From: lizf@kernel.org
To: stable@vger.kernel.org
Cc: linux-kernel@vger.kernel.org, Andrew Hunter <ahh@google.com>,
Thomas Gleixner <tglx@linutronix.de>,
Ingo Molnar <mingo@redhat.com>, Paul Turner <pjt@google.com>,
Richard Cochran <richardcochran@gmail.com>,
Prarit Bhargava <prarit@redhat.com>,
John Stultz <john.stultz@linaro.org>,
Zefan Li <lizefan@huawei.com>
Subject: [PATCH 3.4 41/91] jiffies: Fix timeval conversion to jiffies
Date: Thu, 27 Nov 2014 16:42:24 +0800 [thread overview]
Message-ID: <1417077794-9299-41-git-send-email-lizf@kernel.org> (raw)
In-Reply-To: <1417077368-9217-1-git-send-email-lizf@kernel.org>
From: Andrew Hunter <ahh@google.com>
3.4.105-rc1 review patch. If anyone has any objections, please let me know.
------------------
commit d78c9300c51d6ceed9f6d078d4e9366f259de28c upstream.
timeval_to_jiffies tried to round a timeval up to an integral number
of jiffies, but the logic for doing so was incorrect: intervals
corresponding to exactly N jiffies would become N+1. This manifested
itself particularly repeatedly stopping/starting an itimer:
setitimer(ITIMER_PROF, &val, NULL);
setitimer(ITIMER_PROF, NULL, &val);
would add a full tick to val, _even if it was exactly representable in
terms of jiffies_ (say, the result of a previous rounding.) Doing
this repeatedly would cause unbounded growth in val. So fix the math.
Here's what was wrong with the conversion: we essentially computed
(eliding seconds)
jiffies = usec * (NSEC_PER_USEC/TICK_NSEC)
by using scaling arithmetic, which took the best approximation of
NSEC_PER_USEC/TICK_NSEC with denominator of 2^USEC_JIFFIE_SC =
x/(2^USEC_JIFFIE_SC), and computed:
jiffies = (usec * x) >> USEC_JIFFIE_SC
and rounded this calculation up in the intermediate form (since we
can't necessarily exactly represent TICK_NSEC in usec.) But the
scaling arithmetic is a (very slight) *over*approximation of the true
value; that is, instead of dividing by (1 usec/ 1 jiffie), we
effectively divided by (1 usec/1 jiffie)-epsilon (rounding
down). This would normally be fine, but we want to round timeouts up,
and we did so by adding 2^USEC_JIFFIE_SC - 1 before the shift; this
would be fine if our division was exact, but dividing this by the
slightly smaller factor was equivalent to adding just _over_ 1 to the
final result (instead of just _under_ 1, as desired.)
In particular, with HZ=1000, we consistently computed that 10000 usec
was 11 jiffies; the same was true for any exact multiple of
TICK_NSEC.
We could possibly still round in the intermediate form, adding
something less than 2^USEC_JIFFIE_SC - 1, but easier still is to
convert usec->nsec, round in nanoseconds, and then convert using
time*spec*_to_jiffies. This adds one constant multiplication, and is
not observably slower in microbenchmarks on recent x86 hardware.
