Zenghui reports that since 1396e89e09 ("genirq/msi: Move prepare() call
to per-device allocation"), his Multi-MSI capable device isn't working
anymore.
This is a consequence of 15c72f824b ("PCI/MSI: Add support for per device
MSI[X] domains"), which always creates a MSI domain of size 1, even in the
presence of Multi-MSI.
While this was somehow working until then, moving the .prepare() call ends
up sizing the ITS table with a tiny value for this device, and making the
endpoint driver unhappy.
Instead, always create the domain and call the .prepare() helper with the
maximum expected size.
Fixes: 1396e89e09 ("genirq/msi: Move prepare() call to per-device allocation")
Fixes: 15c72f824b ("PCI/MSI: Add support for per device MSI[X] domains")
Reported-by: Zenghui Yu <yuzenghui@huawei.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Zenghui Yu <yuzenghui@huawei.com>
Reviewed-by: Lorenzo Pieralisi <lpieralisi@kernel.org>
Link: https://lore.kernel.org/all/20250603141801.915305-1-maz@kernel.org
Closes: https://lore.kernel.org/r/0b1d7aec-1eac-a9cd-502a-339e216e08a1@huawei.com
Pull MSI updates from Thomas Gleixner:
"Updates for the MSI subsystem (core code and PCI):
- Switch the MSI descriptor locking to lock guards
- Replace a broken and naive implementation of PCI/MSI-X control word
updates in the PCI/TPH driver with a properly serialized variant in
the PCI/MSI core code.
- Remove the MSI descriptor abuse in the SCCI/UFS/QCOM driver by
replacing the direct access to the MSI descriptors with the proper
API function calls. People will never understand that APIs exist
for a reason...
- Provide core infrastructre for the upcoming PCI endpoint library
extensions. Currently limited to ARM GICv3+, but in theory
extensible to other architectures.
- Provide a MSI domain::teardown() callback, which allows drivers to
undo the effects of the prepare() callback.
- Move the MSI domain::prepare() callback invocation to domain
creation time to avoid redundant (and in case of ARM/GIC-V3-ITS
confusing) invocations on every allocation.
In combination with the new teardown callback this removes some
ugly hacks in the GIC-V3-ITS driver, which pretended to work around
the short comings of the core code so far. With this update the
code is correct by design and implementation.
- Make the irqchip MSI library globally available, provide a MSI
parent domain creation helper and convert a bunch of (PCI/)MSI
drivers over to the modern MSI parent mechanism. This is the first
step to get rid of at least one incarnation of the three PCI/MSI
management schemes.
- The usual small cleanups and improvements"
* tag 'irq-msi-2025-05-25' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (33 commits)
PCI/MSI: Use bool for MSI enable state tracking
PCI: tegra: Convert to MSI parent infrastructure
PCI: xgene: Convert to MSI parent infrastructure
PCI: apple: Convert to MSI parent infrastructure
irqchip/msi-lib: Honour the MSI_FLAG_NO_AFFINITY flag
irqchip/mvebu: Convert to msi_create_parent_irq_domain() helper
irqchip/gic: Convert to msi_create_parent_irq_domain() helper
genirq/msi: Add helper for creating MSI-parent irq domains
irqchip: Make irq-msi-lib.h globally available
irqchip/gic-v3-its: Use allocation size from the prepare call
genirq/msi: Engage the .msi_teardown() callback on domain removal
genirq/msi: Move prepare() call to per-device allocation
irqchip/gic-v3-its: Implement .msi_teardown() callback
genirq/msi: Add .msi_teardown() callback as the reverse of .msi_prepare()
irqchip/gic-v3-its: Add support for device tree msi-map and msi-mask
dt-bindings: PCI: pci-ep: Add support for iommu-map and msi-map
irqchip/gic-v3-its: Set IRQ_DOMAIN_FLAG_MSI_IMMUTABLE for ITS
irqdomain: Add IRQ_DOMAIN_FLAG_MSI_IMMUTABLE and irq_domain_is_msi_immutable()
platform-msi: Add msi_remove_device_irq_domain() in platform_device_msi_free_irqs_all()
genirq/msi: Rename msi_[un]lock_descs()
...
Commit 7d5ec3d361 ("PCI/MSI: Mask all unused MSI-X entries") introduced a
readl() from ENTRY_VECTOR_CTRL before the writel() to ENTRY_DATA.
This is correct, however some hardware, like the Sun Neptune chips, the NIU
module, will cause an error and/or fatal trap if any MSIX table entry is
read before the corresponding ENTRY_DATA field is written to.
Add an optional early writel() in msix_prepare_msi_desc().
Fixes: 7d5ec3d361 ("PCI/MSI: Mask all unused MSI-X entries")
Signed-off-by: Jonathan Currier <dullfire@yahoo.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/all/20241117234843.19236-2-dullfire@yahoo.com
The PCI/TPH driver fiddles with the MSI-X control word of an active
interrupt completely unserialized against concurrent operations issued
from the interrupt core. It also brings the PCI/MSI-X internal cached
control word out of sync.
