the clock in 10^-15 s */
@@ -115,8 +119,8 @@ static long uv_mmtimer_ioctl(struct file
 		ret = hweight64(UVH_RTC_REAL_TIME_CLOCK_MASK);
 		break;
 
-	case MMTIMER_MMAPAVAIL: /* can we mmap the clock into userspace? */
-		ret = (PAGE_SIZE <= (1 << 16)) ? 1 : 0;
+	case MMTIMER_MMAPAVAIL:
+		ret = 1;
 		break;
 
 	case MMTIMER_GETCOUNTER:
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/