18 Sep, 2009
2 commits
-
When booting with pci=nomsi aer causes lost interrupts and
lockdep inversions.So check if MSIs are not disabled before initializing the aer
driver.Signed-off-by: Andi Kleen
Signed-off-by: Jesse Barnes -
The definition of the ASPM support field in the Link Capabilities
Register had been changed by the "ASPM optionality ECN" as follows:00b Reserved
01b L0s Supported
10b Reserved
11b L0s and L1 Supported00b No ASPM Support
01b L0s Supported
10b L1 Supported
11b L0s and L1 SupportedCurrent linux ASPM driver doesn't enable ASPM if the support field is
00b or 10b. So there is no impact about 00b. But current linux ASPM
driver doesn't enable L1 if the support field is 10b. With this patch,
10b (L1 support) is handled properly.Signed-off-by: Kenji Kaneshige
Signed-off-by: Jesse Barnes
15 Sep, 2009
1 commit
-
After commit c82f63e411f1b58427c103bd95af2863b1c96dd1
(PCI: check saved state before restore) pcie_portdrv_slot_reset()
may not work correctly if dev->error_state is equal to
pci_channel_io_frozen, because dev->state_saved need not be set at
that time. Fix this issue by setting dev->state_saved before
pci_restore_state() is called in pcie_portdrv_slot_reset().Signed-off-by: Rafael J. Wysocki
Signed-off-by: Jesse Barnes
11 Sep, 2009
1 commit
-
Remove unused port_data variable left over from the MCH hotplug quirk
cleanup.Signed-off-by: Jesse Barnes
10 Sep, 2009
20 commits
-
There is a very old quirk for the intel E7502 E7320 and E7525 memory
controller hubs that disables usage of msi interrupts on pcie hotplug
bridges of those devices, and disables changing the affinity of irqs.Today all we have to do to disable msi on a specific device is to set
dev->no_msi, which is much more straightforward than the previous
logic.The re-running of this fixup after pci hotplug happens below these
devices is totally bogus. All of the state we change is pure software
state and we don't change the hardware at all. Which means hotplug on
the lower devices doesn't have a chance to change this state. So we
can safely remove the special case from the pciehp driver and the pcie
portdriver.I suspect the special case was someone's expermental debug code that
slipped in. Certainly it isn't mentioned in commit
6fb8880a61510295aece04a542767161f624dffe aka BKrev:
41966101LJ_ogfOU0m2aE6teZfQnuQ where the code first appears.Reviewed-by: Kenji Kaneshige
Signed-off-by: "Eric W. Biederman"
Signed-off-by: Jesse Barnes -
This patch is required not to lost error records by action invoked on
error recovery, such as slot reset etc.Following sample (real machine + dummy record injected by aer-inject)
shows that record of 28:00.1 could not be retrieved by recovery of 28:00.0:- Before:
pcieport-driver 0000:00:02.0: AER: Multiple Uncorrected (Non-Fatal) error received: id=2801
e1000e 0000:28:00.0: PCIE Bus Error: severity=Uncorrected (Non-Fatal), type=Transaction Layer, id=2800(Receiver ID)
e1000e 0000:28:00.0: device [8086:1096] error status/mask=00001000/00100000
e1000e 0000:28:00.0: [12] Poisoned TLP (First)
e1000e 0000:28:00.0: TLP Header: 00000000 00000001 00000002 00000003
e1000e 0000:28:00.0: broadcast error_detected message
e1000e 0000:28:00.0: broadcast slot_reset message
e1000e 0000:28:00.0: setting latency timer to 64
e1000e 0000:28:00.0: restoring config space at offset 0x1 (was 0x100547, writing 0x100147)
e1000e 0000:28:00.0: PME# disabled
e1000e 0000:28:00.0: PME# disabled
e1000e 0000:28:00.1: setting latency timer to 64
e1000e 0000:28:00.1: restoring config space at offset 0x1 (was 0x100547, writing 0x100147)
e1000e 0000:28:00.1: PME# disabled
e1000e 0000:28:00.1: PME# disabled
e1000e 0000:28:00.0: broadcast resume message