Provide a function, which has the required serialization and keeps the
control word cache in sync.
Unfortunately this requires to look up and lock the interrupt descriptor,
which should be only done in the interrupt core code. But confining this
particular oddity in the PCI/MSI core is the lesser of all evil. A
interrupt core implementation would require a larger pile of infrastructure
and indirections for dubious value.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Link: https://lore.kernel.org/all/20250319105506.683663807@linutronix.de
Split the lock protected functionality of msix_capability_init() out into a
helper function and use guard(msi_desc_lock) to replace the lock/unlock
pair.
Simplify the error path in the helper function by utilizing a custom
cleanup to get rid of the remaining gotos.
No functional change intended.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Link: https://lore.kernel.org/all/20250319105506.564105011@linutronix.de
Pull pci updates from Bjorn Helgaas:
"Enumeration:
- Enable Configuration RRS SV, which makes device readiness visible,
early instead of during child bus scanning (Bjorn Helgaas)
- Log debug messages about reset methods being used (Bjorn Helgaas)
- Avoid reset when it has been disabled via sysfs (Nishanth
Aravamudan)
- Add common pci-ep-bus.yaml schema for exporting several peripherals
of a single PCI function via devicetree (Andrea della Porta)
- Create DT nodes for PCI host bridges to enable loading device tree
overlays to create platform devices for PCI devices that have
several features that require multiple drivers (Herve Codina)
Resource management:
- Enlarge devres table[] to accommodate bridge windows, ROM, IOV
BARs, etc., and validate BAR index in devres interfaces (Philipp
Stanner)
- Fix typo that repeatedly distributed resources to a bridge instead
of iterating over subordinate bridges, which resulted in too little
space to assign some BARs (Kai-Heng Feng)
- Relax bridge window tail sizing for optional resources, e.g., IOV
BARs, to avoid failures when removing and re-adding devices (Ilpo
Järvinen)
- Allow drivers to enable devices even if we haven't assigned
optional IOV resources to them (Ilpo Järvinen)
- Rework handling of optional resources (IOV BARs, ROMs) to reduce
failures if we can't allocate them (Ilpo Järvinen)
- Fix a NULL dereference in the SR-IOV VF creation error path (Shay
Drory)
- Fix s390 mmio_read/write syscalls, which didn't cause page faults
in some cases, which broke vfio-pci lazy mapping on first access
(Niklas Schnelle)
- Add pdev->non_mappable_bars to replace CONFIG_VFIO_PCI_MMAP, which
was disabled only for s390 (Niklas Schnelle)
- Support mmap of PCI resources on s390 except for ISM devices
(Niklas Schnelle)
ASPM:
- Delay pcie_link_state deallocation to avoid dangling pointers that
cause invalid references during hot-unplug (Daniel Stodden)
Power management:
- Allow PCI bridges to go to D3Hot when suspending on all non-x86
systems (Manivannan Sadhasivam)
Power control:
- Create pwrctrl devices in pci_scan_device() to make it more
symmetric with pci_pwrctrl_unregister() and make pwrctrl devices
for PCI bridges possible (Manivannan Sadhasivam)
- Unregister pwrctrl devices in pci_destroy_dev() so DOE, ASPM, etc.
can still access devices after pci_stop_dev() (Manivannan
Sadhasivam)
- If there's a pwrctrl device for a PCI device, skip scanning it
because the pwrctrl core will rescan the bus after the device is
powered on (Manivannan Sadhasivam)
- Add a pwrctrl driver for PCI slots based on voltage regulators
described via devicetree (Manivannan Sadhasivam)
Bandwidth control:
- Add set_pcie_speed.sh to TEST_PROGS to fix issue when executing the
set_pcie_cooling_state.sh test case (Yi Lai)
- Avoid a NULL pointer dereference when we run out of bus numbers to
assign for a bridge secondary bus (Lukas Wunner)
Hotplug:
- Drop superfluous pci_hotplug_slot_list, try_module_get() calls, and
NULL pointer checks (Lukas Wunner)
- Drop shpchp module init/exit logging, replace shpchp dbg() with
ctrl_dbg(), and remove unused dbg(), err(), info(), warn() wrappers
(Ilpo Järvinen)
- Drop 'shpchp_debug' module parameter in favor of standard dynamic
debugging (Ilpo Järvinen)
- Drop unused cpcihp .get_power(), .set_power() function pointers
(Guilherme Giacomo Simoes)
- Disable hotplug interrupts in portdrv only when pciehp is not
