Add a module parameter which when enabled, will check on resume, if the
last S0ix attempt was successful. If not, the driver would warn and provide
helpful debug information (which gets latched during the failed suspend
attempt) to debug the S0ix failure.
This information is very useful to debug S0ix failures. Specially since
the latched debug information will be lost (over-written) if the system
attempts to go into runtime (or imminent) S0ix again after that failed
suspend attempt.
Signed-off-by: Rajat Jain <rajatja@google.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
On some platforms such as HP Elite-x2-1013-g3, the platform BIOS
enforces XTAL to remain off before S0ix state can be achieved. This may
not be optimum when we want to enable use cases like Low Power Audio,
Wake on Voice etc which always need 24mhz clock.
This introduces a new quirk to allow S0ix entry when all other
conditions except for XTAL clock are good on a given platform. The extra
power consumed by XTAL clock is about 2mw but it saves much more
platform power compared to the system that remains in just PC10.
Link: https://bit.ly/2UmnrFf
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=201579
Tested-by: "David E. Box" <david.e.box@linux.intel.com>
Reported-and-tested-by: russianneuromancer <russianneuromancer@ya.ru>
Signed-off-by: Rajneesh Bhardwaj <rajneesh.bhardwaj@linux.intel.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Darren Hart (VMware) <dvhart@infradead.org>
This patch introduces a new debugfs entry to read current Package
cstate residency counters. A similar variant of this patch was discussed
earlier "https://patchwork.kernel.org/patch/9908563/" but didn't make it
into mainline for various reasons. Current version only adds debugfs
entry which is quite useful for S0ix debug but excludes the exported API
that was there in initial version. Though there are tools like turbostat
and socwatch which can also show this info but sometimes its more
practical to have it here as it's hard to switch between various tools for
S0ix debug when pmc_core driver is the primary debug tool. Internal and
external customers have requested for this patch to be included in the
PMC driver on many occasions and Google Chrome OS team has already included
it in their builds. This becomes handy when requesting logs from external
customers who may not always have above mentioned tools in their integrated
kernel builds.
Package cstate residency MSRs provide useful debug information about
system idle states. In idle states system must enter deeper Package
cstates. Package cstates depend not only on Core cstates but also on
various IP block's power gating status and LTR values.
For Intel Core SoCs Package C10 entry is a must for deeper sleep states
such as S0ix. "Suspend-to-idle" should ideally take this path:
PC0 -> PC10 -> S0ix. For S0ix debug, its logical to check for
Package C10 residency first if for some reason system fails to enter S0ix.
Please refer to this link for MSR details:
https://software.intel.com/sites/default/files/managed/22/0d/335592-sdm-vol-4.pdf
Usage:
cat /sys/kernel/debug/pmc_core/package_cstate_show
Package C2 : 0xec2e21735f
Package C3 : 0xc30113ba4
Package C6 : 0x9ef4be15c5
Package C7 : 0x1e011904
Package C8 : 0x3c5653cfe5a
Package C9 : 0x0
Package C10 : 0x16fff4289
Cc: Arjan van de Ven <arjan@linux.intel.com>
Cc: "David E. Box" <david.e.box@intel.com>
Cc: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Cc: Anshuman Gupta <anshuman.gupta@intel.com>
Cc: Len Brown <len.brown@intel.com>
Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-and-tested-by: Anshuman Gupta <anshuman.gupta@intel.com>
Signed-off-by: Rajneesh Bhardwaj <rajneesh.bhardwaj@linux.intel.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Darren Hart (VMware) <dvhart@infradead.org>
Icelake can resue most of the CNL PCH IPs as they are mostly similar.
This patch enables the PMC Core driver for ICL family.
It also addresses few other minor issues like upper case conversions and
some tab alignments.
