| Age | Commit message (Collapse) | Author |
|---|
|
- Check IORESOURCE_BUS existence to avoid NULL pointer dereference
(Aleksandr Mishin)
* pci/controller/al:
PCI: al: Check IORESOURCE_BUS existence during probe
|
|
- Use msleep() in DWC core instead of usleep_range() for ~100 ms sleep
(Konrad Dybcio)
- Fix iATU slot management to avoid using the wrong slot after PERST#
assert/deassert, which could potentially cause DMA to go the wrong place
(Frank Li)
- Consolidate dw_pcie_prog_outbound_atu() arguments into a struct to ease
adding new functionality like initiating Message TLPs (Yoshihiro Shimoda)
- Add support for endpoints to initiate PCIe messages (Yoshihiro Shimoda)
- Add #defines for PCIe INTx messages (Yoshihiro Shimoda)
- Add support for endpoints to initiate PCIe PME_Turn_Off messages for
system suspend (Frank Li)
- Add dw_pcie_ep_linkdown() to reinitialize registers that are lost when
the link goes down (Manivannan Sadhasivam)
- Use dw_pcie_ep_linkdown() to reinitialize qcom non-sticky registers that
are lost when the link goes down (Manivannan Sadhasivam)
- Enforce DWC limitation that 64-bit BARs must start with the even numbered
BAR (Niklas Cassel)
* pci/controller/dwc:
PCI: dwc: ep: Enforce DWC specific 64-bit BAR limitation
PCI: layerscape-ep: Use the generic dw_pcie_ep_linkdown() API to handle Link Down event
PCI: qcom-ep: Use the generic dw_pcie_ep_linkdown() API to handle Link Down event
PCI: dwc: ep: Remove dw_pcie_ep_init_notify() wrapper
PCI: dwc: ep: Add a generic dw_pcie_ep_linkdown() API to handle Link Down event
PCI: dwc: Add generic MSG TLP support for sending PME_Turn_Off when system suspend
PCI: Add PCIE_MSG_CODE_PME_TURN_OFF message macro
PCI: Add PCIE_MSG_CODE_ASSERT_INTx message macros
PCI: dwc: Add outbound MSG TLPs support
PCI: dwc: Consolidate args of dw_pcie_prog_outbound_atu() into a structure
PCI: dwc: Fix index 0 incorrectly being interpreted as a free ATU slot
PCI: dwc: Use msleep() in dw_pcie_wait_for_link()
|
|
- Include <linux/irqchip/chained_irq.h> in dra7xx to avoid implicitly
including it elsewhere (Andy Shevchenko)
- Remove unused <linux/of_gpio.h> from aardvark and dwc drivers (dra7xx,
meson, qcom, tegra194) (Andy Shevchenko)
- Convert kirin to use scoped for_each_available_child_of_node() to ease
future error exits (Javier Carrasco)
- Convert imx6 and kirin to use the agnostic GPIO API to simplify GPIO
setup and remove usage of the deprecated of_gpio.h API (Andy Shevchenko)
* pci/controller/gpio:
PCI: kirin: Convert to use agnostic GPIO API
PCI: kirin: Convert kirin_pcie_parse_port() to scoped iterator
PCI: imx6: Convert to use agnostic GPIO API
PCI: dwc: Remove unused of_gpio.h inclusion
PCI: aardvark: Remove unused of_gpio.h inclusion
PCI: dra7xx: Add missing chained IRQ header inclusion
|
|
- Remove unused struct pci_epf_group.type_group (Christophe JAILLET)
- Use cached epc_features instead of pci_epc_get_features() to avoid having
to check for failure (potential NULL pointer dereference) (Manivannan
Sadhasivam)
- Drop pointless local msix_capable variable in pci_epf_test_alloc_space()
(Manivannan Sadhasivam)
- Rename struct pci_epc_event_ops.core_init to .epc_init, since "core" is
no longer meaningful here (Manivannan Sadhasivam)
- Rename pci_epc_bme_notify(), pci_epf_mhi_bme(), pci_epc_bme_notify() to
spell out "bus_master_enable" instead of "bme" (Manivannan Sadhasivam)
- Factor pci_epf_test_clear_bar() and pci_epf_test_free_space() out of
pci_epf_test_unbind() so they can be reused elsewhere (Manivannan
Sadhasivam)
- Move DMA initialization to the pci_epf_mhi_epc_init() callback so
endpoint drivers do this uniformly (Manivannan Sadhasivam)
- Add endpoint testing for Link Down events (Manivannan Sadhasivam)
- Add 'epc_deinit' event so endpoints that can be reset via PERST# (qcom,
tegra194) can notify EPF drivers when this happens (Manivannan
Sadhasivam)
- Make pci_epc_class constant (Greg Kroah-Hartman)
- Fix vpci_scan_bus() error checking to print error for failure (not
success) and clean up after failure (Dan Carpenter)
- Fix epf_ntb_epc_cleanup() error handling to clean up scratchpad BARs and
clean up in mirror order of allocation (Dan Carpenter)
- Add rk3588, which requires 64KB BAR alignment, to pci_endpoint_test
(Niklas Cassel)
- Use memcpy_toio()/memcpy_fromio() for endpoint BAR tests to improve
performance (Niklas Cassel)
- Set DMA mask to 48 bits always to simplify endpoint test, since there's
there's no need to check for error or to fallback to 32 bits (Frank Li)
- Suggest using programmable Vendor/Device ID (when supported) to use
pci_endpoint_test without having to add new entries (Yoshihiro Shimoda)
- Remove unused pci_endpoint_test_bar_{readl,writel}() (Jiapeng Chong)
- Remove 'linkup' and add 'add_cfs' to the endpoint function driver 'ops'