enabled to avoid issuing two hotplug commands too close together
(Feng Tang)
- Skip pciehp 'device replaced' check if the device has been removed
to address a deadlock when resuming after a device was removed
during system sleep (Lukas Wunner)
- Don't enable pciehp hotplug interupt when resuming in poll mode
(Ilpo Järvinen)
Virtualization:
- Fix bugs in 'pci=config_acs=' kernel command line parameter (Tushar
Dave)
DOE:
- Expose supported DOE features via sysfs (Alistair Francis)
- Allow DOE support to be enabled even if CXL isn't enabled (Alistair
Francis)
Endpoint framework:
- Convert PCI device data so pci-epf-test works correctly on
big-endian endpoint systems (Niklas Cassel)
- Add BAR_RESIZABLE type to endpoint framework and add DWC core
support for EPF drivers to set BAR_RESIZABLE type and size (Niklas
Cassel)
- Fix pci-epf-test double free that causes an oops if the host
reboots and PERST# deassertion restarts endpoint BAR allocation
(Christian Bruel)
- Fix endpoint BAR testing so tests can skip disabled BARs instead of
reporting them as failures (Niklas Cassel)
- Widen endpoint test BAR size variable to accommodate BARs larger
than INT_MAX (Niklas Cassel)
- Remove unused tools 'pci' build target left over after moving tests
to tools/testing/selftests/pci_endpoint (Jianfeng Liu)
Altera PCIe controller driver:
- Add DT binding and driver support for Agilex family (P-Tile,
F-Tile, R-Tile) (Matthew Gerlach and D M, Sharath Kumar)
AMD MDB PCIe controller driver:
- Add DT binding and driver for AMD MDB (Multimedia DMA Bridge)
(Thippeswamy Havalige)
Broadcom STB PCIe controller driver:
- Add BCM2712 MSI-X DT binding and interrupt controller drivers and
add softdep on irq_bcm2712_mip driver to ensure that it is loaded
first (Stanimir Varbanov)
- Expand inbound window map to 64GB so it can accommodate BCM2712
(Stanimir Varbanov)
- Add BCM2712 support and DT updates (Stanimir Varbanov)
- Apply link speed restriction before bringing link up, not after
(Jim Quinlan)
- Update Max Link Speed in Link Capabilities via the internal
writable register, not the read-only config register (Jim Quinlan)
- Handle regulator_bulk_get() error to avoid panic when we call
regulator_bulk_free() later (Jim Quinlan)
- Disable regulators only when removing the bus immediately below a
Root Port because we don't support regulators deeper in the
hierarchy (Jim Quinlan)
- Make const read-only arrays static (Colin Ian King)
Cadence PCIe endpoint driver:
- Correct MSG TLP generation so endpoints can generate INTx messages
(Hans Zhang)
Freescale i.MX6 PCIe controller driver:
- Identify the second controller on i.MX8MQ based on devicetree
'linux,pci-domain' instead of DBI 'reg' address (Richard Zhu)
- Remove imx_pcie_cpu_addr_fixup() since dwc core can now derive the
ATU input address (using parent_bus_offset) from devicetree (Frank
Li)
Freescale Layerscape PCIe controller driver:
- Drop deprecated 'num-ib-windows' and 'num-ob-windows' and
unnecessary 'status' from example (Krzysztof Kozlowski)
- Correct the syscon_regmap_lookup_by_phandle_args("fsl,pcie-scfg")
arg_count to fix probe failure on LS1043A (Ioana Ciornei)
HiSilicon STB PCIe controller driver:
- Call phy_exit() to clean up if histb_pcie_probe() fails (Christophe
JAILLET)
Intel Gateway PCIe controller driver:
- Remove intel_pcie_cpu_addr() since dwc core can now derive the ATU
input address (using parent_bus_offset) from devicetree (Frank Li)
Intel VMD host bridge driver:
- Convert vmd_dev.cfg_lock from spinlock_t to raw_spinlock_t so
pci_ops.read() will never sleep, even on PREEMPT_RT where
spinlock_t becomes a sleepable lock, to avoid calling a sleeping
function from invalid context (Ryo Takakura)
MediaTek PCIe Gen3 controller driver:
- Remove leftover mac_reset assert for Airoha EN7581 SoC (Lorenzo
Bianconi)
- Add EN7581 PBUS controller 'mediatek,pbus-csr' DT property and
program host bridge memory aperture to this syscon node (Lorenzo
Bianconi)
Qualcomm PCIe controller driver:
- Add qcom,pcie-ipq5332 binding (Varadarajan Narayanan)
- Add qcom i.MX8QM and i.MX8QXP/DXP optional DMA interrupt (Alexander
Stein)
- Add optional dma-coherent DT property for Qualcomm SA8775P (Dmitry
Baryshkov)
- Make DT iommu property required for SA8775P and prohibited for
SDX55 (Dmitry Baryshkov)
- Add DT IOMMU and DMA-related properties for Qualcomm SM8450 (Dmitry
Baryshkov)
- Add endpoint DT properties for SAR2130P and enable endpoint mode in