e1000e 0000:28:00.0: AER driver successfully recovered
e1000e: eth0 NIC Link is Up 1000 Mbps Full Duplex, Flow Control: RX/TX- After:
pcieport-driver 0000:00:02.0: AER: Multiple Uncorrected (Non-Fatal) error received: id=2801
e1000e 0000:28:00.0: PCIE Bus Error: severity=Uncorrected (Non-Fatal), type=Transaction Layer, id=2800(Receiver ID)
e1000e 0000:28:00.0: device [8086:1096] error status/mask=00001000/00100000
e1000e 0000:28:00.0: [12] Poisoned TLP (First)
e1000e 0000:28:00.0: TLP Header: 00000000 00000001 00000002 00000003
e1000e 0000:28:00.1: PCIE Bus Error: severity=Uncorrected (Non-Fatal), type=Transaction Layer, id=2801(Receiver ID)
e1000e 0000:28:00.1: device [8086:1096] error status/mask=00081000/00100000
e1000e 0000:28:00.1: [12] Poisoned TLP (First)
e1000e 0000:28:00.1: [19] ECRC
e1000e 0000:28:00.1: TLP Header: 00000000 00000001 00000002 00000003
e1000e 0000:28:00.1: Error of this Agent(2801) is reported first
e1000e 0000:28:00.0: broadcast error_detected message
e1000e 0000:28:00.0: broadcast slot_reset message
e1000e 0000:28:00.0: setting latency timer to 64
e1000e 0000:28:00.0: restoring config space at offset 0x1 (was 0x100547, writing 0x100147)
e1000e 0000:28:00.0: PME# disabled
e1000e 0000:28:00.0: PME# disabled
e1000e 0000:28:00.1: setting latency timer to 64
e1000e 0000:28:00.1: restoring config space at offset 0x1 (was 0x100547, writing 0x100147)
e1000e 0000:28:00.1: PME# disabled
e1000e 0000:28:00.1: PME# disabled
e1000e 0000:28:00.0: broadcast resume message
e1000e 0000:28:00.0: AER driver successfully recovered
e1000e: eth0 NIC Link is Up 1000 Mbps Full Duplex, Flow Control: RX/TXSigned-off-by: Hidetoshi Seto
Signed-off-by: Jesse Barnes -
Use dev_printk like format.
Sample (real machine + dummy error injected by aer-inject):
- Before:
+------ PCI-Express Device Error ------+
Error Severity : Corrected
PCIE Bus Error type : Data Link Layer
Bad TLP :
Receiver ID : 2800
VendorID=8086h, DeviceID=1096h, Bus=28h, Device=00h, Function=00h
+------ PCI-Express Device Error ------+
Error Severity : Corrected
PCIE Bus Error type : Data Link Layer
Bad TLP :
Bad DLLP :
Receiver ID : 2801
VendorID=8086h, DeviceID=1096h, Bus=28h, Device=00h, Function=01h
Error of this Agent(2801) is reported first- After:
pcieport-driver 0000:00:02.0: AER: Multiple Corrected error received: id=2801
e1000e 0000:28:00.0: PCIE Bus Error: severity=Corrected, type=Data Link Layer, id=2800(Receiver ID)
e1000e 0000:28:00.0: device [8086:1096] error status/mask=00000040/00000000
e1000e 0000:28:00.0: [ 6] Bad TLP
e1000e 0000:28:00.1: PCIE Bus Error: severity=Corrected, type=Data Link Layer, id=2801(Receiver ID)
e1000e 0000:28:00.1: device [8086:1096] error status/mask=000000c0/00000000
e1000e 0000:28:00.1: [ 6] Bad TLP
e1000e 0000:28:00.1: [ 7] Bad DLLP
e1000e 0000:28:00.1: Error of this Agent(2801) is reported firstSigned-off-by: Hidetoshi Seto
Signed-off-by: Jesse Barnes -
Compact struct and codes.
Signed-off-by: Hidetoshi Seto
Signed-off-by: Jesse Barnes -
Cleanup.
Signed-off-by: Hidetoshi Seto
Signed-off-by: Jesse Barnes -
Multiple bits might be set in the Uncorrectable Error Status
register. But aer_print_error_source() only report a error of
the lowest bit set in the error status register.So print strings for all bits unmasked and set.
And check First Error Pointer to mark the error occured first.
This FEP is not valid when the corresponing bit of the Uncorrectable
Error Status register is not set, or unimplemented or undefined.Signed-off-by: Hidetoshi Seto
Signed-off-by: Jesse Barnes -
ERR_{,UN}CORRECTABLE_ERROR_MASK are set of error bits which linux know,
set of PCI_ERR_COR_* and PCI_ERR_UNC_* defined in linux/pci_regs.h.