Cc: "David E. Box" <david.e.box@intel.com>
Cc: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Acked-and-tested-by: Anshuman Gupta <anshuman.gupta@intel.com>
Signed-off-by: Rajneesh Bhardwaj <rajneesh.bhardwaj@linux.intel.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Darren Hart (VMware) <dvhart@infradead.org>
Recently introduced commit "platform/x86: intel_pmc_core: Show Latency
Tolerance info <51337cd94d18184601ac0fb4cf1a02b8bbabc3d7> skipped the
LTR from a reserved IP. Though this doesn't cause any functional issue
but it is needed for the consumers of "ltr_ignore" as the index printing
for "ltr_show" is missing. For example, w/o this change, a user that wants
to ignore LTR from ME would do something like
echo 5 > ltr_ignore
but the index for ME is 6. Printing a reserved IP helps to properly
calculate LTR ignore offsets.
Cc: "David E. Box" <david.e.box@intel.com>
Cc: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Signed-off-by: Rajneesh Bhardwaj <rajneesh.bhardwaj@linux.intel.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
A previous commit "platform/x86: intel_pmc_core: Make the driver PCH
family agnostic <c977b98bbef5898ed3d30b08ea67622e9e82082a>" provided
better abstraction to this driver but has some fundamental issues.
e.g. the following condition
for (index = 0; index < pmcdev->map->ppfear_buckets &&
index < PPFEAR_MAX_NUM_ENTRIES; index++, iter++)
is wrong because for CNL, PPFEAR_MAX_NUM_ENTRIES is hardcoded as 5 which
is _wrong_ and even though ppfear_buckets is 8, the loop fails to read
all eight registers needed for CNL PCH i.e. PPFEAR0 and PPFEAR1. This
patch refactors the pfear show logic to correctly read PCH IP power
gating status for Cannonlake and beyond.
Cc: "David E. Box" <david.e.box@intel.com>
Cc: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Fixes: c977b98bbe ("platform/x86: intel_pmc_core: Make the driver PCH family agnostic")
Signed-off-by: Rajneesh Bhardwaj <rajneesh.bhardwaj@linux.intel.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
The LTR values follow PCIE LTR encoding format and can be decoded as per
https://pcisig.com/sites/default/files/specification_documents/ECN_LatencyTolnReporting_14Aug08.pdf
This adds support to translate the raw LTR values as read from the PMC
to meaningful values in nanosecond units of time.
Signed-off-by: Rajneesh Bhardwaj <rajneesh.bhardwaj@linux.intel.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Cannonlake PCH allows us to ignore LTR from more IPs than Sunrisepoint
PCH so make the LTR ignore platform specific.
Signed-off-by: Rajneesh Bhardwaj <rajneesh.bhardwaj@linux.intel.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
This adds support to show the Latency Tolerance Reporting for the IPs on
the PCH as reported by the PMC. The format shown here is raw LTR data
payload that can further be decoded as per the PCI specification.
This also fixes some minor alignment issues in the header file by
removing spaces and converting to tabs at some places.
Signed-off-by: Rajneesh Bhardwaj <rajneesh.bhardwaj@linux.intel.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Reduce size of duplicated comments by switching to use SPDX identifier.
No functional change.
While here, correct MODULE_LICENSE() string to be aligned with license text.
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Adds debugfs access to registers in the Cannon Point PCH PMC that are
useful for debugging #SLP_S0 signal assertion and other low power relate
activities. Device pm states are latched in these registers whenever the
package enters C10 and can be read from slp_s0_debug_status. The pm
states may also be latched by writing 1 to slp_s0_dbg_latch which will
immediately capture the current state on the next read of
slp_s0_debug_status.
Signed-off-by: Box, David E <david.e.box@intel.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Intel CoffeeLake SoC uses CPU ID of KabyLake but has Cannonlake PCH, so in
this case PMC register details from Cannonlake PCH must be used.
In order to identify whether the given platform is Coffeelake, scan for the
Sunrisepoint PMC PCI Id.
KBL CPUID SPT PCIID
------------------------------------
KBL | Y | Y |
------------------------------------
CFL | Y | N |
------------------------------------
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Signed-off-by: Rajneesh Bhardwaj <rajneesh.bhardwaj@intel.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
This adds support for Cannonlake PCH which is used by Cannonlake and
Coffeelake SoCs.