Tested: the following program:
int main() {
struct itimerval zero = {{0, 0}, {0, 0}};
/* Initially set to 10 ms. */
struct itimerval initial = zero;
initial.it_interval.tv_usec = 10000;
setitimer(ITIMER_PROF, &initial, NULL);
/* Save and restore several times. */
for (size_t i = 0; i < 10; ++i) {
struct itimerval prev;
setitimer(ITIMER_PROF, &zero, &prev);
/* on old kernels, this goes up by TICK_USEC every iteration */
printf("previous value: %ld %ld %ld %ld\n",
prev.it_interval.tv_sec, prev.it_interval.tv_usec,
prev.it_value.tv_sec, prev.it_value.tv_usec);
setitimer(ITIMER_PROF, &prev, NULL);
}
return 0;
}
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Paul Turner <pjt@google.com>
Cc: Richard Cochran <richardcochran@gmail.com>
Cc: Prarit Bhargava <prarit@redhat.com>
Reviewed-by: Paul Turner <pjt@google.com>
Reported-by: Aaron Jacobs <jacobsa@google.com>
Signed-off-by: Andrew Hunter <ahh@google.com>
[jstultz: Tweaked to apply to 3.17-rc]
Signed-off-by: John Stultz <john.stultz@linaro.org>
[lizf: Backported to 3.4: adjust filename]
Signed-off-by: Zefan Li <lizefan@huawei.com>
---
include/linux/jiffies.h | 12 -----------
kernel/time.c | 54 +++++++++++++++++++++++++++----------------------
2 files changed, 30 insertions(+), 36 deletions(-)
diff --git a/include/linux/jiffies.h b/include/linux/jiffies.h
index f5df3dc..f4e8578 100644
--- a/include/linux/jiffies.h
+++ b/include/linux/jiffies.h
@@ -259,23 +259,11 @@ extern unsigned long preset_lpj;
#define SEC_JIFFIE_SC (32 - SHIFT_HZ)
#endif
#define NSEC_JIFFIE_SC (SEC_JIFFIE_SC + 29)
-#define USEC_JIFFIE_SC (SEC_JIFFIE_SC + 19)
#define SEC_CONVERSION ((unsigned long)((((u64)NSEC_PER_SEC << SEC_JIFFIE_SC) +\
TICK_NSEC -1) / (u64)TICK_NSEC))
#define NSEC_CONVERSION ((unsigned long)((((u64)1 << NSEC_JIFFIE_SC) +\
TICK_NSEC -1) / (u64)TICK_NSEC))
-#define USEC_CONVERSION \
- ((unsigned long)((((u64)NSEC_PER_USEC << USEC_JIFFIE_SC) +\
- TICK_NSEC -1) / (u64)TICK_NSEC))
-/*
- * USEC_ROUND is used in the timeval to jiffie conversion. See there
- * for more details. It is the scaled resolution rounding value. Note
- * that it is a 64-bit value. Since, when it is applied, we are already
- * in jiffies (albit scaled), it is nothing but the bits we will shift
- * off.
- */
-#define USEC_ROUND (u64)(((u64)1 << USEC_JIFFIE_SC) - 1)
/*
* The maximum jiffie value is (MAX_INT >> 1). Here we translate that
* into seconds. The 64-bit case will overflow if we are not careful,
diff --git a/kernel/time.c b/kernel/time.c
index ba744cf..a095290 100644
--- a/kernel/time.c
+++ b/kernel/time.c
@@ -487,17 +487,20 @@ EXPORT_SYMBOL(usecs_to_jiffies);
* that a remainder subtract here would not do the right thing as the
* resolution values don't fall on second boundries. I.e. the line:
* nsec -= nsec % TICK_NSEC; is NOT a correct resolution rounding.
+ * Note that due to the small error in the multiplier here, this
+ * rounding is incorrect for sufficiently large values of tv_nsec, but
+ * well formed timespecs should have tv_nsec < NSEC_PER_SEC, so we're
+ * OK.
*
* Rather, we just shift the bits off the right.
*
* The >> (NSEC_JIFFIE_SC - SEC_JIFFIE_SC) converts the scaled nsec
* value to a scaled second value.