documentation to match the code (Alexander Stein)
-
* pci/endpoint:
Documentation: PCI: pci-endpoint: Fix EPF ops list
misc: pci_endpoint_test: Remove unused pci_endpoint_test_bar_{readl,writel} functions
misc: pci_endpoint_test: Document policy about adding pci_device_id
misc: pci_endpoint_test: Refactor dma_set_mask_and_coherent() logic
misc: pci_endpoint_test: Use memcpy_toio()/memcpy_fromio() for BAR tests
misc: pci_endpoint_test: Add support for Rockchip rk3588
PCI: endpoint: Fix error handling in epf_ntb_epc_cleanup()
PCI: endpoint: Clean up error handling in vpci_scan_bus()
PCI: endpoint: Make pci_epc_class struct constant
PCI: endpoint: Introduce 'epc_deinit' event and notify the EPF drivers
PCI: endpoint: pci-epf-test: Handle Link Down event
PCI: endpoint: pci-epf-{mhi/test}: Move DMA initialization to EPC init callback
PCI: endpoint: pci-epf-test: Refactor pci_epf_test_unbind() function
PCI: endpoint: Rename BME to Bus Master Enable
PCI: endpoint: Rename core_init() callback in 'struct pci_epc_event_ops' to epc_init()
PCI: endpoint: pci-epf-test: Use 'msix_capable' flag directly in pci_epf_test_alloc_space()
PCI: endpoint: pci-epf-test: Make use of cached 'epc_features' in pci_epf_test_core_init()
PCI: endpoint: Remove unused field in struct pci_epf_group
|
|
- Rename find_resource() to find_resource_space() to make it more
descriptive for exporting outside resource.c (Ilpo Järvinen)
- Document find_resource_space() and the resource_constraint struct it uses
(Ilpo Järvinen)
- Add typedef resource_alignf to make it simpler to declare allocation
constraint alignf callbacks (Ilpo Järvinen)
- Open-code the no-constraint simple alignment case to make the
simple_align_resource() default callback unnecessary (Ilpo Järvinen)
- Export find_resource_space() because PCI bridge window allocation needs
to learn whether there's space for a window (Ilpo Järvinen)
- Fix a double-counting problem in PCI calculate_memsize() that led to
allocating larger windows each time a bus was removed and rescanned (Ilpo
Järvinen)
- When we don't have space to allocate larger bridge windows, allocate
windows only large enough for the downstream devices to prevent cases
where a device worked originally, but not after being removed and
re-added (Ilpo Järvinen)
* pci/resource:
PCI: Relax bridge window tail sizing rules
PCI: Make minimum bridge window alignment reference more obvious
PCI: Fix resource double counting on remove & rescan
resource: Export find_resource_space()
resource: Handle simple alignment inside __find_resource_space()
resource: Use typedef for alignf callback
resource: Document find_resource_space() and resource_constraint
resource: Rename find_resource() to find_resource_space()
|
|
- Warn about doing a Secondary Bus Reset without holding the device lock
(Dan Williams)
- Lock bridge in addition to downstream hierarchy before doing a Secondary
Bus Reset (Dan Williams)
* pci/reset:
PCI: Add missing bridge lock to pci_bus_lock()
PCI: Warn on missing cfg_access_lock during secondary bus reset
|
|
- Detect if a device was removed or replaced during system sleep so we
don't assume a new device is the one that used to be there. This uses
Vendor/Device/Subsystem/Class/Revision and Device Serial Number (if
implemented), so it's not fool-proof and drivers may know how to detect
more cases (Lukas Wunner)
- Add missing MODULE_DESCRIPTION() macro (Jeff Johnson)
* pci/hotplug:
PCI: acpiphp: Add missing MODULE_DESCRIPTION() macro
PCI: pciehp: Detect device replacement during system sleep
|
|
- Disable AER and DPC during suspend so that if they share an interrupt
with PME and errors occur during suspend, the AER or DPC interrupt
doesn't cause spurious wakeups (Kai-Heng Feng)
* pci/err:
PCI/DPC: Disable DPC service on suspend
PCI/AER: Disable AER service on suspend
|
|
- Move the PRESERVE_BOOT_CONFIG ACPI _DSM evaluation from drivers/acpi to
drivers/pci so we can unify with similar DT functionality (Vidya Sagar)
- Add of_pci_preserve_config() to check for a DT "linux,pci-probe-only"
property on a per-host bridge basis in addition to a global basis (Vidya
Sagar)
- Unify ACPI PRESERVE_BOOT_CONFIG _DSM and DT "linux,pci-probe-only" in a
generic pci_preserve_config() path (Vidya Sagar)
* pci/enumeration:
PCI: Use preserve_config in place of pci_flags
PCI: Unify ACPI and DT 'preserve config' support
PCI: of: Add of_pci_preserve_config() for per-host bridge support
PCI: Move PRESERVE_BOOT_CONFIG _DSM evaluation to pci_register_host_bridge()
|
|
- If there's a device below a bridge, prevent a use-after-free by holding a
reference to the device while waiting for the secondary bus to be ready
in case the device is concurrently removed, e.g., by DPC (Lukas Wunner)
* pci/dpc:
PCI/DPC: Fix use-after-free on concurrent DPC and hot-removal
|
|
- Add pcim_add_mapping_to_legacy_table() and