driver (Dmitry Baryshkov)
- Describe endpoint BAR0 and BAR2 as 64-bit only and BAR1 and BAR3 as
RESERVED (Manivannan Sadhasivam)
Rockchip DesignWare PCIe controller driver:
- Describe rk3568 and rk3588 BARs as Resizable, not Fixed (Niklas
Cassel)
Synopsys DesignWare PCIe controller driver:
- Add debugfs-based Silicon Debug, Error Injection, Statistical
Counter support for DWC (Shradha Todi)
- Add debugfs property to expose LTSSM status of DWC PCIe link (Hans
Zhang)
- Add Rockchip support for DWC debugfs features (Niklas Cassel)
- Add dw_pcie_parent_bus_offset() to look up the parent bus address
of a specified 'reg' property and return the offset from the CPU
physical address (Frank Li)
- Use dw_pcie_parent_bus_offset() to derive CPU -> ATU addr offset
via 'reg[config]' for host controllers and 'reg[addr_space]' for
endpoint controllers (Frank Li)
- Apply struct dw_pcie.parent_bus_offset in ATU users to remove use
of .cpu_addr_fixup() when programming ATU (Frank Li)
TI J721E PCIe driver:
- Correct the 'link down' interrupt bit for J784S4 (Siddharth
Vadapalli)
TI Keystone PCIe controller driver:
- Describe AM65x BARs 2 and 5 as Resizable (not Fixed) and reduce
alignment requirement from 1MB to 64KB (Niklas Cassel)
Xilinx Versal CPM PCIe controller driver:
- Free IRQ domain in probe error path to avoid leaking it
(Thippeswamy Havalige)
- Add DT .compatible "xlnx,versal-cpm5nc-host" and driver support for
Versal Net CPM5NC Root Port controller (Thippeswamy Havalige)
- Add driver support for CPM5_HOST1 (Thippeswamy Havalige)
Miscellaneous:
- Convert fsl,mpc83xx-pcie binding to YAML (J. Neuschäfer)
- Use for_each_available_child_of_node_scoped() to simplify apple,
kirin, mediatek, mt7621, tegra drivers (Zhang Zekun)"
* tag 'pci-v6.15-changes' of git://git.kernel.org/pub/scm/linux/kernel/git/pci/pci: (197 commits)
PCI: layerscape: Fix arg_count to syscon_regmap_lookup_by_phandle_args()
PCI: j721e: Fix the value of .linkdown_irq_regfield for J784S4
misc: pci_endpoint_test: Add support for PCITEST_IRQ_TYPE_AUTO
PCI: endpoint: pci-epf-test: Expose supported IRQ types in CAPS register
PCI: dw-rockchip: Endpoint mode cannot raise INTx interrupts
PCI: endpoint: Add intx_capable to epc_features struct
dt-bindings: PCI: Add common schema for devices accessible through PCI BARs
PCI: intel-gw: Remove intel_pcie_cpu_addr()
PCI: imx6: Remove imx_pcie_cpu_addr_fixup()
PCI: dwc: Use parent_bus_offset to remove need for .cpu_addr_fixup()
PCI: dwc: ep: Ensure proper iteration over outbound map windows
PCI: dwc: ep: Use devicetree 'reg[addr_space]' to derive CPU -> ATU addr offset
PCI: dwc: ep: Consolidate devicetree handling in dw_pcie_ep_get_resources()
PCI: dwc: ep: Call epc_create() early in dw_pcie_ep_init()
PCI: dwc: Use devicetree 'reg[config]' to derive CPU -> ATU addr offset
PCI: dwc: Add dw_pcie_parent_bus_offset() checking and debug
PCI: dwc: Add dw_pcie_parent_bus_offset()
PCI/bwctrl: Fix NULL pointer dereference on bus number exhaustion
PCI: xilinx-cpm: Add cpm_csr register mapping for CPM5_HOST1 variant
PCI: brcmstb: Make const read-only arrays static
...
This reverts commit 36f5f026df, reversing
changes made to 43a7eec035.
Thomas says:
"I just noticed that for some incomprehensible reason, probably sheer
incompetemce when trying to utilize b4, I managed to merge an outdated
_and_ buggy version of that series.
Can you please revert that merge completely?"
Done.
Requested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Pull xen updates from Juergen Gross:
- cleanup: remove an used function
- add support for a XenServer specific virtual PCI device
- fix the handling of a sparse Xen hypervisor symbol table
- avoid warnings when building the kernel with gcc 15
- fix use of devices behind a VMD bridge when running as a Xen PV dom0
* tag 'for-linus-6.15-rc1-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip:
PCI/MSI: Convert pci_msi_ignore_mask to per MSI domain flag
PCI: vmd: Disable MSI remapping bypass under Xen
xen/pci: Do not register devices with segments >= 0x10000
xen/pciback: Remove unused pcistub_get_pci_dev
xenfs/xensyms: respect hypervisor's "next" indication
xen/mcelog: Add __nonstring annotations for unterminated strings
xen: Add support for XenServer 6.1 platform device
Setting pci_msi_ignore_mask inhibits the toggling of the mask bit for both
MSI and MSI-X entries globally, regardless of the IRQ chip they are using.