This masks make aerdrv not to report errors of unknown bit, while aerdrv
have ability to report such undefined errors as "Unknown Error Bit %2d".OTOH aerdrv_errprint does not have any check of setting in mask register.
So it could report masked wrong error by finding bit in status without
knowing that the bit is masked in the mask register.This patch changes aerdrv to use mask state in mask register propely
instead of defined/hardcoded ERR_{,UN}CORRECTABLE_ERROR_MASK.
This change prevents aerdrv from reporting masked error, and also enable
reporting unknown errors.Signed-off-by: Hidetoshi Seto
Reviewed-by: Andrew Patterson
Signed-off-by: Jesse Barnes -
The static buffer errmsg_buff[] is used only for building error
message in fixed format, and is protected by a spinlock.This patch removes this buffer and the spinlock.
Signed-off-by: Hidetoshi Seto
Reviewed-by: Andrew Patterson
Signed-off-by: Jesse Barnes -
The flag AER_MULTI_ERROR_VALID_FLAG in info->flag does mean that the
root port receives multiple error messages. Error messages can be
posted from different devices, so it does not mean that each reported
device has multiple errors.If there are multiple error devices and the root port has valid error
source ID, it would be nice to report which device is the error source
reported first.Signed-off-by: Hidetoshi Seto
Signed-off-by: Jesse Barnes -
In case of multiple errors, struct aer_err_info would be reused among
all reported devices. So the info->status should be initialized before
recycled. Otherwise error of one device might be reported as the error
of another device. Also info->flags has similar problem on reporting
TLP header.Signed-off-by: Hidetoshi Seto
Signed-off-by: Jesse Barnes -
Definitions of MASK macros in aerdrv_errprint.c are tricky and unsafe.
For example, AER_AGENT_TRANSMITTER_MASK(_sev, _stat) does work like:
static inline func(int _sev, int _stat)
{
if (_sev == AER_CORRECTABLE)
return (_stat & (PCI_ERR_COR_REP_ROLL|PCI_ERR_COR_REP_TIMER));
else
return (_stat & PCI_ERR_COR_REP_ROLL);
}
In case of else path here, for uncorrectable errors, testing bits in
_stat by PCI_ERR_COR_* does not make sense because _stat should have only
PCI_ERR_UNC_* bits originated in uncorrectable error status register.
But at this time this is safe because uncorrectable error using bit
position same to PCI_ERR_COR_REP_ROLL(= bit position 8) is not defined.
Likewise, AER_AGENT_COMPLETER_MASK is always PCI_ERR_UNC_COMP_ABORT but
it works because bit 15 of correctable error status is not defined.It means that these MASK macros will turn to be wrong once if new error
is defined. (In fact, bit 15 of correctable is now defined in PCIe 2.1)This patch changes these MASK macros to be more strict, not to return
PCI_ERR_COR_* bits for uncorrectable error status and vise versa.Signed-off-by: Hidetoshi Seto
Reviewed-by: Andrew Patterson
Signed-off-by: Jesse Barnes -
Add workaround macro to reduce the number of checkpatch warning:
WARNING: printk() should include KERN_ facility levelBefore:
total: 0 errors, 10 warnings, 247 lines checked
After:
total: 0 errors, 1 warnings, 243 lines checkedSigned-off-by: Hidetoshi Seto
Signed-off-by: Jesse Barnes -
Before:
drivers/pci/pcie/aer/aer_inject.c
total: 4 errors, 4 warnings, 473 lines checked
drivers/pci/pcie/aer/aerdrv.c
total: 5 errors, 2 warnings, 333 lines checked
drivers/pci/pcie/aer/aerdrv.h
total: 1 errors, 0 warnings, 139 lines checked
drivers/pci/pcie/aer/aerdrv_core.c
total: 4 errors, 3 warnings, 872 lines checked
drivers/pci/pcie/aer/aerdrv_errprint.c
total: 12 errors, 11 warnings, 248 lines checkedAfter:
drivers/pci/pcie/aer/aer_inject.c
total: 0 errors, 0 warnings, 466 lines checked
drivers/pci/pcie/aer/aerdrv.c
total: 0 errors, 0 warnings, 335 lines checked
drivers/pci/pcie/aer/aerdrv.h
total: 0 errors, 0 warnings, 139 lines checked