pcim_remove_mapping_from_legacy_table() helper functions to simplify
devres iomap table (Philipp Stanner)
- Reimplement devres that take a bit mask of BARs in a way that can be used
to map partial BARs as well as entire BARs (Philipp Stanner)
- Deprecate pcim_iomap_table() and pcim_iomap_regions_request_all() in
favor of pcim_* request plus pcim_* mapping (Philipp Stanner)
- Add pcim_request_region(), a managed interface to request a single BAR
(Philipp Stanner)
- Use the existing pci_is_enabled() interface to replace the struct
devres.enabled bit (Philipp Stanner)
- Move the struct pci_devres.pinned bit to struct pci_dev (Philipp Stanner)
- Reimplement pcim_set_mwi() so it uses its own devres cleanup callback
instead of a special-purpose bit in struct pci_devres (Philipp Stanner)
- Add pcim_intx(), which is unambiguously managed, unlike pci_intx(), which
is managed if pcim_enable_device() has been called but unmanaged
otherwise (Philipp Stanner)
- Remove pcim_release(), which is no longer needed after previous cleanups
of pcim_set_mwi() and pci_intx() (Philipp Stanner)
- Add pcim_iomap_range(), a managed interface to map part of a BAR (Philipp
Stanner)
- Fix vboxvideo leak by using the new pcim_iomap_range() instead of the
unmanaged pci_iomap_range() (Philipp Stanner)
* pci/devres:
drm/vboxvideo: fix mapping leaks
PCI: Add managed pcim_iomap_range()
PCI: Remove legacy pcim_release()
PCI: Add managed pcim_intx()
PCI: Give pcim_set_mwi() its own devres cleanup callback
PCI: Move struct pci_devres.pinned bit to struct pci_dev
PCI: Remove struct pci_devres.enabled status bit
PCI: Document hybrid devres hazards
PCI: Add managed pcim_request_region()
PCI: Deprecate pcim_iomap_table(), pcim_iomap_regions_request_all()
PCI: Add managed partial-BAR request and map infrastructure
PCI: Add devres helpers for iomap table
PCI: Add and use devres helper for bit masks
|
|
PCIe ACS settings control the level of isolation and the possible P2P paths
between devices. With greater isolation the kernel will create smaller
iommu_groups and with less isolation there is more HW that can achieve P2P
transfers. From a virtualization perspective all devices in the same
iommu_group must be assigned to the same VM as they lack security
isolation.
There is no way for the kernel to automatically know the correct ACS
settings for any given system and workload. Existing command line options
(e.g., disable_acs_redir) allow only for large scale change, disabling all
isolation, but this is not sufficient for more complex cases.
Add a kernel command-line option 'config_acs' to directly control all the
ACS bits for specific devices, which allows the operator to setup the right
level of isolation to achieve the desired P2P configuration. The
definition is future proof; when new ACS bits are added to the spec the
open syntax can be extended.
ACS needs to be setup early in the kernel boot as the ACS settings affect
how iommu_groups are formed. iommu_group formation is a one time event
during initial device discovery, so changing ACS bits after kernel boot can
result in an inaccurate view of the iommu_groups compared to the current
isolation configuration.
ACS applies to PCIe Downstream Ports and multi-function devices. The
default ACS settings are strict and deny any direct traffic between two
functions. This results in the smallest iommu_group the HW can support.
Frequently these values result in slow or non-working P2PDMA.
ACS offers a range of security choices controlling how traffic is
allowed to go directly between two devices. Some popular choices:
- Full prevention
- Translated requests can be direct, with various options
- Asymmetric direct traffic, A can reach B but not the reverse
- All traffic can be direct
Along with some other less common ones for special topologies.
The intention is that this option would be used with expert knowledge of
the HW capability and workload to achieve the desired configuration.
Link: https://lore.kernel.org/r/20240625153150.159310-1-vidyas@nvidia.com
Signed-off-by: Vidya Sagar <vidyas@nvidia.com>
[bhelgaas: add example, tidy printk formats]
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
|
|
One of the true positives that the cfg_access_lock lockdep effort
identified is this sequence:
WARNING: CPU: 14 PID: 1 at drivers/pci/pci.c:4886 pci_bridge_secondary_bus_reset+0x5d/0x70
RIP: 0010:pci_bridge_secondary_bus_reset+0x5d/0x70
Call Trace:
<TASK>
? __warn+0x8c/0x190
? pci_bridge_secondary_bus_reset+0x5d/0x70
? report_bug+0x1f8/0x200
? handle_bug+0x3c/0x70
? exc_invalid_op+0x18/0x70
? asm_exc_invalid_op+0x1a/0x20
? pci_bridge_secondary_bus_reset+0x5d/0x70
pci_reset_bus+0x1d8/0x270
vmd_probe+0x778/0xa10
pci_device_probe+0x95/0x120
Where pci_reset_bus() users are triggering unlocked secondary bus resets.
