[<prev] [next>] [day] [month] [year] [list]
Message-ID: <CALurfZULPK_tXpqGpGcXM53wCBXf0VCds-E1iKCKZcw2OWoqvg@mail.gmail.com>
Date: Tue, 16 Oct 2012 10:04:03 -0400
From: Yclept Nemo <orbisvicis@...il.com>
To: "Rafael J. Wysocki" <rjw@...k.pl>
Cc: cpufreq@...r.kernel.org, linux-pm@...r.kernel.org,
linux-kernel@...r.kernel.org
Subject: PROBLEM: CPUFreq BIOS limit
[1.] One line summary of the problem:
CPUFreq BIOS limit problems
[2.] Full description of the problem/report:
The frequency steps of my laptop range from 1GHz to 1.83GHz. When I
disconnect my AC adapter the BIOS enforces an upper frequency limit of
1GHz for a 'settling' period of around 15 seconds. After these
15-or-so seconds have elapsed, the BIOS limit is lifted, and the cpu
frequency is restored to its previous setting - the original policy
without any CPUFREQ_ADJUST/CPUFREQ_INCOMPATIBLE notification
adjustments.
Example:
1GHz-1.83GHz-performance ->
(disconnect AC Adapter) 1GHz-1GHz-performance ->
(approx. 15 seconds later) 1GHz-1.83GHz-performance
There are two problems with how CPUFreq handles the BIOS limit:
[2.1]
Unable to restore the previous limits if (previous_limit->min >
bios_limit->max). When this occurs my laptop is effectively
permanently stuck in the lowest operating frequency even though the
BIOS limit has been lifted.
The problem is at: cpufreq.c:__cpufreq_set_policy:1674
if (policy->min > data->max || policy->max < data->min) {
ret = -EINVAL;
goto error_out;
}
Debugging excerpt from the CPUFreq module:
Also available as attachment "dmesg.filter.test.bug.reset.txt".
# description:
# governor: performance
# limits: 1.83GHz - 1.83GHz
# test: remove AC, triggering BIOS limit
# wait till BIOS limit lifted
# result: fail to reset correct limits
[ 1817.701944] starting test
[ 1824.463248] cpufreq:cpufreq_update_policy:1770: updating policy for CPU 0
[ 1824.463258] cpufreq:__cpufreq_set_policy:1669: setting new policy
for CPU 0: 1830000 - 1830000 kHz
[ 1824.463265] cpufreq:__cpufreq_set_policy:1706: new min and max
freqs are 1000000 - 1000000 kHz
[ 1824.463279] cpufreq:__cpufreq_set_policy:1739: governor: change or
update limits
[ 1824.463282] cpufreq:__cpufreq_governor:1562: __cpufreq_governor for
CPU 0, event 3
[ 1824.463286] cpufreq:__cpufreq_driver_target:1476: target for CPU 0:
1000000 kHz, relation 1
[ 1824.463291] cpufreq:cpufreq_notify_transition:267: notification 0
of frequency transition to 1000000 kHz
[ 1824.463295] cpufreq:cpufreq_notify_transition:267: notification 1
of frequency transition to 1000000 kHz
[ 1824.463310] cpufreq:cpufreq_notify_transition:294: FREQ: 1000000 - CPU: 0
[ 1824.463544] cpufreq:cpufreq_update_policy:1770: updating policy for CPU 1
[ 1824.463552] cpufreq:__cpufreq_set_policy:1669: setting new policy
for CPU 1: 1830000 - 1830000 kHz
[ 1824.463567] cpufreq:__cpufreq_set_policy:1706: new min and max
freqs are 1000000 - 1000000 kHz
[ 1824.463570] cpufreq:__cpufreq_set_policy:1739: governor: change or
update limits
[ 1824.463573] cpufreq:__cpufreq_governor:1562: __cpufreq_governor for
CPU 1, event 3
[ 1824.463576] cpufreq:__cpufreq_driver_target:1476: target for CPU 1:
1000000 kHz, relation 1
[ 1824.463580] cpufreq:cpufreq_notify_transition:267: notification 0
of frequency transition to 1000000 kHz
[ 1824.463609] cpufreq:cpufreq_notify_transition:267: notification 1
of frequency transition to 1000000 kHz
[ 1824.463614] cpufreq:cpufreq_notify_transition:294: FREQ: 1000000 - CPU: 1
[ 1826.285314] cpufreq:cpufreq_update_policy:1770: updating policy for CPU 0
[ 1826.285328] cpufreq:__cpufreq_set_policy:1669: setting new policy
for CPU 0: 1830000 - 1830000 kHz
[ 1826.285551] cpufreq:cpufreq_update_policy:1770: updating policy for CPU 1
[ 1826.285564] cpufreq:__cpufreq_set_policy:1669: setting new policy
for CPU 1: 1830000 - 1830000 kHz
[ 1835.187612] ending test
[2.2]
Race condition between kernel and userspace code when BIOS limit is
lifted. If a single event triggers both an ACPI event and a change in
the BIOS limit, both the kernel (cpufreq:cpufreq_update_policy) and
userspace code (any CPUFreq daemon listening to ACPI events) will
adjust the cpu frequency policy.
For example I run the cpu frequency scaling daemon cpufreqd which
responds to ACPI events. In the following sequence of events:
1GHz-1.83GHz-performance ->
(disconnect AC Adapter) 1GHz-1GHz-performance (BIOS limit) ->
(reconnect AC Adapter before BIOS limit lifted)
reconnecting the AC adapter lifts the BIOS limit and triggers an ACPI
event. The daemon cpufreqd reacts almost instantaneously and sets a
new policy. The kernel, most likely due to calling
blocking_notifier_call_chain three times with CPUFREQ_ADJUST,
CPUFREQ_INCOMPATIBLE, and CPUFREQ_NOTIFY, responds much more slowly
and overrides the new policy cpufreqd set. Cpufreqd is unaware of this
change and thus the wrong policy has been set.