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Signed-off-by: Rajneesh Bhardwaj <rajneesh.bhardwaj@intel.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
The Only use of PCI device enumeration here is to get the PMC base address
which is a fixed value i.e. 0xFE000000. On some platforms this can be read
through a non standard PCI BAR. But after Kabylake, PMC is not exposed as a
PCI device anymore. There are other non standard methods like ACPI LPIT
which can also be used for obtaining this value.
For simplicity, this value can be hardcoded as it won't change.
Since we don't have a PMC PCI device on any platform after Kabylake, this
creates a foundation for future SoC support.
Signed-off-by: Rajneesh Bhardwaj <rajneesh.bhardwaj@intel.com>
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Fix invalid field information and add missing fields in kernel doc comments.
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Signed-off-by: Rajneesh Bhardwaj <rajneesh.bhardwaj@intel.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
base_address field is redundant and unused in the driver so get rid of it.
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Signed-off-by: Rajneesh Bhardwaj <rajneesh.bhardwaj@intel.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Though ChromeOs uses the exported API as part of their S0ix failsafe
mechanism, there is no active consumer of this API in upstream kernel.
We can revisit this when ChromeOs kernel team is able to get their S0ix
failsafe framework in mainline.
Cc: Derek Basehore <dbasehore@chromium.org>
Cc: Rajat Jain <rajatja@google.com>
Link: https://patchwork.kernel.org/patch/9831229/
Suggested-by: Andriy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Signed-off-by: Rajneesh Bhardwaj <rajneesh.bhardwaj@intel.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Although this driver did pretty good job in abstracting PCH specific
interfaces, but still there are some loose ends. For example
SLP_S0 counter (for reading SLP_S0 residency), PM config offset (for
checking permissions to read XRAM) and PPFEAR offset (for reading IP
status) is still hardcoded for a specific family of PCH.
This change extended the struct pmc_reg_map to allow per family
configuration of offsets and bits.
No functional change is expected with this change. This change allows
seamless additions to new PCH and create a baseline for other platform
specific extensions.
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
SPT LTR_IGN register provides a means to make the PMC ignore the LTR values
reported by the individual PCH devices.
echo <IP Offset> > /sys/kernel/debug/pmc_core/ltr_ignore.
When a particular IP Offset bit is set the PMC will ignore the LTR value
reported by the corresponding IP when the PMC performs the latency
coalescing.
IP Offset IP Name
0 SPA
1 SPB
2 SATA
3 GBE
4 XHCI
5 RSVD
6 ME
7 EVA
8 SPC
9 Azalia/ADSP
10 RSVD
11 LPSS
12 SPD
13 SPE
14 Camera
15 ESPI
16 SCC
17 ISH
Signed-off-by: Rajneesh Bhardwaj <rajneesh.bhardwaj@intel.com>
[dvhart: pmc_core_ltr_ignore_write local declaration order cleanup]
Signed-off-by: Darren Hart <dvhart@linux.intel.com>
ModPhy Common lanes can provide the clock gating status for the important
system PLLs such as Gen2 USB3PCIE2 PLL, DMIPCIE3 PLL, SATA PLL and MIPI
PLL.
On SPT, in addition to the crystal oscillator clock, the 100Mhz Gen2
USB3PCI2 PLL clock is used as the PLL reference clock and Gen2 PLL idling
is a necessary condition for the platform to go into low power states like
PC10 and S0ix.
Signed-off-by: Rajneesh Bhardwaj <rajneesh.bhardwaj@intel.com>
Signed-off-by: Darren Hart <dvhart@linux.intel.com>
The PCH implements a number of High Speed I/O (HSIO) lanes that are split
between PCIe*, USB 3.0, SATA, GbE, USB OTG and SSIC. This patch shows the
current power gating status of the available ModPhy Core lanes. This is
done by sending a message to the PMC (MTPMC) that contains the XRAM
register offset for the MPHY_CORE_STS_0 and MPHY_CORE_STS_1 and then by
reading the response sent by the PMC (MFPMC).
While enabling low power modes we often encounter situations when the
ModPhy lanes are not power gated and it becomes hard to debug which lane is
active and which is not in the absence of an external hardware debugger
(JTAG/ITP). This patch eliminates the dependency on an external hardware
debugger for reading the ModPhy Lanes power gating status.