Only Xen sets the pci_msi_ignore_mask when routing physical interrupts over
event channels, to prevent PCI code from attempting to toggle the maskbit,
as it's Xen that controls the bit.
However, the pci_msi_ignore_mask being global will affect devices that use
MSI interrupts but are not routing those interrupts over event channels
(not using the Xen pIRQ chip). One example is devices behind a VMD PCI
bridge. In that scenario the VMD bridge configures MSI(-X) using the
normal IRQ chip (the pIRQ one in the Xen case), and devices behind the
bridge configure the MSI entries using indexes into the VMD bridge MSI
table. The VMD bridge then demultiplexes such interrupts and delivers to
the destination device(s). Having pci_msi_ignore_mask set in that scenario
prevents (un)masking of MSI entries for devices behind the VMD bridge.
Move the signaling of no entry masking into the MSI domain flags, as that
allows setting it on a per-domain basis. Set it for the Xen MSI domain
that uses the pIRQ chip, while leaving it unset for the rest of the
cases.
Remove pci_msi_ignore_mask at once, since it was only used by Xen code, and
with Xen dropping usage the variable is unneeded.
This fixes using devices behind a VMD bridge on Xen PV hardware domains.
Albeit Devices behind a VMD bridge are not known to Xen, that doesn't mean
Linux cannot use them. By inhibiting the usage of
VMD_FEAT_CAN_BYPASS_MSI_REMAP and the removal of the pci_msi_ignore_mask
bodge devices behind a VMD bridge do work fine when use from a Linux Xen
hardware domain. That's the whole point of the series.
Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Juergen Gross <jgross@suse.com>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Message-ID: <20250219092059.90850-4-roger.pau@citrix.com>
Signed-off-by: Juergen Gross <jgross@suse.com>
The PCI/TPH driver fiddles with the MSI-X control word of an active
interrupt completely unserialized against concurrent operations issued
from the interrupt core. It also brings the PCI/MSI-X internal cached
control word out of sync.
Provide a function, which has the required serialization and keeps the
control word cache in sync.
Unfortunately this requires to look up and lock the interrupt descriptor,
which should be only done in the interrupt core code. But confining this
particular oddity in the PCI/MSI core is the lesser of all evil. A
interrupt core implementation would require a larger pile of infrastructure
and indirections for dubious value.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Link: https://lore.kernel.org/all/20250313130321.822790423@linutronix.de
Alexandre observed a warning emitted from pci_msi_setup_msi_irqs() on a
RISCV platform which does not provide PCI/MSI support:
WARNING: CPU: 1 PID: 1 at drivers/pci/msi/msi.h:121 pci_msi_setup_msi_irqs+0x2c/0x32
__pci_enable_msix_range+0x30c/0x596
pci_msi_setup_msi_irqs+0x2c/0x32
pci_alloc_irq_vectors_affinity+0xb8/0xe2
RISCV uses hierarchical interrupt domains and correctly does not implement
the legacy fallback. The warning triggers from the legacy fallback stub.
That warning is bogus as the PCI/MSI layer knows whether a PCI/MSI parent
domain is associated with the device or not. There is a check for MSI-X,
which has a legacy assumption. But that legacy fallback assumption is only
valid when legacy support is enabled, but otherwise the check should simply
return -ENOTSUPP.
Loongarch tripped over the same problem and blindly enabled legacy support
without implementing the legacy fallbacks. There are weak implementations
which return an error, so the problem was papered over.
Correct pci_msi_domain_supports() to evaluate the legacy mode and add
the missing supported check into the MSI enable path to complete it.
Fixes: d2a463b297 ("PCI/MSI: Reject multi-MSI early")
Reported-by: Alexandre Ghiti <alexghiti@rivosinc.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Alexandre Ghiti <alexghiti@rivosinc.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/all/87ed2a8ow5.ffs@tglx
Pull MSI interrupt updates from Thomas Gleixner:
"Switch ARM/ARM64 over to the modern per device MSI domains.
This simplifies the handling of platform MSI and wire to MSI
controllers and removes about 500 lines of legacy code.
Aside of that it paves the way for ARM/ARM64 to utilize the dynamic
allocation of PCI/MSI interrupts and to support the upcoming non
standard IMS (Interrupt Message Store) mechanism on PCIe devices"
* tag 'irq-msi-2024-07-22' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (25 commits)
irqchip/gic-v3-its: Correctly fish out the DID for platform MSI
irqchip/gic-v3-its: Correctly honor the RID remapping
genirq/msi: Move msi_device_data to core
genirq/msi: Remove platform MSI leftovers
irqchip/irq-mvebu-icu: Remove platform MSI leftovers
irqchip/irq-mvebu-sei: Switch to MSI parent
irqchip/mvebu-odmi: Switch to parent MSI
irqchip/mvebu-gicp: Switch to MSI parent
irqchip/irq-mvebu-icu: Prepare for real per device MSI
irqchip/imx-mu-msi: Switch to MSI parent
irqchip/gic-v2m: Switch to device MSI
irqchip/gic_v3_mbi: Switch over to parent domain
genirq/msi: Remove platform_msi_create_device_domain()
irqchip/mbigen: Remove platform_msi_create_device_domain() fallback
irqchip/gic-v3-its: Switch platform MSI to MSI parent
irqchip/irq-msi-lib: Prepare for DOMAIN_BUS_WIRED_TO_MSI
irqchip/mbigen: Prepare for real per device MSI
irqchip/irq-msi-lib: Prepare for DEVICE MSI to replace platform MSI
irqchip/gic-v3-its: Provide MSI parent for PCI/MSI[-X]
irqchip/irq-msi-lib: Prepare for PCI MSI/MSIX
...
