msi_desc::masked is a misnomer. For MSI it's used to cache the MSI mask
bits when the device supports per vector masking. For MSI-X it's used to
cache the content of the vector control word which contains the mask bit
for the vector.
Replace it with a union of msi_mask and msix_ctrl to make the purpose clear
and fix up the usage sites.
No functional change
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Marc Zyngier <maz@kernel.org>
Reviewed-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20210729222543.045993608@linutronix.de
Multi-MSI uses a single MSI descriptor and there is a single mask register
when the device supports per vector masking. To avoid reading back the mask
register the value is cached in the MSI descriptor and updates are done by
clearing and setting bits in the cache and writing it to the device.
But nothing protects msi_desc::masked and the mask register from being
modified concurrently on two different CPUs for two different Linux
interrupts which belong to the same multi-MSI descriptor.
Add a lock to struct device and protect any operation on the mask and the
mask register with it.
This makes the update of msi_desc::masked unconditional, but there is no
place which requires a modification of the hardware register without
updating the masked cache.
msi_mask_irq() is now an empty wrapper which will be cleaned up in follow
up changes.
The problem goes way back to the initial support of multi-MSI, but picking
the commit which introduced the mask cache is a valid cut off point
(2.6.30).
Fixes: f2440d9acb ("PCI MSI: Refactor interrupt masking code")
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Marc Zyngier <maz@kernel.org>
Reviewed-by: Marc Zyngier <maz@kernel.org>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20210729222542.726833414@linutronix.de
msi_mask_irq() takes a mask and a flags argument. The mask argument is used
to mask out bits from the cached mask and the flags argument to set bits.
Some places invoke it with a flags argument which sets bits which are not
used by the device, i.e. when the device supports up to 8 vectors a full
unmask in some places sets the mask to 0xFFFFFF00. While devices probably
do not care, it's still bad practice.
Fixes: 7ba1930db0 ("PCI MSI: Unmask MSI if setup failed")
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Marc Zyngier <maz@kernel.org>
Reviewed-by: Marc Zyngier <maz@kernel.org>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20210729222542.568173099@linutronix.de
Nothing enforces the posted writes to be visible when the function
returns. Flush them even if the flush might be redundant when the entry is
masked already as the unmask will flush as well. This is either setup or a
rare affinity change event so the extra flush is not the end of the world.
While this is more a theoretical issue especially the logic in the X86
specific msi_set_affinity() function relies on the assumption that the
update has reached the hardware when the function returns.
Again, as this never has been enforced the Fixes tag refers to a commit in:
git://git.kernel.org/pub/scm/linux/kernel/git/tglx/history.git
Fixes: f036d4ea5fa7 ("[PATCH] ia32 Message Signalled Interrupt support")
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Marc Zyngier <maz@kernel.org>
Reviewed-by: Marc Zyngier <maz@kernel.org>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20210729222542.515188147@linutronix.de
The specification (PCIe r5.0, sec 6.1.4.5) states:
For MSI-X, a function is permitted to cache Address and Data values
from unmasked MSI-X Table entries. However, anytime software unmasks a
currently masked MSI-X Table entry either by clearing its Mask bit or
by clearing the Function Mask bit, the function must update any Address
or Data values that it cached from that entry. If software changes the
Address or Data value of an entry while the entry is unmasked, the
result is undefined.
The Linux kernel's MSI-X support never enforced that the entry is masked
before the entry is modified hence the Fixes tag refers to a commit in:
git://git.kernel.org/pub/scm/linux/kernel/git/tglx/history.git
Enforce the entry to be masked across the update.
There is no point in enforcing this to be handled at all possible call
sites as this is just pointless code duplication and the common update
function is the obvious place to enforce this.
Fixes: f036d4ea5fa7 ("[PATCH] ia32 Message Signalled Interrupt support")
Reported-by: Kevin Tian <kevin.tian@intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Marc Zyngier <maz@kernel.org>
Reviewed-by: Marc Zyngier <maz@kernel.org>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20210729222542.462096385@linutronix.de
When MSI-X is enabled the ordering of calls is:
msix_map_region();
msix_setup_entries();
pci_msi_setup_msi_irqs();
msix_program_entries();
This has a few interesting issues:
1) msix_setup_entries() allocates the MSI descriptors and initializes them
except for the msi_desc:masked member which is left zero initialized.