This patch requires PMC_READ_DISABLE setting to be disabled in the platform
bios.
cat /sys/kernel/debug/pmc_core/mphy_lanes_power_gating_status
Signed-off-by: Rajneesh Bhardwaj <rajneesh.bhardwaj@intel.com>
Signed-off-by: Darren Hart <dvhart@linux.intel.com>
This patch adds the support for reading the power gating status of various
devices present on Sunrise Point PCH. This is intended to be used for
debugging purpose while tuning the platform for power optimizations and
also to understand which devices (on PCH) are blocking the system to enter
a low power state.
Power Management Controller on Sunrise Point PCH provides access to "PGD
PFET Enable Ack Status Registers (ppfear)". This patch reads and decodes
this register and dumps the output in formatted manner showing various
devices present on the PCH and their "Power Gating" status.
Further documentation can be found in Intel 7th Gen Core family mobile u/y
processor io datasheet volume 2.
Sample output (stripped and not in order):
cat /sys/kernel/debug/pmc_core/pch_ip_power_gating_status
PMC State: Not Power gated
OPI-DMI State: Not Power gated
XHCI State: Power gated
LPSS State: Power gated
CSME_PSF State: Not power gated
Signed-off-by: Rajneesh Bhardwaj <rajneesh.bhardwaj@intel.com>
Signed-off-by: Darren Hart <dvhart@linux.intel.com>
On Sunrise Point PCH, the Power Management Controller provides 4K bytes of
memory space for various power management and debug registers. This fix is
needed to access power management & debug registers that are mapped at a
higher offset.
Also, this provides a fix for correctly masking the PWRMBASE as the initial
bits (0-11) are reserved.
Signed-off-by: Rajneesh Bhardwaj <rajneesh.bhardwaj@intel.com>
Signed-off-by: Darren Hart <dvhart@linux.intel.com>
Refactor the code to use the recently introduced
DEFINE_DEBUGFS_ATTRIBUTE() macro to eliminate boilerplate code.
Make the absence of DEBUG_FS a non-fatal error.
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Reviewed-and-tested-by: Rajneesh Bhardwaj <rajneesh.bhardwaj@intel.com>
Signed-off-by: Darren Hart <dvhart@linux.intel.com>
This patch adds the Power Management Controller driver as a PCI driver
for Intel Core SoC architecture.
This driver can utilize debugging capabilities and supported features
as exposed by the Power Management Controller.
Please refer to the below specification for more details on PMC features.
http://www.intel.in/content/www/in/en/chipsets/100-series-chipset-datasheet-vol-2.html
The current version of this driver exposes SLP_S0_RESIDENCY counter.
This counter can be used for detecting fragile SLP_S0 signal related
failures and take corrective actions when PCH SLP_S0 signal is not
asserted after kernel freeze as part of suspend to idle flow
(echo freeze > /sys/power/state).
Intel Platform Controller Hub (PCH) asserts SLP_S0 signal when it
detects favorable conditions to enter its low power mode. As a
pre-requisite the SoC should be in deepest possible Package C-State
and devices should be in low power mode. For example, on Skylake SoC
the deepest Package C-State is Package C10 or PC10. Suspend to idle
flow generally leads to PC10 state but PC10 state may not be sufficient
for realizing the platform wide power potential which SLP_S0 signal
assertion can provide.
SLP_S0 signal is often connected to the Embedded Controller (EC) and the
Power Management IC (PMIC) for other platform power management related
optimizations.
In general, SLP_S0 assertion == PC10 + PCH low power mode + ModPhy Lanes
power gated + PLL Idle.
As part of this driver, a mechanism to read the SLP_S0_RESIDENCY is exposed
as an API and also debugfs features are added to indicate SLP_S0 signal
assertion residency in microseconds.
echo freeze > /sys/power/state
wake the system
cat /sys/kernel/debug/pmc_core/slp_s0_residency_usec
Signed-off-by: Rajneesh Bhardwaj <rajneesh.bhardwaj@intel.com>
Signed-off-by: Vishwanath Somayaji <vishwanath.somayaji@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Darren Hart <dvhart@linux.intel.com>
Rajat Jain