drivers/pci/pcie/aer/aerdrv_core.c
total: 0 errors, 0 warnings, 869 lines checked
drivers/pci/pcie/aer/aerdrv_errprint.c
total: 0 errors, 10 warnings, 247 lines checkedSigned-off-by: Hidetoshi Seto
Reviewed-by: Andrew Patterson
Signed-off-by: Jesse Barnes -
The L0s state can be managed separately for each direction (upstream
direction and downstream direction) of the link. But in the current
implementation, those are mixed up. With this patch, L0s for each
direction are managed separately.To maintain three states (upstream direction L0s, downstream L0s and
L1), 'aspm_support', 'aspm_enabled', 'aspm_capable', 'aspm_disable'
and 'aspm_default' fields in struct pcie_link_state are changed to
3-bit from 2-bit. The 'latency' field is separated to two 'latency_up'
and 'latency_dw' fields to maintain exit latencies for each direction
of the link. For L0, 'latency_up.l0' and 'latency_dw.l0' are used to
configure upstream direction L0s and downstream direction L0s
respectively. For L1, larger value of 'latency_up.l1' and
'latency_dw.l1' is considered as L1 exit latency.Acked-by: Shaohua Li
Signed-off-by: Kenji Kaneshige
Signed-off-by: Jesse Barnes -
In the current implementation, ASPM L0s/L1 is disabled for all links
in the hierarchy if one of the link doesn't meet latency requirement.
But we can partially enable ASPM L0s/L1 on sub-tree in the hierarchy.
This patch allows partial L0s/L1 enablement in the hierarchy. And it
also reduce the calculation cost of ASPM configuration very much.In the previous implementation, all links were enabled with the same
state. With this patch, enabled state for each link is determined
simply as follows (the 'requested' is from policy_to_aspm_state()).enabled = requested & (link->aspm_capable & link->aspm_disable)
Acked-by: Shaohua Li
Signed-off-by: Kenji Kaneshige
Signed-off-by: Jesse Barnes -
Introduce 'aspm_capable' field to maintain the capable ASPM setting of
the link. By the 'aspm_capable', we don't need to recheck latency
every time ASPM policy is changed.Each bit in 'aspm_capable' is associated to ASPM state (L0S/L1). The
bit is set if the associated ASPM state is supported by the link and
it satisfies the latency requirement (i.e. exit latency < endpoint
acceptable latency). The 'aspm_capable' is updated when- an endpoint device is added (boot time or hot-plug time)
- an endpoint device is removed (hot-unplug time)
- PCI power state is changed.Acked-by: Shaohua Li
Signed-off-by: Kenji Kaneshige
Signed-off-by: Jesse Barnes -
Introduce 'aspm_disable' flag to manage disabled ASPM state more
robust way.Acked-by: Shaohua Li
Signed-off-by: Kenji Kaneshige
Signed-off-by: Jesse Barnes -
Fix possible NULL dereference in pcie_aspm_exit_link_state(). This
patch also cleanup some code.Acked-by: Shaohua Li
Signed-off-by: Kenji Kaneshige
Signed-off-by: Jesse Barnes -
Remove the following check in __pcie_aspm_config_link() because it
nerver be true.Acked-by: Shaohua Li
Signed-off-by: Kenji Kaneshige
Signed-off-by: Jesse Barnes -
We must not clear bits in 'aspm_enabled' using 'aspm_support', or
'aspm_enabled' and 'aspm_default' might be different from the actual
state. In addtion, 'aspm_default' should be intialized even if
'aspm_support' is 0.Acked-by: Shaohua Li
Signed-off-by: Kenji Kaneshige
Signed-off-by: Jesse Barnes
30 Jun, 2009
1 commit
-
Acked-by: Andrew Patterson
Signed-off-by: Joe Perches
Signed-off-by: Jesse Barnes
19 Jun, 2009
14 commits
-
By having a pointer to the root port link, we can remove loops in
get_root_port_link() to search the root port link.Acked-by: Shaohua Li
Signed-off-by: Kenji Kaneshige
Signed-off-by: Jesse Barnes -
Minor cleanup for pcie_aspm_sanity_check().
Acked-by: Shaohua Li
Signed-off-by: Kenji Kaneshige
Signed-off-by: Jesse Barnes -
We don't need the 'has_switch' field in the struct pcie_link_state.