2) pci_msi_setup_msi_irqs() allocates the interrupt descriptors and sets
up the MSI interrupts which ends up in pci_write_msi_msg() unless the
interrupt chip provides its own irq_write_msi_msg() function.
3) msix_program_entries() does not do what the name suggests. It solely
updates the entries array (if not NULL) and initializes the masked
member for each MSI descriptor by reading the hardware state and then
masks the entry.
Obviously this has some issues:
1) The uninitialized masked member of msi_desc prevents the enforcement
of masking the entry in pci_write_msi_msg() depending on the cached
masked bit. Aside of that half initialized data is a NONO in general
2) msix_program_entries() only ensures that the actually allocated entries
are masked. This is wrong as experimentation with crash testing and
crash kernel kexec has shown.
This limited testing unearthed that when the production kernel had more
entries in use and unmasked when it crashed and the crash kernel
allocated a smaller amount of entries, then a full scan of all entries
found unmasked entries which were in use in the production kernel.
This is obviously a device or emulation issue as the device reset
should mask all MSI-X table entries, but obviously that's just part
of the paper specification.
Cure this by:
1) Masking all table entries in hardware
2) Initializing msi_desc::masked in msix_setup_entries()
3) Removing the mask dance in msix_program_entries()
4) Renaming msix_program_entries() to msix_update_entries() to
reflect the purpose of that function.
As the masking of unused entries has never been done the Fixes tag refers
to a commit in:
git://git.kernel.org/pub/scm/linux/kernel/git/tglx/history.git
Fixes: f036d4ea5fa7 ("[PATCH] ia32 Message Signalled Interrupt support")
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Marc Zyngier <maz@kernel.org>
Reviewed-by: Marc Zyngier <maz@kernel.org>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20210729222542.403833459@linutronix.de
The ordering of MSI-X enable in hardware is dysfunctional:
1) MSI-X is disabled in the control register
2) Various setup functions
3) pci_msi_setup_msi_irqs() is invoked which ends up accessing
the MSI-X table entries
4) MSI-X is enabled and masked in the control register with the
comment that enabling is required for some hardware to access
the MSI-X table
Step #4 obviously contradicts #3. The history of this is an issue with the
NIU hardware. When #4 was introduced the table access actually happened in
msix_program_entries() which was invoked after enabling and masking MSI-X.
This was changed in commit d71d6432e1 ("PCI/MSI: Kill redundant call of
irq_set_msi_desc() for MSI-X interrupts") which removed the table write
from msix_program_entries().
Interestingly enough nobody noticed and either NIU still works or it did
not get any testing with a kernel 3.19 or later.
Nevertheless this is inconsistent and there is no reason why MSI-X can't be
enabled and masked in the control register early on, i.e. move step #4
above to step #1. This preserves the NIU workaround and has no side effects
on other hardware.
Fixes: d71d6432e1 ("PCI/MSI: Kill redundant call of irq_set_msi_desc() for MSI-X interrupts")
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Marc Zyngier <maz@kernel.org>
Reviewed-by: Ashok Raj <ashok.raj@intel.com>
Reviewed-by: Marc Zyngier <maz@kernel.org>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20210729222542.344136412@linutronix.de
Reading/writing 4 bytes should be fast enough even on a slow bus, therefore
pci_vpd_wait() doesn't have to be interruptible. Making it uninterruptible
allows to simplify the code.
In addition make VPD writes uninterruptible in general. It's about vital
data, and allowing writes to be interruptible may leave the VPD in an
inconsistent state.
Link: https://lore.kernel.org/r/258bf994-bc2a-2907-9181-2c7a562986d5@gmail.com
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
vpd->len is initialized to PCI_VPD_MAX_SIZE, and if a quirk is used to set
a specific VPD size, then pci_vpd_set_size() sets vpd->valid, resulting in
pci_vpd_size() not being called. Therefore we can remove the old_size
argument. Note that we don't have to check off < PCI_VPD_MAX_SIZE because
that's implicitly done by pci_read_vpd().