Ironically pci_bus_reset(), several calls down from pci_reset_bus(), uses
pci_bus_lock() before issuing the reset which locks everything *but* the
bridge itself.
For the same motivation as adding:
bridge = pci_upstream_bridge(dev);
if (bridge)
pci_dev_lock(bridge);
to pci_reset_function() for the "bus" and "cxl_bus" reset cases, add
pci_dev_lock() for @bus->self to pci_bus_lock().
Link: https://lore.kernel.org/r/171711747501.1628941.15217746952476635316.stgit@dwillia2-xfh.jf.intel.com
Reported-by: Imre Deak <imre.deak@intel.com>
Closes: http://lore.kernel.org/r/6657833b3b5ae_14984b29437@dwillia2-xfh.jf.intel.com.notmuch
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Keith Busch <kbusch@kernel.org>
[bhelgaas: squash in recursive locking deadlock fix from Keith Busch:
https://lore.kernel.org/r/20240711193650.701834-1-kbusch@meta.com]
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Tested-by: Hans de Goede <hdegoede@redhat.com>
Tested-by: Kalle Valo <kvalo@kernel.org>
Reviewed-by: Dave Jiang <dave.jiang@intel.com>
|
|
The only managed mapping function currently is pcim_iomap() which doesn't
allow for mapping an area starting at a certain offset, which many drivers
want.
Add pcim_iomap_range() as an exported function.
Link: https://lore.kernel.org/r/20240613115032.29098-13-pstanner@redhat.com
Signed-off-by: Philipp Stanner <pstanner@redhat.com>
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
|
|
Thanks to preceding cleanup steps, pcim_release() is now not needed
anymore and can be replaced by pcim_disable_device(), which is the exact
counterpart to pcim_enable_device().
This permits removing further parts of the old PCI devres implementation.
Replace pcim_release() with pcim_disable_device(). Remove the now unused
function get_pci_dr(). Remove the struct pci_devres from pci.h.
Link: https://lore.kernel.org/r/20240613115032.29098-12-pstanner@redhat.com
Signed-off-by: Philipp Stanner <pstanner@redhat.com>
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
|
|
pci_intx() is a "hybrid" function, i.e., it is managed if
pcim_enable_device() has been called, but unmanaged otherwise.
Add pcim_intx(), which is always managed, and implement pci_intx() using
it.
Remove the now-unused struct pci_devres.orig_intx and .restore_intx and
find_pci_dr().
Link: https://lore.kernel.org/r/20240613115032.29098-11-pstanner@redhat.com
Signed-off-by: Philipp Stanner <pstanner@redhat.com>
[kwilczynski: squashed in
https://lore.kernel.org/r/426645d40776198e0fcc942f4a6cac4433c7a9aa.camel@redhat.com
to fix problem reported and tested by Ashish Kalra <Ashish.Kalra@amd.com>:
https://lore.kernel.org/r/20240708214656.4721-1-Ashish.Kalra@amd.com
https://lore.kernel.org/r/8c4634e9-4f02-4c54-9c89-d75e2f4bf026@amd.com/]
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
[bhelgaas: commit log]
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
|
|
Managing pci_set_mwi() with devres can easily be done with its own
callback, without the necessity to store any state about it in a
device-related struct.
Remove the MWI state from struct pci_devres. Give pcim_set_mwi() a
separate devres cleanup callback.
Link: https://lore.kernel.org/r/20240613115032.29098-10-pstanner@redhat.com
Signed-off-by: Philipp Stanner <pstanner@redhat.com>
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
|
|
The bit describing whether the PCI device is currently pinned is stored
in struct pci_devres. To clean up and simplify the PCI devres API, it's
better if this information is stored in struct pci_dev.
This will later permit simplifying pcim_enable_device().
Move the 'pinned' boolean bit to struct pci_dev.
Restructure bits in struct pci_dev so the pm / pme fields are next to
each other.
Link: https://lore.kernel.org/r/20240613115032.29098-9-pstanner@redhat.com
Signed-off-by: Philipp Stanner <pstanner@redhat.com>
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
|
|
The struct pci_devres has a separate boolean to track whether a device is
enabled. That, however, can easily be tracked in an agnostic manner through
the function pci_is_enabled().
Using it allows for simplifying the PCI devres implementation.
Replace the separate 'enabled' status bit from struct pci_devres with
calls to pci_is_enabled() at the appropriate places.
Link: https://lore.kernel.org/r/20240613115032.29098-8-pstanner@redhat.com
Signed-off-by: Philipp Stanner <pstanner@redhat.com>
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
|
|
These functions:
pci_request_region()
pci_request_regions()
pci_request_regions_exclusive()
pci_request_selected_regions()
pci_request_selected_regions_exclusive()
pci_intx()
are "hybrid" functions that are managed if pcim_enable_device() has been
called, but unmanaged otherwise.
