ea4497337f
[ Upstream commit 7ffb791423 ]
When CONFIG_PCI_P2PDMA=y (which is basically enabled on all
large x86 distros), it maps the PFN's via a ZONE_DEVICE
mapping using devm_memremap_pages(). The mapped virtual
address range corresponds to the pci_resource_start()
of the BAR address and size corresponding to the BAR length.
When KASLR is enabled, the direct map range of the kernel is
reduced to the size of physical memory plus additional padding.
If the BAR address is beyond this limit, PCI peer to peer DMA
mappings fail.
Fix this by not shrinking the size of the direct map when
CONFIG_PCI_P2PDMA=y.
This reduces the total available entropy, but it's better than
the current work around of having to disable KASLR completely.
[ mingo: Clarified the changelog to point out the broad impact ... ]
Signed-off-by: Balbir Singh <balbirs@nvidia.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Kees Cook <kees@kernel.org>
Acked-by: Bjorn Helgaas <bhelgaas@google.com> # drivers/pci/Kconfig
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Link: https://lore.kernel.org/lkml/20250206023201.1481957-1-balbirs@nvidia.com/
Link: https://lore.kernel.org/r/20250206234234.1912585-1-balbirs@nvidia.com
--
arch/x86/mm/kaslr.c | 10 ++++++++--
drivers/pci/Kconfig | 6 ++++++
2 files changed, 14 insertions(+), 2 deletions(-)
Signed-off-by: Sasha Levin <sashal@kernel.org>
8.2 KiB
SPDX-License-Identifier: GPL-2.0
PCI configuration
select this to offer the PCI prompt
config HAVE_PCI bool
select this to unconditionally force on PCI support
config FORCE_PCI bool select HAVE_PCI select PCI
menuconfig PCI bool "PCI support" depends on HAVE_PCI help This option enables support for the PCI local bus, including support for PCI-X and the foundations for PCI Express support. Say 'Y' here unless you know what you are doing.
if PCI
config PCI_DOMAINS bool depends on PCI
config PCI_DOMAINS_GENERIC bool select PCI_DOMAINS
config PCI_SYSCALL bool
source "drivers/pci/pcie/Kconfig"
config PCI_MSI bool "Message Signaled Interrupts (MSI and MSI-X)" select GENERIC_MSI_IRQ help This allows device drivers to enable MSI (Message Signaled Interrupts). Message Signaled Interrupts enable a device to generate an interrupt using an inbound Memory Write on its PCI bus instead of asserting a device IRQ pin.
Use of PCI MSI interrupts can be disabled at kernel boot time
by using the 'pci=nomsi' option. This disables MSI for the
entire system.
If you don't know what to do here, say Y.
config PCI_MSI_IRQ_DOMAIN def_bool y depends on PCI_MSI select GENERIC_MSI_IRQ_DOMAIN
config PCI_MSI_ARCH_FALLBACKS bool
config PCI_QUIRKS default y bool "Enable PCI quirk workarounds" if EXPERT help This enables workarounds for various PCI chipset bugs/quirks. Disable this only if your target machine is unaffected by PCI quirks.
config PCI_DEBUG bool "PCI Debugging" depends on DEBUG_KERNEL help Say Y here if you want the PCI core to produce a bunch of debug messages to the system log. Select this if you are having a problem with PCI support and want to see more of what is going on.
When in doubt, say N.
config PCI_REALLOC_ENABLE_AUTO bool "Enable PCI resource re-allocation detection" depends on PCI_IOV help Say Y here if you want the PCI core to detect if PCI resource re-allocation needs to be enabled. You can always use pci=realloc=on or pci=realloc=off to override it. It will automatically re-allocate PCI resources if SR-IOV BARs have not been allocated by the BIOS.
When in doubt, say N.
config PCI_STUB tristate "PCI Stub driver" help Say Y or M here if you want be able to reserve a PCI device when it is going to be assigned to a guest operating system.
When in doubt, say N.
config PCI_PF_STUB tristate "PCI PF Stub driver" depends on PCI_IOV help Say Y or M here if you want to enable support for devices that require SR-IOV support, while at the same time the PF (Physical Function) itself is not providing any actual services on the host itself such as storage or networking.
When in doubt, say N.
config XEN_PCIDEV_FRONTEND tristate "Xen PCI Frontend" depends on XEN_PV select PCI_XEN select XEN_XENBUS_FRONTEND default y help The PCI device frontend driver allows the kernel to import arbitrary PCI devices from a PCI backend to support PCI driver domains.
config PCI_ATS bool
config PCI_DOE bool
config PCI_ECAM bool
config PCI_LOCKLESS_CONFIG bool
config PCI_BRIDGE_EMUL bool
config PCI_IOV bool "PCI IOV support" select PCI_ATS help I/O Virtualization is a PCI feature supported by some devices which allows them to create virtual devices which share their physical resources.