Link: https://lore.kernel.org/r/ede36c16-5335-6867-43a1-293641348430@gmail.com
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Previously, if we found any error in the VPD, we returned size 0, which
prevents access to all of VPD. But there may be valid resources in VPD
before the error, and there's no reason to prevent access to those.
"off" covers only VPD resources known to have valid header tags. In case
of error, return "off" (which may be zero if we haven't found any valid
header tags at all).
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
VPD consists of a series of Small and Large Resources. Computing the size
of VPD requires only the length of each, which is specified in the generic
tag of each resource. We only expect to see ID_STRING, RO_DATA, and
RW_DATA in VPD, but it's not a problem if it contains other resource types
because all we care about is the size.
Drop the validity checking of Large Resource items.
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
VPD is limited in size by the 15-bit VPD Address field in the VPD
Capability. Each resource tag includes a length that determines the
overall size of the resource. Reject any resources that would extend past
the maximum VPD size.
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
sysfs: Allow deferred execution of iomem_get_mapping()
Tag for toerh trees/branches to pull from in order to have a stable base
to build off of for the "Allow deferred execution of
iomem_get_mapping()" set of sysfs changes
Link: https://lore.kernel.org/r/20210729233235.1508920-1-kw@linux.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
* tag 'sysfs_defferred_iomem_get_mapping-5.15' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core:
sysfs: Rename struct bin_attribute member to f_mapping
sysfs: Invoke iomem_get_mapping() from the sysfs open callback
There are two users of iomem_get_mapping(), the struct file and struct
bin_attribute. The former has a member called "f_mapping" and the
latter has a member called "mapping", and both are poniters to struct
address_space.
Rename struct bin_attribute member to "f_mapping" to keep both meaning
and the usage consistent with other users of iomem_get_mapping().
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Signed-off-by: Krzysztof Wilczyński <kw@linux.com>
Link: https://lore.kernel.org/r/20210729233235.1508920-3-kw@linux.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Defer invocation of the iomem_get_mapping() to the sysfs open callback
so that it can be executed as needed when the binary sysfs object has
been accessed.
To do that, convert the "mapping" member of the struct bin_attribute
from a pointer to the struct address_space into a function pointer with
a signature that requires the same return type, and then updates the
sysfs_kf_bin_open() to invoke provided function should the function
pointer be valid.
Also, convert every invocation of iomem_get_mapping() into a function
pointer assignment, therefore allowing for the iomem_get_mapping()
invocation to be deferred to when the sysfs open callback runs.
Thus, this change removes the need for the fs_initcalls to complete
before any other sub-system that uses the iomem_get_mapping() would be
able to invoke it safely without leading to a failure and an Oops
related to an invalid iomem_get_mapping() access.
Suggested-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Krzysztof Wilczyński <kw@linux.com>
Link: https://lore.kernel.org/r/20210729233235.1508920-2-kw@linux.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
devm_ioremap_resource() internally calls __devm_ioremap_resource() which
is where error checking and handling is actually taking place. i
Therefore, the dev_err() call in xgene_msi_probe() is redundant.
Remove it.
Link: https://lore.kernel.org/r/20210408132751.1198171-1-yangerkun@huawei.com
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: ErKun Yang <yangerkun@huawei.com>
[lorenzo.pieralisi@arm.com: commit log]
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Krzysztof Wilczyński <kw@linux.com>
Don't populate the array err_msg on the stack but instead make it
static. Makes the object code smaller by 64 bytes.
While at it, add a missing const, as reported by checkpatch.
Compiled with gcc 11.0.1
Before:
$ size drivers/pci/controller/pci-tegra.o
text data bss dec hex filename
25623 2844 32 28499 6f53 drivers/pci/controller/pci-tegra.o
After:
$ size drivers/pci/controller/pci-tegra.o
text data bss dec hex filename
25559 2844 32 28435 6f13 drivers/pci/controller/pci-tegra.o
Link: https://lore.kernel.org/r/5f3f35296b944b94546cc7d1e9cc6186484620d8.1620148539.git.christophe.jaillet@wanadoo.fr
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Vidya Sagar <vidyas@nvidia.com>
Set CRSVIS flag in emulated root PCI bridge to indicate support for
Completion Retry Status.