This is confusing and has already caused a bug (in 8558de401b5f
("drm/vboxvideo: use managed pci functions")) because users believe all PCI
functions, such as pci_iomap_range(), can become managed that way, which is
not the case.
Add comments to the relevant functions' docstrings that warn users about
this behavior.
Link: https://lore.kernel.org/r/20240613115032.29098-7-pstanner@redhat.com
Signed-off-by: Philipp Stanner <pstanner@redhat.com>
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
[bhelgaas: commit log]
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
|
|
These existing functions:
pci_request_region()
pci_request_selected_regions()
pci_request_selected_regions_exclusive()
are "hybrid" functions built on __pci_request_region() and are managed if
pcim_enable_device() has been called, but unmanaged otherwise.
Add these new functions:
pcim_request_region()
pcim_request_region_exclusive()
These are *always* managed and use the new pcim_addr_devres tracking
infrastructure instead of find_pci_dr() and struct pci_devres.region_mask.
Implement the hybrid functions using the new "pure" functions and remove
struct pci_devres.region_mask, which is no longer needed.
Link: https://lore.kernel.org/r/20240613115032.29098-6-pstanner@redhat.com
Signed-off-by: Philipp Stanner <pstanner@redhat.com>
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
[bhelgaas: commit log]
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
|
|
Deprecate pcim_iomap_table(). It returns a pointer to a table of
ioremapped BARs, or NULL if it fails. This makes uses like this:
addr = pcim_iomap_table(pdev)[0];
problematic because it causes a NULL pointer dereference on failure.
Callers should use pcim_iomap() instead.
Deprecate pcim_iomap_regions_request_all() because it is built on
__pci_request_region() and is managed if pcim_enable_device() has been
called, but unmanaged otherwise, which is prone to errors.
Callers should either use pcim_iomap_regions() to request and map BARs, or
use pcim_request_region() followed by pcim_iomap().
Link: https://lore.kernel.org/r/20240613115032.29098-5-pstanner@redhat.com
Signed-off-by: Philipp Stanner <pstanner@redhat.com>
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
[bhelgaas: commit log, sphinx markup]
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
|
|
The pcim_iomap_devres table tracks entire-BAR mappings, so we can't use it
to build a managed version of pci_iomap_range(), which maps partial BARs.
Add struct pcim_addr_devres, which can track request and mapping of both
entire BARs and partial BARs.
Add the following internal devres functions based on struct
pcim_addr_devres:
pcim_iomap_region() # request & map entire BAR
pcim_iounmap_region() # unmap & release entire BAR
pcim_request_region() # request entire BAR
pcim_release_region() # release entire BAR
pcim_request_all_regions() # request all entire BARs
pcim_release_all_regions() # release all entire BARs
Rework the following public interfaces using the new infrastructure
listed above:
pcim_iomap() # map partial BAR
pcim_iounmap() # unmap partial BAR
pcim_iomap_regions() # request & map specified BARs
pcim_iomap_regions_request_all() # request all BARs, map specified BARs
pcim_iounmap_regions() # unmap & release specified BARs
Link: https://lore.kernel.org/r/20240613115032.29098-4-pstanner@redhat.com
Signed-off-by: Philipp Stanner <pstanner@redhat.com>
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
[bhelgaas: commit log]
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
|
|
The pcim_iomap_devres.table administrated by pcim_iomap_table() has its
entries set and unset at several places throughout devres.c using manual
iterations which are effectively code duplications.
Add pcim_add_mapping_to_legacy_table() and
pcim_remove_mapping_from_legacy_table() helper functions and use them where
possible.
Link: https://lore.kernel.org/r/20240613115032.29098-3-pstanner@redhat.com
Signed-off-by: Philipp Stanner <pstanner@redhat.com>
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
|
|
The current devres implementation uses manual shift operations to check
whether a bit in a mask is set. The code can be made more readable by
writing a small helper function for that.
Implement mask_contains_bar() and use it where applicable.
Link: https://lore.kernel.org/r/20240613115032.29098-2-pstanner@redhat.com
Signed-off-by: Philipp Stanner <pstanner@redhat.com>
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
|
|
From the DWC EP databook 5.96a, section "3.5.7.1.4 General Rules for BAR
Setup (Fixed Mask or Programmable Mask Schemes Only)":
"Any pair (for example BARs 0 and 1) can be configured as one 64-bit BAR,
two 32-bit BARs, or one 32-bit BAR."
"BAR pairs cannot overlap to form a 64-bit BAR. For example, you cannot
combine BARs 1 and 2 to form a 64-bit BAR."
While this limitation does exist in some other PCI endpoint controllers,
e.g. cdns_pcie_ep_set_bar(), the limitation does not appear to be defined
in the PCIe specification itself, thus add an explicit check for this in
dw_pcie_ep_set_bar() (rather than pci_epc_set_bar()).
Link: https://lore.kernel.org/linux-pci/20240528134839.8817-2-cassel@kernel.org
Signed-off-by: Niklas Cassel <cassel@kernel.org>
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
|
|
Down event
Now that dw_pcie_ep_linkdown() is available, use it. This also handles the
reinitialization of DWC non-sticky registers in addition to sending the
notification to EPF drivers.