*/
-unsigned long
-timespec_to_jiffies(const struct timespec *value)
+static unsigned long
+__timespec_to_jiffies(unsigned long sec, long nsec)
{
- unsigned long sec = value->tv_sec;
- long nsec = value->tv_nsec + TICK_NSEC - 1;
+ nsec = nsec + TICK_NSEC - 1;
if (sec >= MAX_SEC_IN_JIFFIES){
sec = MAX_SEC_IN_JIFFIES;
@@ -508,6 +511,13 @@ timespec_to_jiffies(const struct timespec *value)
(NSEC_JIFFIE_SC - SEC_JIFFIE_SC))) >> SEC_JIFFIE_SC;
}
+
+unsigned long
+timespec_to_jiffies(const struct timespec *value)
+{
+ return __timespec_to_jiffies(value->tv_sec, value->tv_nsec);
+}
+
EXPORT_SYMBOL(timespec_to_jiffies);
void
@@ -524,31 +534,27 @@ jiffies_to_timespec(const unsigned long jiffies, struct timespec *value)
}
EXPORT_SYMBOL(jiffies_to_timespec);
-/* Same for "timeval"
+/*
+ * We could use a similar algorithm to timespec_to_jiffies (with a
+ * different multiplier for usec instead of nsec). But this has a
+ * problem with rounding: we can't exactly add TICK_NSEC - 1 to the
+ * usec value, since it's not necessarily integral.
*
- * Well, almost. The problem here is that the real system resolution is
- * in nanoseconds and the value being converted is in micro seconds.
- * Also for some machines (those that use HZ = 1024, in-particular),
- * there is a LARGE error in the tick size in microseconds.
-
- * The solution we use is to do the rounding AFTER we convert the
- * microsecond part. Thus the USEC_ROUND, the bits to be shifted off.
- * Instruction wise, this should cost only an additional add with carry
- * instruction above the way it was done above.
+ * We could instead round in the intermediate scaled representation
+ * (i.e. in units of 1/2^(large scale) jiffies) but that's also
+ * perilous: the scaling introduces a small positive error, which
+ * combined with a division-rounding-upward (i.e. adding 2^(scale) - 1
+ * units to the intermediate before shifting) leads to accidental
+ * overflow and overestimates.
+ *
+ * At the cost of one additional multiplication by a constant, just
+ * use the timespec implementation.
*/
unsigned long
timeval_to_jiffies(const struct timeval *value)
{
- unsigned long sec = value->tv_sec;
- long usec = value->tv_usec;
-
- if (sec >= MAX_SEC_IN_JIFFIES){
- sec = MAX_SEC_IN_JIFFIES;
- usec = 0;
- }
- return (((u64)sec * SEC_CONVERSION) +
- (((u64)usec * USEC_CONVERSION + USEC_ROUND) >>
- (USEC_JIFFIE_SC - SEC_JIFFIE_SC))) >> SEC_JIFFIE_SC;
+ return __timespec_to_jiffies(value->tv_sec,
+ value->tv_usec * NSEC_PER_USEC);
}
EXPORT_SYMBOL(timeval_to_jiffies);
--
1.9.1
next prev parent reply other threads:[~2014-11-27 8:47 UTC|newest]
Thread overview: 98+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-11-27 8:36 [PATCH 3.4 00/91] 3.4.105-rc1 review lizf
2014-11-27 8:41 ` [PATCH 3.4 01/91] KVM: s390: Fix user triggerable bug in dead code lizf
2014-11-27 8:41 ` [PATCH 3.4 02/91] regmap: Fix handling of volatile registers for format_write() chips lizf
2014-11-27 8:41 ` [PATCH 3.4 03/91] drm/i915: Remove bogus __init annotation from DMI callbacks lizf