Some possible solutions:
* write lock on the cpu policy
* kernel->userspace notifier
[3.] Keywords (i.e., modules, networking, kernel):
[4.] Kernel information
[4.1.] Kernel version (from /proc/version):
[4.2.] Kernel .config file:
[5.] Most recent kernel version which did not have the bug:
[6.] Output of Oops.. message (if applicable) with symbolic information
resolved (see Documentation/oops-tracing.txt)
[7.] A small shell script or example program which triggers the
problem (if possible)
[8.] Environment
[8.1.] Software (add the output of the ver_linux script here)
Linux 3.6.2-1-ARCH #1 SMP PREEMPT Mon Oct 15 02:54:09 EDT 2012 i686 GNU/Linux
Gnu C 4.7.2
Gnu make 3.82
binutils 2.22.0.20120323
util-linux 2.22
mount debug
module-init-tools 10
e2fsprogs 1.42.5
jfsutils 1.1.15
reiserfsprogs 3.6.21
xfsprogs 3.1.8
pcmciautils 018
quota-tools 4.01.
PPP 2.4.5
Linux C Library 2.16
Dynamic linker (ldd) 2.16
Linux C++ Library 6.0.17
Procps 3.3.3
Net-tools 1.60
Kbd 1.15.3
oprofile 0.9.7
Sh-utils 8.19
wireless-tools 29
Modules Loaded fuse act_police cls_basic cls_flow cls_fw
cls_u32 sch_tbf sch_prio sch_htb sch_hfsc sch_ingress sch_sfq
ipt_rpfilter ip6t_rpfilter xt_statistic xt_CT xt_LOG xt_connlimit
xt_realm xt_addrtype xt_comment ipt_ULOG ipt_REJECT ipt_REDIRECT
ipt_NETMAP aes_i586 xt_recent ipt_MASQUERADE ipt_ECN ipt_CLUSTERIP
ipt_ah ablk_helper cryptd aes_generic nf_nat_tftp nf_nat_snmp_basic
nf_conntrack_snmp nf_nat_sip nf_nat_pptp nf_nat_proto_gre nf_nat_irc
nf_nat_h323 nf_nat_ftp xt_set ip_set nf_nat_amanda nf_conntrack_sane
nf_conntrack_tftp nf_conntrack_sip nf_conntrack_proto_udplite
nf_conntrack_proto_sctp ts_kmp nf_conntrack_pptp
nf_conntrack_proto_gre nf_conntrack_amanda nf_conntrack_netlink
nf_conntrack_netbios_ns nf_conntrack_broadcast nf_conntrack_irc
nf_conntrack_h323 nf_conntrack_ftp xt_time xt_TCPMSS xt_sctp xt_TPROXY
nf_tproxy_core xt_tcpmss xt_policy xt_pkttype xt_physdev xt_NFLOG
xt_owner nfnetlink_log xt_NFQUEUE xt_multiport xt_mark xt_mac xt_limit
xt_length xt_iprange xt_helper xt_hashlimit xt_DSCP xt_dscp xt_dccp
xt_connmark xt_CLASSIFY xt_AUDIT ip6t_REJECT xt_tcpudp xt_state
iptable_raw nf_conntrack_ipv6 nf_defrag_ipv6 iptable_nat nf_nat
xt_conntrack nf_conntrack_ipv4 nf_defrag_ipv4 ip6table_raw
nf_conntrack ip6table_mangle iptable_mangle nfnetlink ip6table_filter
iptable_filter ip6_tables ip_tables x_tables ext4 jbd2 mbcache arc4
snd_usb_audio snd_usbmidi_lib snd_rawmidi iwl3945 iTCO_wdt
snd_seq_device iTCO_vendor_support iwlegacy btusb nvidia gpio_ich
joydev serio_raw b44 bluetooth acpi_cpufreq mac80211 microcode
kvm_intel kvm mperf crc16 dell_wmi sparse_keymap ssb pcmcia mii
snd_hda_codec_idt coretemp evdev lpc_ich pcmcia_core sr_mod cdrom
firewire_ohci firewire_core crc_itu_t dell_laptop dcdbas cfg80211
rfkill psmouse r852 sm_common nand snd_hda_intel nand_ecc
snd_hda_codec nand_ids intel_agp snd_pcm intel_gtt agpgart mtd
snd_page_alloc snd_hwdep sdhci_pci sdhci mmc_core snd_timer snd
soundcore r592 memstick i2c_i801 i2c_core wmi button ac thermal
processor video battery nfs lockd sunrpc fscache i8k hid_generic
usbhid hid xfs sd_mod ata_piix uhci_hcd libata scsi_mod ehci_hcd
usbcore usb_common
[8.2.] Processor information (from /proc/cpuinfo):
Intel(R) CPU T2400 @ 1.83GHz
[8.3.] Module information (from /proc/modules):
[8.4.] Loaded driver and hardware information (/proc/ioports, /proc/iomem)
[8.5.] PCI information ('lspci -vvv' as root)
[8.6.] SCSI information (from /proc/scsi/scsi)
[8.7.] Other information that might be relevant to the problem
(please look in /proc and include all information that you
think to be relevant):
laptop model: DELL E1705/9400
[X.] Other notes, patches, fixes, workarounds:
View attachment "dmesg.filter.test.bug.reset.txt" of type "text/plain" (2302 bytes)
Powered by blists - more mailing lists