Most ARM(64) PCI/MSI domains mask and unmask in the parent domain after or
before the PCI mask/unmask operation takes place. So there are more than a
dozen of the same wrapper implementation all over the place.
Don't make the same mistake with the new per device PCI/MSI domains and
provide a new MSI feature flag, which lets the domain implementation
enable this sequence in the PCI/MSI code.
Signed-off-by: Shivamurthy Shastri <shivamurthy.shastri@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Link: https://lore.kernel.org/r/87ed8j34pj.ffs@tglx
KFENCE reports the following UAF:
BUG: KFENCE: use-after-free read in __pci_enable_msi_range+0x2c0/0x488
Use-after-free read at 0x0000000024629571 (in kfence-#12):
__pci_enable_msi_range+0x2c0/0x488
pci_alloc_irq_vectors_affinity+0xec/0x14c
pci_alloc_irq_vectors+0x18/0x28
kfence-#12: 0x0000000008614900-0x00000000e06c228d, size=104, cache=kmalloc-128
allocated by task 81 on cpu 7 at 10.808142s:
__kmem_cache_alloc_node+0x1f0/0x2bc
kmalloc_trace+0x44/0x138
msi_alloc_desc+0x3c/0x9c
msi_domain_insert_msi_desc+0x30/0x78
msi_setup_msi_desc+0x13c/0x184
__pci_enable_msi_range+0x258/0x488
pci_alloc_irq_vectors_affinity+0xec/0x14c
pci_alloc_irq_vectors+0x18/0x28
freed by task 81 on cpu 7 at 10.811436s:
msi_domain_free_descs+0xd4/0x10c
msi_domain_free_locked.part.0+0xc0/0x1d8
msi_domain_alloc_irqs_all_locked+0xb4/0xbc
pci_msi_setup_msi_irqs+0x30/0x4c
__pci_enable_msi_range+0x2a8/0x488
pci_alloc_irq_vectors_affinity+0xec/0x14c
pci_alloc_irq_vectors+0x18/0x28
Descriptor allocation done in:
__pci_enable_msi_range
msi_capability_init
msi_setup_msi_desc
msi_insert_msi_desc
msi_domain_insert_msi_desc
msi_alloc_desc
...
Freed in case of failure in __msi_domain_alloc_locked()
__pci_enable_msi_range
msi_capability_init
pci_msi_setup_msi_irqs
msi_domain_alloc_irqs_all_locked
msi_domain_alloc_locked
__msi_domain_alloc_locked => fails
msi_domain_free_locked
...
That failure propagates back to pci_msi_setup_msi_irqs() in
msi_capability_init() which accesses the descriptor for unmasking in the
error exit path.
Cure it by copying the descriptor and using the copy for the error exit path
unmask operation.
[ tglx: Massaged change log ]
Fixes: bf6e054e0e ("genirq/msi: Provide msi_device_populate/destroy_sysfs()")
Suggested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Mostafa Saleh <smostafa@google.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Bjorn Heelgas <bhelgaas@google.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20240624203729.1094506-1-smostafa@google.com
- Update coding style to "mainline is normal path, errors are the
exceptions" (Andy Shevchenko)
* pci/msi:
PCI/MSI: Make error path handling follow the standard pattern
While calculating the hardware interrupt number for a MSI interrupt, the
higher bits (i.e. from bit-5 onwards a.k.a domain_nr >= 32) of the PCI
domain number gets truncated because of the shifted value casting to return
type of pci_domain_nr() which is 'int'. This for example is resulting in
same hardware interrupt number for devices 0019:00:00.0 and 0039:00:00.0.
To address this cast the PCI domain number to 'irq_hw_number_t' before left
shifting it to calculate the hardware interrupt number.
Please note that this fixes the issue only on 64-bit systems and doesn't
change the behavior for 32-bit systems i.e. the 32-bit systems continue to
have the issue. Since the issue surfaces only if there are too many PCIe
controllers in the system which usually is the case in modern server
systems and they don't tend to run 32-bit kernels.