2014-11-27 8:41 ` [PATCH 3.4 04/91] get rid of propagate_umount() mistakenly treating slaves as busy lizf
2014-11-27 8:41 ` [PATCH 3.4 05/91] drm/vmwgfx: Fix a potential infinite spin waiting for fifo idle lizf
2014-11-27 8:41 ` [PATCH 3.4 06/91] ALSA: hda - Fix COEF setups for ALC1150 codec lizf
2014-11-27 8:41 ` [PATCH 3.4 07/91] ACPI / cpuidle: fix deadlock between cpuidle_lock and cpu_hotplug.lock lizf
2014-11-27 8:41 ` [PATCH 3.4 08/91] regulatory: add NUL to alpha2 lizf
2014-11-27 8:41 ` [PATCH 3.4 09/91] percpu: fix pcpu_alloc_pages() failure path lizf
2014-11-27 8:41 ` [PATCH 3.4 10/91] percpu: perform tlb flush after pcpu_map_pages() failure lizf
2014-11-27 8:41 ` [PATCH 3.4 11/91] percpu: free percpu allocation info for uniprocessor system lizf
2014-11-27 8:41 ` [PATCH 3.4 12/91] cgroup: reject cgroup names with ' ' lizf
2014-11-27 8:41 ` [PATCH 3.4 13/91] rtlwifi: rtl8192cu: Add new ID lizf
2014-11-27 8:41 ` [PATCH 3.4 14/91] ahci: Add Device IDs for Intel 9 Series PCH lizf
2014-11-27 8:41 ` [PATCH 3.4 15/91] ata_piix: " lizf
2014-11-27 8:41 ` [PATCH 3.4 16/91] USB: ftdi_sio: add support for NOVITUS Bono E thermal printer lizf
2014-11-27 8:42 ` [PATCH 3.4 17/91] USB: sierra: avoid CDC class functions on "68A3" devices lizf
2014-11-27 8:42 ` [PATCH 3.4 18/91] USB: sierra: add 1199:68AA device ID lizf
2014-11-27 8:42 ` [PATCH 3.4 19/91] xen/manage: Always freeze/thaw processes when suspend/resuming lizf
2014-11-27 8:42 ` [PATCH 3.4 20/91] block: Fix dev_t minor allocation lifetime lizf
2014-11-27 8:42 ` [PATCH 3.4 21/91] usb: dwc3: core: fix order of PM runtime calls lizf
2014-11-27 8:42 ` [PATCH 3.4 22/91] ahci: add pcid for Marvel 0x9182 controller lizf
2014-11-27 8:42 ` [PATCH 3.4 23/91] drm/radeon: add connector quirk for fujitsu board lizf
2014-11-27 8:42 ` [PATCH 3.4 24/91] usb: host: xhci: fix compliance mode workaround lizf
2014-11-27 8:42 ` [PATCH 3.4 25/91] Input: elantech - fix detection of touchpad on ASUS s301l lizf
2014-11-27 8:42 ` [PATCH 3.4 26/91] USB: ftdi_sio: Add support for GE Healthcare Nemo Tracker device lizf
2014-11-27 8:42 ` [PATCH 3.4 27/91] uwb: init beacon cache entry before registering uwb device lizf
2014-11-27 8:42 ` [PATCH 3.4 28/91] Input: synaptics - add support for ForcePads lizf
2014-11-27 8:42 ` [PATCH 3.4 29/91] libceph: gracefully handle large reply messages from the mon lizf
2014-11-27 8:42 ` [PATCH 3.4 30/91] libceph: add process_one_ticket() helper lizf
2014-11-27 8:42 ` [PATCH 3.4 31/91] libceph: do not hard code max auth ticket len lizf
2014-11-27 8:42 ` [PATCH 3.4 32/91] Input: serport - add compat handling for SPIOCSTYPE ioctl lizf
2014-11-27 8:42 ` [PATCH 3.4 33/91] usb: hub: take hub->hdev reference when processing from eventlist lizf
2014-11-27 8:42 ` [PATCH 3.4 34/91] storage: Add single-LUN quirk for Jaz USB Adapter lizf
2014-11-27 8:42 ` [PATCH 3.4 35/91] xhci: Fix null pointer dereference if xhci initialization fails lizf
2014-11-27 8:42 ` [PATCH 3.4 36/91] futex: Unlock hb->lock in futex_wait_requeue_pi() error path lizf