Add check for CRSSVE flag from root PCI brige when issuing Configuration
Read Request via PIO to correctly returns fabricated CRS value as it is
required by PCIe spec.
Link: https://lore.kernel.org/r/20210722144041.12661-5-pali@kernel.org
Fixes: 8a3ebd8de3 ("PCI: aardvark: Implement emulated root PCI bridge config space")
Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Cc: stable@vger.kernel.org # e0d9d30b73 ("PCI: pci-bridge-emul: Fix big-endian support")
Measurements in different conditions showed that aardvark hardware PIO
response can take up to 1.44s. Increase wait timeout from 1ms to 1.5s to
ensure that we do not miss responses from hardware. After 1.44s hardware
returns errors (e.g. Completer abort).
The previous two patches fixed checking for PIO status, so now we can use
it to also catch errors which are reported by hardware after 1.44s.
After applying this patch, kernel can detect and print PIO errors to dmesg:
[ 6.879999] advk-pcie d0070000.pcie: Non-posted PIO Response Status: CA, 0xe00 @ 0x100004
[ 6.896436] advk-pcie d0070000.pcie: Posted PIO Response Status: COMP_ERR, 0x804 @ 0x100004
[ 6.913049] advk-pcie d0070000.pcie: Posted PIO Response Status: COMP_ERR, 0x804 @ 0x100010
[ 6.929663] advk-pcie d0070000.pcie: Non-posted PIO Response Status: CA, 0xe00 @ 0x100010
[ 6.953558] advk-pcie d0070000.pcie: Posted PIO Response Status: COMP_ERR, 0x804 @ 0x100014
[ 6.970170] advk-pcie d0070000.pcie: Non-posted PIO Response Status: CA, 0xe00 @ 0x100014
[ 6.994328] advk-pcie d0070000.pcie: Posted PIO Response Status: COMP_ERR, 0x804 @ 0x100004
Without this patch kernel prints only a generic error to dmesg:
[ 5.246847] advk-pcie d0070000.pcie: config read/write timed out
Link: https://lore.kernel.org/r/20210722144041.12661-3-pali@kernel.org
Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Marek Behún <kabel@kernel.org>
Cc: stable@vger.kernel.org # 7fbcb5da81 ("PCI: aardvark: Don't rely on jiffies while holding spinlock")
There is an issue that when PCIe switch is connected to an Armada 3700
board, there will be lots of warnings about PIO errors when reading the
config space. According to Aardvark PIO read and write sequence in HW
specification, the current way to check PIO status has the following
issues:
1) For PIO read operation, it reports the error message, which should be
avoided according to HW specification.
2) For PIO read and write operations, it only checks PIO operation complete
status, which is not enough, and error status should also be checked.
This patch aligns the code with Aardvark PIO read and write sequence in HW
specification on PIO status check and fix the warnings when reading config
space.
[pali: Fix CRS handling when CRSSVE is not enabled]
Link: https://lore.kernel.org/r/20210722144041.12661-2-pali@kernel.org
Tested-by: Victor Gu <xigu@marvell.com>
Signed-off-by: Evan Wang <xswang@marvell.com>
Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Victor Gu <xigu@marvell.com>
Reviewed-by: Marek Behún <kabel@kernel.org>
Cc: stable@vger.kernel.org # b1bd571447 ("PCI: aardvark: Indicate error in 'val' when config read fails")
PME signaling is only enabled by __pci_enable_wake() if the target
device can signal PME from the given target power state (to avoid
pointless reconfiguration of the device), but if the hierarchy above
the device goes into D3cold, the device itself will end up in D3cold
too, so if it can signal PME from D3cold, it should be enabled to
do so in __pci_enable_wake().