Closes: https://lore.kernel.org/linux-pci/20240528195539.GA458945@bhelgaas
Link: https://lore.kernel.org/linux-pci/20240606-pci-deinit-v1-5-4395534520dc@linaro.org
Reported-by: Bjorn Helgaas <helgaas@kernel.org>
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Frank Li <Frank.Li@nxp.com>
Reviewed-by: Niklas Cassel <cassel@kernel.org>
|
|
event
Now that the generic dw_pcie_ep_linkdown() API is available, use it. This
also handles the reinitialization of DWC non-sticky registers in addition
to sending the notification to EPF drivers.
Link: https://lore.kernel.org/linux-pci/20240430-pci-epf-rework-v4-9-22832d0d456f@linaro.org
Tested-by: Niklas Cassel <cassel@kernel.org>
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Niklas Cassel <cassel@kernel.org>
|
|
Currently dw_pcie_ep_init_notify() wrapper just calls pci_epc_init_notify()
directly, so this wrapper provides no benefit to the glue drivers.
Remove it and call pci_epc_init_notify() directly from glue drivers.
Suggested-by: Bjorn Helgaas <helgaas@kernel.org>
Link: https://lore.kernel.org/linux-pci/20240606-pci-deinit-v1-1-4395534520dc@linaro.org
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Siddharth Vadapalli <s-vadapalli@ti.com>
Reviewed-by: Frank Li <Frank.Li@nxp.com>
Reviewed-by: Niklas Cassel <cassel@kernel.org>
|
|
Per PCIe r6.0, sec 5.2, a Link Down event can happen under any of the
following circumstances:
1. Fundamental/Hot reset
2. Link disable transmission by upstream component
3. Moving from L2/L3 to L0
In those cases, Link Down causes some non-sticky DWC registers to lose the
state (like REBAR, etc.), so drivers need to reinitialize them to function
properly once the link comes back again.
This is not a problem for drivers supporting PERST# IRQ, since they can
reinitialize the registers in the PERST# IRQ callback. But for the drivers
not supporting PERST#, there is no way they can reinitialize the registers
other than relying on Link Down IRQ received when the link goes down. So
add a DWC generic API dw_pcie_ep_linkdown() that reinitializes the
non-sticky registers and also notifies the EPF drivers about link going
down.
This API can also be used by the drivers supporting PERST# to handle the
scenario (2) mentioned above.
NOTE: For the sake of code organization, move the dw_pcie_ep_linkup()
definition just above dw_pcie_ep_linkdown().
Link: https://lore.kernel.org/linux-pci/20240430-pci-epf-rework-v4-8-22832d0d456f@linaro.org
Tested-by: Niklas Cassel <cassel@kernel.org>
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
[bhelgaas: update spec citation]
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Niklas Cassel <cassel@kernel.org>
|
|
suspend
Instead of relying on the vendor specific implementations to send the
PME_Turn_Off message, introduce a generic way of sending the message using
the MSG TLP.
This is achieved by reserving a region for MSG TLP of size
'pci->region_align', at the end of the first IORESOURCE_MEM window of the
host bridge. And then sending the PME_Turn_Off message during system
suspend with the help of iATU.
The reason for reserving the MSG TLP region at the end of the
IORESOURCE_MEM is to avoid generating holes in between, because when the
region is allocated using allocate_resource(), memory will be allocated
from the start of the window. Later, if memory gets allocated for an
endpoint of size bigger than 'region_align', there will be a hole between
MSG TLP region and endpoint memory.
This generic implementation is optional for the glue drivers and can be
overridden by a custom 'pme_turn_off' callback.
Link: https://lore.kernel.org/linux-pci/20240418-pme_msg-v8-5-a54265c39742@nxp.com
Signed-off-by: Frank Li <Frank.Li@nxp.com>
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
|
|
Add PCIE_MSG_CODE_PME_TURN_OFF macros to enable a PCIe host driver to send
PME_Turn_Off messages.
Link: https://lore.kernel.org/linux-pci/20240418-pme_msg-v8-4-a54265c39742@nxp.com
Signed-off-by: Frank Li <Frank.Li@nxp.com>
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
|
|
Add "Message Routing" and "INTx Mechanism Messages" macros to enable
a PCIe driver to send messages for INTx Interrupt Signaling.
Values from PCIe r6.1, sec 2.2.8 and 2.2.8.1.
Link: https://lore.kernel.org/linux-pci/20240418-pme_msg-v8-1-a54265c39742@nxp.com
Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Signed-off-by: Frank Li <Frank.Li@nxp.com>
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Reviewed-by: Serge Semin <fancer.lancer@gmail.com>
|
|
Add "code" and "routing" into struct dw_pcie_ob_atu_cfg for triggering
INTx IRQs by iATU in the PCIe endpoint mode in near the future.
PCIE_ATU_INHIBIT_PAYLOAD is set to issue TLP type of Msg instead of
MsgD. This implementation supports the data-less messages only for now.
Link: https://lore.kernel.org/linux-pci/20240418-pme_msg-v8-3-a54265c39742@nxp.com
Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Signed-off-by: Frank Li <Frank.Li@nxp.com>
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Reviewed-by: Serge Semin <fancer.lancer@gmail.com>
|
|
This is a preparation before adding the Msg-type outbound iATU
mapping. The respective update will require two more arguments added
to __dw_pcie_prog_outbound_atu(). That will make the already
complicated function prototype even more hard to comprehend accepting
_eight_ arguments.