2014-11-27 8:42 ` [PATCH 3.4 37/91] alarmtimer: Return relative times in timer_gettime lizf
2014-11-27 8:42 ` [PATCH 3.4 38/91] alarmtimer: Do not signal SIGEV_NONE timers lizf
2014-11-27 8:42 ` [PATCH 3.4 39/91] alarmtimer: Lock k_itimer during timer callback lizf
2014-11-27 8:42 ` [PATCH 3.4 40/91] don't bugger nd->seq on set_root_rcu() from follow_dotdot_rcu() lizf
2014-11-27 8:42 ` lizf [this message]
2014-11-27 8:42 ` [PATCH 3.4 42/91] MIPS: ZBOOT: add missing <linux/string.h> include lizf
2014-11-27 8:42 ` [PATCH 3.4 43/91] perf: Fix a race condition in perf_remove_from_context() lizf
2014-12-01 18:43 ` Sukadev Bhattiprolu
2014-12-02 1:22 ` Zefan Li
2014-11-27 8:42 ` [PATCH 3.4 44/91] ASoC: samsung-i2s: Check secondary DAI exists before referencing lizf
2014-11-27 8:42 ` [PATCH 3.4 45/91] Input: i8042 - add Fujitsu U574 to no_timeout dmi table lizf
2014-11-27 8:42 ` [PATCH 3.4 46/91] Input: i8042 - add nomux quirk for Avatar AVIU-145A6 lizf
2014-11-27 8:42 ` [PATCH 3.4 47/91] iscsi-target: Fix memory corruption in iscsit_logout_post_handler_diffcid lizf
2014-11-27 8:42 ` [PATCH 3.4 48/91] iscsi-target: avoid NULL pointer in iscsi_copy_param_list failure lizf
2014-11-27 8:42 ` [PATCH 3.4 49/91] NFSv4: Fix another bug in the close/open_downgrade code lizf
2014-11-27 8:42 ` [PATCH 3.4 50/91] libiscsi: fix potential buffer overrun in __iscsi_conn_send_pdu lizf
2014-11-27 8:42 ` [PATCH 3.4 51/91] USB: storage: Add quirk for Adaptec USBConnect 2000 USB-to-SCSI Adapter lizf
2014-11-27 8:42 ` [PATCH 3.4 52/91] USB: storage: Add quirk for Ariston Technologies iConnect USB to SCSI adapter lizf
2014-11-27 8:42 ` [PATCH 3.4 53/91] USB: storage: Add quirks for Entrega/Xircom USB to SCSI converters lizf
2014-11-27 8:42 ` [PATCH 3.4 54/91] can: flexcan: mark TX mailbox as TX_INACTIVE lizf
2014-11-27 8:42 ` [PATCH 3.4 55/91] can: flexcan: correctly initialize mailboxes lizf
2014-11-27 8:42 ` [PATCH 3.4 56/91] can: flexcan: implement workaround for errata ERR005829 lizf
2014-11-27 8:42 ` [PATCH 3.4 57/91] can: flexcan: put TX mailbox into TX_INACTIVE mode after tx-complete lizf
2014-11-27 8:42 ` [PATCH 3.4 58/91] can: at91_can: add missing prepare and unprepare of the clock lizf
2014-11-27 8:42 ` [PATCH 3.4 59/91] ALSA: pcm: fix fifo_size frame calculation lizf
2014-11-27 8:42 ` [PATCH 3.4 60/91] Fix nasty 32-bit overflow bug in buffer i/o code lizf
2014-11-27 8:42 ` [PATCH 3.4 61/91] parisc: Only use -mfast-indirect-calls option for 32-bit kernel builds lizf
2014-11-27 8:42 ` [PATCH 3.4 62/91] sched: Fix unreleased llc_shared_mask bit during CPU hotplug lizf
2014-11-27 8:42 ` [PATCH 3.4 63/91] sched: add macros to define bitops for task atomic flags lizf
2014-11-27 8:42 ` [PATCH 3.4 64/91] cpuset: PF_SPREAD_PAGE and PF_SPREAD_SLAB should be " lizf
2014-11-27 8:42 ` [PATCH 3.4 65/91] MIPS: mcount: Adjust stack pointer for static trace in MIPS32 lizf
2014-11-27 8:42 ` [PATCH 3.4 66/91] nilfs2: fix data loss with mmap() lizf
2014-11-27 8:42 ` [PATCH 3.4 67/91] ocfs2/dlm: do not get resource spinlock if lockres is new lizf