Fixes: 3878eaefb8 ("PCI/MSI: Enhance core to support hierarchy irqdomain")
Signed-off-by: Vidya Sagar <vidyas@nvidia.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Shanker Donthineni <sdonthineni@nvidia.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20240115135649.708536-1-vidyas@nvidia.com
pci_msix_validate_entries() validates the entries array which is handed in
by the caller for a MSI-X interrupt allocation. Aside of consistency
failures it also detects a failure when the size of the MSI-X hardware table
in the device is smaller than the size of the entries array.
That's wrong for the case of range allocations where the caller provides
the minimum and the maximum number of vectors to allocate, when the
hardware size is greater or equal than the mininum, but smaller than the
maximum.
Remove the hardware size check completely from that function and just
ensure that the entires array up to the maximum size is consistent.
The limitation and range checking versus the hardware size happens
independently of that afterwards anyway because the entries array is
optional.
Fixes: 4644d22eb6 ("PCI/MSI: Validate MSI-X contiguous restriction early")
Reported-by: David Laight <David.Laight@aculab.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/87v8i3sg62.ffs@tglx
pci_msix_free_irq() is used to free an interrupt on a PCI/MSI-X interrupt
domain.
The API description specifies that the interrupt to be freed was allocated
via pci_msix_alloc_irq_at(). This description limits the usage of
pci_msix_free_irq() since pci_msix_free_irq() can also be used to free
MSI-X interrupts allocated with, for example, pci_alloc_irq_vectors().
Remove the text stating that the interrupt to be freed had to be allocated
with pci_msix_alloc_irq_at(). The needed struct msi_map need not be from
pci_msix_alloc_irq_at() but can be created from scratch using
pci_irq_vector() to obtain the Linux IRQ number. Highlight that
pci_msix_free_irq() cannot be used to disable MSI-X to guide users that,
for example, pci_free_irq_vectors() remains to be needed.
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Kevin Tian <kevin.tian@intel.com>
Link: https://lore.kernel.org/lkml/87r0xsd8j4.ffs@tglx
Link: https://lore.kernel.org/r/4c3e7a50d6e70f408812cd7ab199c6b4b326f9de.1676408572.git.reinette.chatre@intel.com
IMS (Interrupt Message Store) is a new specification which allows
implementation specific storage of MSI messages contrary to the
strict standard specified MSI and MSI-X message stores.
This requires new device specific interrupt domains to handle the
implementation defined storage which can be an array in device memory or
host/guest memory which is shared with hardware queues.
Add a function to create IMS domains for PCI devices. IMS domains are using
the new per device domain mechanism and are configured by the device driver
via a template. IMS domains are created as secondary device domains so they
work side on side with MSI[-X] on the same device.
The IMS domains have a few constraints:
- The index space is managed by the core code.
Device memory based IMS provides a storage array with a fixed size
which obviously requires an index. But there is no association between
index and functionality so the core can randomly allocate an index in
the array.
System memory based IMS does not have the concept of an index as the
storage is somewhere in memory. In that case the index is purely
software based to keep track of the allocations.
- There is no requirement for consecutive index ranges
This is currently a limitation of the MSI core and can be implemented
if there is a justified use case by changing the internal storage from
xarray to maple_tree. For now it's single vector allocation.
- The interrupt chip must provide the following callbacks:
- irq_mask()
- irq_unmask()
- irq_write_msi_msg()
- The interrupt chip must provide the following optional callbacks
when the irq_mask(), irq_unmask() and irq_write_msi_msg() callbacks
cannot operate directly on hardware, e.g. in the case that the
interrupt message store is in queue memory:
- irq_bus_lock()
- irq_bus_unlock()
These callbacks are invoked from preemptible task context and are
allowed to sleep. In this case the mandatory callbacks above just
store the information. The irq_bus_unlock() callback is supposed to
make the change effective before returning.
- Interrupt affinity setting is handled by the underlying parent
interrupt domain and communicated to the IMS domain via
irq_write_msi_msg(). IMS domains cannot have a irq_set_affinity()
callback. That's a reasonable restriction similar to the PCI/MSI
device domain implementations.
The domain is automatically destroyed when the PCI device is removed.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Kevin Tian <kevin.tian@intel.com>
Acked-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20221124232326.904316841@linutronix.de
MSI-X vectors can be allocated after the initial MSI-X enablement, but this
needs explicit support of the underlying interrupt domains.
Provide a function to query the ability and functions to allocate/free
individual vectors post-enable.
The allocation can either request a specific index in the MSI-X table or
with the index argument MSI_ANY_INDEX it allocates the next free vector.
The return value is a struct msi_map which on success contains both index
and the Linux interrupt number. In case of failure index is negative and
the Linux interrupt number is 0.
The allocation function is for a single MSI-X index at a time as that's
sufficient for the most urgent use case VFIO to get rid of the 'disable
MSI-X, reallocate, enable-MSI-X' cycle which is prone to lost interrupts
and redirections to the legacy and obviously unhandled INTx.