To prevent that and keep the code more-or-less readable, move all the
outbound iATU-related arguments to a new config structure: struct
dw_pcie_ob_atu_cfg, and pass a pointer to dw_pcie_prog_outbound_atu(). The
structure should be locally defined and populated with the outbound iATU
settings implied by the caller context.
As a result of this change there is no longer need in having the two
distinctive methods for the Host and Endpoint outbound iATU setups since
the code can directly call the dw_pcie_prog_outbound_atu() method with the
config structure populated, so drop dw_pcie_prog_ep_outbound_atu().
[kwilczynski: commit log]
Link: https://lore.kernel.org/linux-pci/20240418-pme_msg-v8-2-a54265c39742@nxp.com
Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Signed-off-by: Frank Li <Frank.Li@nxp.com>
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Reviewed-by: Serge Semin <fancer.lancer@gmail.com>
|
|
When PERST# assert and deassert happens on the PERST# supported platforms,
both iATU0 and iATU6 will map inbound window to BAR0. DMA will access the
area that was previously allocated (iATU0) for BAR0, instead of the new
area (iATU6) for BAR0.
Right now, this isn't an issue because both iATU0 and iATU6 should
translate inbound accesses to BAR0 to the same allocated memory area.
However, having two separate inbound mappings for the same BAR is a
disaster waiting to happen.
The mappings between PCI BAR and iATU inbound window are maintained in the
dw_pcie_ep::bar_to_atu[] array. While allocating a new inbound iATU map for
a BAR, dw_pcie_ep_inbound_atu() API checks for the availability of the
existing mapping in the array and if it is not found (i.e., value in the
array indexed by the BAR is found to be 0), it allocates a new map value
using find_first_zero_bit().
The issue is the existing logic failed to consider the fact that the map
value '0' is a valid value for BAR0, so find_first_zero_bit() will return
'0' as the map value for BAR0 (note that it returns the first zero bit
position).
Due to this, when PERST# assert + deassert happens on the PERST# supported
platforms, the inbound window allocation restarts from BAR0 and the
existing logic to find the BAR mapping will return '6' for BAR0 instead of
'0' due to the fact that it considers '0' as an invalid map value.
Fix this issue by always incrementing the map value before assigning to
bar_to_atu[] array and then decrementing it while fetching. This will make
sure that the map value '0' always represents the invalid mapping."
Fixes: 4284c88fff0e ("PCI: designware-ep: Allow pci_epc_set_bar() update inbound map address")
Closes: https://lore.kernel.org/linux-pci/ZXsRp+Lzg3x%2Fnhk3@x1-carbon/
Link: https://lore.kernel.org/linux-pci/20240412160841.925927-1-Frank.Li@nxp.com
Reported-by: Niklas Cassel <Niklas.Cassel@wdc.com>
Tested-by: Niklas Cassel <niklas.cassel@wdc.com>
Signed-off-by: Frank Li <Frank.Li@nxp.com>
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Reviewed-by: Niklas Cassel <niklas.cassel@wdc.com>
|
|
According to [1], msleep should be used for large sleeps, such as the
100-ish ms one in this function. Comply with the guide and use it.
[1] https://docs.kernel.org/timers/timers-howto.html
[kwilczynski: commit log]
Link: https://lore.kernel.org/linux-pci/20240215-topic-pci_sleep-v2-1-79334884546b@linaro.org
Signed-off-by: Konrad Dybcio <konrad.dybcio@linaro.org>
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Johan Hovold <johan+linaro@kernel.org>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
|
|
The of_gpio.h legacy API is going to be removed. In preparation for that,
convert the driver to the agnostic API.
[kwilczynski: commit log]
Link: https://lore.kernel.org/linux-pci/20240506142142.4042810-6-andriy.shevchenko@linux.intel.com
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Reviewed-by: Rob Herring <robh@kernel.org>
|
|
Convert loops in kirin_pcie_parse_port() to use the _scoped() version of
for_each_available_child_of_node() so the refcounts of children are
implicitly decremented when the loop is exited.
No functional change intended here, but it will make future error exits
from these loops easier.
Link: https://lore.kernel.org/linux-pci/20240609-pcie-kirin-memleak-v1-1-62b45b879576@gmail.com
Signed-off-by: Javier Carrasco <javier.carrasco.cruz@gmail.com>
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
[bhelgaas: move to GPIO series to avoid bisection hole, commit log]
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
|
|
There are two issues related to epf_ntb_epc_cleanup():
1) It should call epf_ntb_config_sspad_bar_clear()
2) The epf_ntb_bind() function should call epf_ntb_epc_cleanup()
to cleanup.
I also changed the ordering a bit. Unwinding should be done in the
mirror order from how they are allocated.
Fixes: e35f56bb0330 ("PCI: endpoint: Support NTB transfer between RC and EP")
Link: https://lore.kernel.org/linux-pci/aaffbe8d-7094-4083-8146-185f4a84e8a1@moroto.mountain
Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
|
|
Smatch complains about inconsistent NULL checking in vpci_scan_bus():
drivers/pci/endpoint/functions/pci-epf-vntb.c:1024 vpci_scan_bus() error: we previously assumed 'vpci_bus' could be null (see line 1021)
Instead of printing an error message and then crashing we should return
an error code and clean up.