Acked-by: Shaohua Li
Signed-off-by: Kenji Kaneshige
Signed-off-by: Jesse Barnes -
Cleanup for calc_L0S_latency() and calc_L1_latency().
- Separate exit latency and acceptable latency calculation.
- Some minor cleanups.Acked-by: Shaohua Li
Signed-off-by: Kenji Kaneshige
Signed-off-by: Jesse Barnes -
Minor cleanup for pcie_aspm_get_cap_device().
Acked-by: Shaohua Li
Signed-off-by: Kenji Kaneshige
Signed-off-by: Jesse Barnes -
In the current ASPM implementation, callers of pcie_set_clock_pm() check
Clock PM capability of the link or current Clock PM state of the link.
This check should be done in pcie_set_clock_pm() itself.This patch moves those checks into pcie_set_clock_pm(). It also
introduces pcie_set_clkpm_nocheck() that is equivalent to old
pcie_set_clock_pm(), for the caller who wants to change Clocl PM state
regardless of the Clock PM capability or current Clock PM state. In
addition, this patch changes the function name from
pcie_set_clock_pm() to pcie_set_clkpm() for consistency.Acked-by: Shaohua Li
Signed-off-by: Kenji Kaneshige
Signed-off-by: Jesse Barnes -
Clean up and simplify __pcie_aspm_check_state_one().
Acked-by: Shaohua Li
Signed-off-by: Kenji Kaneshige
Signed-off-by: Jesse Barnes -
Clean up ASPM initialization by refactoring some functionality, renaming
functions, and moving things around.Acked-by: Shaohua Li
Signed-off-by: Kenji Kaneshige
Signed-off-by: Jesse Barnes -
In the current ASPM implementation, there are many functions that
take a pointer to struct pci_dev corresponding to the upstream component
of the link as a parameter. But, since those functions handle PCI
express link state, a pointer to struct pcie_link_state is more
suitable than a pointer to struct pci_dev. Changing a parameter to a
pointer to struct pcie_link_state makes ASPM code much simpler and
easier to read. This patch also contains some minor cleanups. This patch
doesn't have any functional change.Acked-by: Shaohua Li
Signed-off-by: Kenji Kaneshige
Signed-off-by: Jesse Barnes -
Cleanup for some fields in pcie_link_state.
- Add comments.
- make "downstream_has_switch" field 1-bit.Acked-by: Shaohua Li
Signed-off-by: Kenji Kaneshige
Signed-off-by: Jesse Barnes -
The "clk_pm_capable", "clk_pm_enable" and "bios_clk_state" fields in
the struct pcie_link_state only take 1-bit value. So those fields
don't need to be defined as unsigned int. This patch makes those
fields 1-bit, and cleans up some related code.Acked-by: Shaohua Li
Signed-off-by: Kenji Kaneshige
Signed-off-by: Jesse Barnes -
Clean up latency related data structures for ASPM.
- Introduce struct acpi_latency for exit latency and acceptable
latency management. With this change, struct endpoint_state is no
longer needed.- We don't need to hold both upstream latency and downstream latency
in the current implementation.Acked-by: Shaohua Li
Signed-off-by: Kenji Kaneshige
Signed-off-by: Jesse Barnes -
The "support_state", "enabled_state" and "bios_aspm_state" fields in
the struct pcie_link_state take 2-bit value. So those fields don't
need to be defined as unsigned int. This patch makes those fields
2-bit, and cleans up some related code.Acked-by: Shaohua Li
Signed-off-by: Kenji Kaneshige
Signed-off-by: Jesse Barnes -
Fix a typo in struct pcie_link_state.
The "sibiling" field in the struct pcie_link_state should be
"sibling".Acked-by: Shaohua Li
Signed-off-by: Kenji Kaneshige
Signed-off-by: Jesse Barnes
17 Jun, 2009
1 commit
-
Debugging PCIE AER code can be very difficult because it is hard
to trigger various real hardware errors. This patch provide a
software based error injection tool, which can fake various PCIE
errors with a user space helper tool named "aer-inject". Which
can be gotten from:http://www.kernel.org/pub/linux/kernel/people/yhuang/
The patch fakes AER error by faking some PCIE AER related
registers and an AER interrupt for specified the PCIE device.Signed-off-by: Huang Ying
Signed-off-by: Jesse Barnes