If unsure, say N.
config PCI_PRI bool "PCI PRI support" select PCI_ATS help PRI is the PCI Page Request Interface. It allows PCI devices that are behind an IOMMU to recover from page faults.
If unsure, say N.
config PCI_PASID bool "PCI PASID support" select PCI_ATS help Process Address Space Identifiers (PASIDs) can be used by PCI devices to access more than one IO address space at the same time. To make use of this feature an IOMMU is required which also supports PASIDs. Select this option if you have such an IOMMU and want to compile the driver for it into your kernel.
If unsure, say N.
config PCI_P2PDMA bool "PCI peer-to-peer transfer support" depends on ZONE_DEVICE # # The need for the scatterlist DMA bus address flag means PCI P2PDMA # requires 64bit # depends on 64BIT select GENERIC_ALLOCATOR help Enableѕ drivers to do PCI peer-to-peer transactions to and from BARs that are exposed in other devices that are the part of the hierarchy where peer-to-peer DMA is guaranteed by the PCI specification to work (ie. anything below a single PCI bridge).
Many PCIe root complexes do not support P2P transactions and
it's hard to tell which support it at all, so at this time,
P2P DMA transactions must be between devices behind the same root
port.
Enabling this option will reduce the entropy of x86 KASLR memory
regions. For example - on a 46 bit system, the entropy goes down
from 16 bits to 15 bits. The actual reduction in entropy depends
on the physical address bits, on processor features, kernel config
(5 level page table) and physical memory present on the system.
If unsure, say N.
config PCI_LABEL def_bool y if (DMI || ACPI) select NLS
config PCI_HYPERV tristate "Hyper-V PCI Frontend" depends on ((X86 && X86_64) || ARM64) && HYPERV && PCI_MSI && PCI_MSI_IRQ_DOMAIN && SYSFS select PCI_HYPERV_INTERFACE help The PCI device frontend driver allows the kernel to import arbitrary PCI devices from a PCI backend to support PCI driver domains.
choice prompt "PCI Express hierarchy optimization setting" default PCIE_BUS_DEFAULT depends on PCI && EXPERT help MPS (Max Payload Size) and MRRS (Max Read Request Size) are PCIe device parameters that affect performance and the ability to support hotplug and peer-to-peer DMA.
The following choices set the MPS and MRRS optimization strategy
at compile-time. The choices are the same as those offered for
the kernel command-line parameter 'pci', i.e.,
'pci=pcie_bus_tune_off', 'pci=pcie_bus_safe',
'pci=pcie_bus_perf', and 'pci=pcie_bus_peer2peer'.
This is a compile-time setting and can be overridden by the above
command-line parameters. If unsure, choose PCIE_BUS_DEFAULT.
config PCIE_BUS_TUNE_OFF bool "Tune Off" depends on PCI help Use the BIOS defaults; don't touch MPS at all. This is the same as booting with 'pci=pcie_bus_tune_off'.
config PCIE_BUS_DEFAULT bool "Default" depends on PCI help Default choice; ensure that the MPS matches upstream bridge.
config PCIE_BUS_SAFE bool "Safe" depends on PCI help Use largest MPS that boot-time devices support. If you have a closed system with no possibility of adding new devices, this will use the largest MPS that's supported by all devices. This is the same as booting with 'pci=pcie_bus_safe'.
config PCIE_BUS_PERFORMANCE bool "Performance" depends on PCI help Use MPS and MRRS for best performance. Ensure that a given device's MPS is no larger than its parent MPS, which allows us to keep all switches/bridges to the max MPS supported by their parent. This is the same as booting with 'pci=pcie_bus_perf'.
config PCIE_BUS_PEER2PEER bool "Peer2peer" depends on PCI help Set MPS = 128 for all devices. MPS configuration effected by the other options could cause the MPS on one root port to be different than that of the MPS on another, which may cause hot-added devices or peer-to-peer DMA to fail. Set MPS to the smallest possible value (128B) system-wide to avoid these issues. This is the same as booting with 'pci=pcie_bus_peer2peer'.
endchoice
config VGA_ARB bool "VGA Arbitration" if EXPERT default y depends on (PCI && !S390) help Some "legacy" VGA devices implemented on PCI typically have the same hard-decoded addresses as they did on ISA. When multiple PCI devices are accessed at same time they need some kind of coordination. Please see Documentation/gpu/vgaarbiter.rst for more details. Select this to enable VGA arbiter.
config VGA_ARB_MAX_GPUS int "Maximum number of GPUs" default 16 depends on VGA_ARB help Reserves space in the kernel to maintain resource locking for multiple GPUS. The overhead for each GPU is very small.
source "drivers/pci/hotplug/Kconfig" source "drivers/pci/controller/Kconfig" source "drivers/pci/endpoint/Kconfig" source "drivers/pci/switch/Kconfig"
endif