As single index allocation is also sufficient for the use cases Jason
Gunthorpe pointed out: Allocation of a MSI-X or IMS vector for a network
queue. See Link below.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Kevin Tian <kevin.tian@intel.com>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Acked-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/all/20211126232735.547996838@linutronix.de
Link: https://lore.kernel.org/r/20221124232326.731233614@linutronix.de
The setup of MSI descriptors for PCI/MSI-X interrupts depends partially on
the MSI index for which the descriptor is initialized.
Dynamic MSI-X vector allocation post MSI-X enablement allows to allocate
vectors at a given index or at any free index in the available table
range. The latter requires that the descriptor is initialized after the
MSI core has chosen an index.
Implement the prepare_desc() op in the PCI/MSI-X specific msi_domain_ops
which is invoked before the core interrupt descriptor and the associated
Linux interrupt number is allocated.
That callback is also provided for the upcoming PCI/IMS implementations so
the implementation specific interrupt domain can do their domain specific
initialization of the MSI descriptors.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Reviewed-by: Kevin Tian <kevin.tian@intel.com>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Acked-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20221124232326.673658806@linutronix.de
The upcoming mechanism to allocate MSI-X vectors after enabling MSI-X needs
to share some of the MSI-X descriptor setup.
The regular descriptor setup on enable has the following code flow:
1) Allocate descriptor
2) Setup descriptor with PCI specific data
3) Insert descriptor
4) Allocate interrupts which in turn scans the inserted
descriptors
This cannot be easily changed because the PCI/MSI code needs to handle the
legacy architecture specific allocation model and the irq domain model
where quite some domains have the assumption that the above flow is how it
works.
Ideally the code flow should look like this:
1) Invoke allocation at the MSI core
2) MSI core allocates descriptor
3) MSI core calls back into the irq domain which fills in
the domain specific parts
This could be done for underlying parent MSI domains which support
post-enable allocation/free but that would create significantly different
code pathes for MSI/MSI-X enable.
Though for dynamic allocation which wants to share the allocation code with
the upcoming PCI/IMS support it's the right thing to do.
Split the MSI-X descriptor setup into the preallocation part which just sets
the index and fills in the horrible hack of virtual IRQs and the real PCI
specific MSI-X setup part which solely depends on the index in the
descriptor. This allows to provide a common dynamic allocation interface at
the MSI core level for both PCI/MSI-X and PCI/IMS.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Kevin Tian <kevin.tian@intel.com>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Acked-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20221124232326.616292598@linutronix.de
Provide a template and the necessary callbacks to create PCI/MSI and
PCI/MSI-X domains.
The domains are created when MSI or MSI-X is enabled. The domain's lifetime
is either the device lifetime or in case that e.g. MSI-X was tried first
and failed, then the MSI-X domain is removed and a MSI domain is created as
both are mutually exclusive and reside in the default domain ID slot of the
per device domain pointer array.
Also expand pci_msi_domain_supports() to handle feature checks correctly
even in the case that the per device domain was not yet created by checking
the features supported by the MSI parent.
Add the necessary setup calls into the MSI and MSI-X enable code path.
These setup calls are backwards compatible. They return success when there
is no parent domain found, which means the existing global domains or the
legacy allocation path keep just working.
Co-developed-by: Ahmed S. Darwish <darwi@linutronix.de>
Signed-off-by: Ahmed S. Darwish <darwi@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Kevin Tian <kevin.tian@intel.com>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Acked-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20221124232325.975388241@linutronix.de
With interrupt domains the sanity check for MSI-X vector validation can be
done _before_ any allocation happens. The sanity check only applies to the
allocation functions which have an 'entries' array argument. The entries
array is filled by the caller with the requested MSI-X indices. Some drivers
have gaps in the index space which is not supported on all architectures.
The PCI/MSI irq domain has a 'feature' bit to enforce this validation late
during the allocation phase.
Just do it right away before doing any other work along with the other
sanity checks on that array.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Link: https://lore.kernel.org/r/20221111122015.691357406@linutronix.de
When hierarchical MSI interrupt domains are enabled then there is no point
to do tons of work and detect the missing support for multi-MSI late in the
allocation path.
Just query the domain feature flags right away. The query function is going
to be used for other purposes later and has a mode argument which influences
the result:
ALLOW_LEGACY returns true when:
- there is no irq domain attached (legacy support)
- there is a irq domain attached which has the feature flag set
DENY_LEGACY returns only true when:
- there is a irq domain attached which has the feature flag set
This allows to use the function universally without ifdeffery in the
calling code.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Link: https://lore.kernel.org/r/20221111122015.574339988@linutronix.de
There is no way to navigate msi.c without banging the head against the wall
every now and then because MSI and MSI-X specific functions are
intermingled and the code flow is completely non-obvious.
Reorder everthing so common helpers, MSI and MSI-X specific functions are
grouped together.
Suggested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ahmed S. Darwish <darwi@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Link: https://lore.kernel.org/r/20221111122015.459089736@linutronix.de