[Note that if the device does not end up in D3cold and it cannot
signal PME from the original target power state, it will not signal
PME, so in that case the behavior does not change.]
Link: https://lore.kernel.org/linux-pm/3149540.aeNJFYEL58@kreacher/
Fixes: 5bcc2fb4e8 ("PCI PM: Simplify PCI wake-up code")
Reported-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reported-by: Utkarsh H Patel <utkarsh.h.patel@intel.com>
Reported-by: Koba Ko <koba.ko@canonical.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Tested-by: Mika Westerberg <mika.westerberg@linux.intel.com>
It is inconsistent to return PCI_D0 from pci_target_state() instead
of the original target state if 'wakeup' is true and the device
cannot signal PME from D0.
This only happens when the device cannot signal PME from the original
target state and any shallower power states (including D0) and that
case is effectively equivalent to the one in which PME singaling is
not supported at all. Since the original target state is returned in
the latter case, make the function do that in the former one too.
Link: https://lore.kernel.org/linux-pm/3149540.aeNJFYEL58@kreacher/
Fixes: 666ff6f83e ("PCI/PM: Avoid using device_may_wakeup() for runtime PM")
Reported-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reported-by: Utkarsh H Patel <utkarsh.h.patel@intel.com>
Reported-by: Koba Ko <koba.ko@canonical.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Tested-by: Mika Westerberg <mika.westerberg@linux.intel.com>
In suspend_noirq() call if link doesn't goto L2, PERST# is asserted
to bring link to detect state. However, this is causing surprise
link down AER error. Since Kernel is executing noirq suspend calls,
AER interrupt is not processed. PME and AER are shared interrupts
and PCIe subsystem driver enables wake capability of PME irq during
suspend. So this AER will cause suspend failure due to pending
AER interrupt.
After PCIe link is in L2, interrupts are not expected since PCIe
controller will be in reset state. Disable PCIe interrupts before
going to L2 state to avoid pending AER interrupt.
Link: https://lore.kernel.org/r/20210623100525.19944-4-omp@nvidia.com
Signed-off-by: Om Prakash Singh <omp@nvidia.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Bjorn Helgaas <bhelgaas@google.com>
Acked-by: Vidya Sagar <vidyas@nvidia.com>
The pciconfig_read() syscall reads PCI configuration space using
hardware-dependent config accessors.
If the read fails on PCI, most accessors don't return an error; they
pretend the read was successful and got ~0 data from the device, so the
syscall returns success with ~0 data in the buffer.
When the accessor does return an error, pciconfig_read() normally fills the
user's buffer with ~0 and returns an error in errno. But after
e4585da22a ("pci syscall.c: Switch to refcounting API"), we don't fill
the buffer with ~0 for the EPERM "user lacks CAP_SYS_ADMIN" error.
Userspace may rely on the ~0 data to detect errors, but after e4585da22a,
that would not detect CAP_SYS_ADMIN errors.
Restore the original behaviour of filling the buffer with ~0 when the
CAP_SYS_ADMIN check fails.
[bhelgaas: commit log, fold in Nathan's fix
https://lore.kernel.org/r/20210803200836.500658-1-nathan@kernel.org]
Fixes: e4585da22a ("pci syscall.c: Switch to refcounting API")
Link: https://lore.kernel.org/r/20210729233755.1509616-1-kw@linux.com
Signed-off-by: Krzysztof Wilczyński <kw@linux.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Cc: stable@vger.kernel.org
Previously we assumed that the first tag being 0x00 meant an EEPROM was
missing. The first tag being 0xff means the same thing; check for that
also.
[bhelgaas: rework error mesage]
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Previously, we checked for PCI_VPD_STIN_END, PCI_VPD_LTIN_ID_STRING, etc.,
outside the Large and Small Resource cases, so we checked Large Resource
Item Names against a Small Resource name and vice versa.
Move these tests into the Large and Small Resource cases, so we only check
PCI_VPD_STIN_END for Small Resources and PCI_VPD_LTIN_* for Large
Resources.
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Previously, when a VPD read failed, we warned about an "invalid large
VPD tag". Warn about the VPD read failure instead.
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