2014-11-27 8:42 ` [PATCH 3.4 68/91] shmem: fix nlink for rename overwrite directory lizf
2014-11-27 8:42 ` [PATCH 3.4 69/91] ARM: 8165/1: alignment: don't break misaligned NEON load/store lizf
2014-11-27 8:42 ` [PATCH 3.4 70/91] ASoC: core: fix possible ZERO_SIZE_PTR pointer dereferencing error lizf
2014-11-27 8:42 ` [PATCH 3.4 71/91] mm: migrate: Close race between migration completion and mprotect lizf
2014-11-27 8:42 ` [PATCH 3.4 72/91] perf: fix perf bug in fork() lizf
2014-11-27 8:42 ` [PATCH 3.4 73/91] init/Kconfig: Hide printk log config if CONFIG_PRINTK=n lizf
2014-11-27 8:42 ` [PATCH 3.4 74/91] genhd: fix leftover might_sleep() in blk_free_devt() lizf
2014-11-27 8:42 ` [PATCH 3.4 75/91] nl80211: clear skb cb before passing to netlink lizf
2014-11-27 8:42 ` [PATCH 3.4 76/91] ext4: propagate errors up to ext4_find_entry()'s callers lizf
2014-11-27 8:43 ` [PATCH 3.4 77/91] ext4: avoid trying to kfree an ERR_PTR pointer lizf
2014-11-27 8:43 ` [PATCH 3.4 78/91] NFS: fix stable regression lizf
2014-11-27 8:43 ` [PATCH 3.4 79/91] perf: Handle compat ioctl lizf
2014-11-27 8:43 ` [PATCH 3.4 80/91] bluetooth: hci_ldisc: fix deadlock condition lizf
2014-11-27 8:43 ` [PATCH 3.4 81/91] mnt: Only change user settable mount flags in remount lizf
2014-11-27 8:43 ` [PATCH 3.4 82/91] dm crypt: fix access beyond the end of allocated space lizf
2014-11-27 8:43 ` [PATCH 3.4 83/91] Fix spurious request sense in error handling lizf
2014-11-27 8:43 ` [PATCH 3.4 84/91] ipv4: move route garbage collector to work queue lizf
2014-11-27 8:43 ` [PATCH 3.4 85/91] ipv4: avoid parallel route cache gc executions lizf
2014-11-27 8:43 ` [PATCH 3.4 86/91] ipv4: disable bh while doing route gc lizf
2014-11-27 8:43 ` [PATCH 3.4 87/91] rtl8192ce: Fix null dereference in watchdog lizf
2014-11-27 8:43 ` [PATCH 3.4 88/91] ipv6: reuse ip6_frag_id from ip6_ufo_append_data lizf
2014-11-27 8:43 ` [PATCH 3.4 89/91] net: Do not enable tx-nocache-copy by default lizf
2014-11-27 8:43 ` [PATCH 3.4 90/91] ixgbevf: Prevent RX/TX statistics getting reset to zero lizf
2014-11-27 8:43 ` [PATCH 3.4 91/91] l2tp: fix race while getting PMTU on PPP pseudo-wire lizf
2014-11-27 9:07 ` [PATCH 3.4 00/91] 3.4.105-rc1 review Zefan Li
2014-11-27 20:15 ` Guenter Roeck
2014-11-29 5:54 ` Zefan Li
-- strict thread matches above, loose matches on Subject: below --
2015-01-28 4:07 [PATCH 3.4 000/177] 3.4.106-rc1 review lizf
2015-01-28 4:07 ` [PATCH 3.4 41/91] jiffies: Fix timeval conversion to jiffies lizf
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=1417077794-9299-41-git-send-email-lizf@kernel.org \
--to=lizf@kernel.org \
--cc=ahh@google.com \
--cc=john.stultz@linaro.org \
--cc=linux-kernel@vger.kernel.org \
--cc=lizefan@huawei.com \
--cc=mingo@redhat.com \
--cc=pjt@google.com \
--cc=prarit@redhat.com \
--cc=richardcochran@gmail.com \
--cc=stable@vger.kernel.org \
--cc=tglx@linutronix.de \
/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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.