Also the NULL check is reversed so it prints an error for success
instead of failure.
Fixes: e35f56bb0330 ("PCI: endpoint: Support NTB transfer between RC and EP")
Link: https://lore.kernel.org/linux-pci/68e0f6a4-fd57-45d0-945b-0876f2c8cb86@moroto.mountain
Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
|
|
Now that the driver core allows for struct class to be in read-only
memory, we should make all 'class' structures declared at build time
placing them into read-only memory, instead of having to be dynamically
allocated at runtime.
Link: https://lore.kernel.org/linux-pci/2024061011-citable-herbicide-1095@gregkh
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
|
|
As like the 'epc_init' event, that is used to signal the EPF drivers about
the EPC initialization, let's introduce 'epc_deinit' event that is used to
signal EPC deinitialization.
The EPC deinitialization applies only when any sort of fundamental reset
is supported by the endpoint controller as per the PCIe spec.
Reference: PCIe r6.0, sec 4.2.5.9.1 and 6.6.1.
Currently, some EPC drivers like pcie-qcom-ep and pcie-tegra194 support
PERST# as the fundamental reset. So the 'deinit' event will be notified to
the EPF drivers when PERST# assert happens in the above mentioned EPC
drivers.
The EPF drivers, on receiving the event through the epc_deinit() callback
should reset the EPF state machine and also cleanup any configuration that
got affected by the fundamental reset like BAR, DMA etc...
This change also warrants skipping the cleanups in unbind() if already done
in epc_deinit().
Link: https://lore.kernel.org/r/20240606-pci-deinit-v1-2-4395534520dc@linaro.org
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Reviewed-by: Niklas Cassel <cassel@kernel.org>
Reviewed-by: Siddharth Vadapalli <s-vadapalli@ti.com>
Reviewed-by: Frank Li <Frank.Li@nxp.com>
|
|
Keith reports a use-after-free when a DPC event occurs concurrently to
hot-removal of the same portion of the hierarchy:
The dpc_handler() awaits readiness of the secondary bus below the
Downstream Port where the DPC event occurred. To do so, it polls the
config space of the first child device on the secondary bus. If that
child device is concurrently removed, accesses to its struct pci_dev
cause the kernel to oops.
That's because pci_bridge_wait_for_secondary_bus() neglects to hold a
reference on the child device. Before v6.3, the function was only
called on resume from system sleep or on runtime resume. Holding a
reference wasn't necessary back then because the pciehp IRQ thread
could never run concurrently. (On resume from system sleep, IRQs are
not enabled until after the resume_noirq phase. And runtime resume is
always awaited before a PCI device is removed.)
However starting with v6.3, pci_bridge_wait_for_secondary_bus() is also
called on a DPC event. Commit 53b54ad074de ("PCI/DPC: Await readiness
of secondary bus after reset"), which introduced that, failed to
appreciate that pci_bridge_wait_for_secondary_bus() now needs to hold a
reference on the child device because dpc_handler() and pciehp may
indeed run concurrently. The commit was backported to v5.10+ stable
kernels, so that's the oldest one affected.
Add the missing reference acquisition.
Abridged stack trace:
BUG: unable to handle page fault for address: 00000000091400c0
CPU: 15 PID: 2464 Comm: irq/53-pcie-dpc 6.9.0
RIP: pci_bus_read_config_dword+0x17/0x50
pci_dev_wait()
pci_bridge_wait_for_secondary_bus()
dpc_reset_link()
pcie_do_recovery()
dpc_handler()
Fixes: 53b54ad074de ("PCI/DPC: Await readiness of secondary bus after reset")
Closes: https://lore.kernel.org/r/20240612181625.3604512-3-kbusch@meta.com/
Link: https://lore.kernel.org/linux-pci/8e4bcd4116fd94f592f2bf2749f168099c480ddf.1718707743.git.lukas@wunner.de
Reported-by: Keith Busch <kbusch@kernel.org>
Tested-by: Keith Busch <kbusch@kernel.org>
Signed-off-by: Lukas Wunner <lukas@wunner.de>
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Reviewed-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Cc: stable@vger.kernel.org # v5.10+
|
|
If the link is powered off during suspend, electrical noise may cause
errors that trigger DPC. If the DPC interrupt is enabled and shares an IRQ
with PME, that causes a spurious wakeup during suspend.
Disable DPC triggering and the DPC interrupt during suspend to prevent
this. Clear DPC interrupt status before re-enabling DPC interrupts during
resume so we don't get an interrupt for errors that occurred during the
suspend/resume process.
Link: https://bugzilla.kernel.org/show_bug.cgi?id=209149
Link: https://bugzilla.kernel.org/show_bug.cgi?id=216295
Link: https://bugzilla.kernel.org/show_bug.cgi?id=218090
Link: https://lore.kernel.org/r/20240416043225.1462548-3-kai.heng.feng@canonical.com
Signed-off-by: Kai-Heng Feng <kai.heng.feng@canonical.com>
[bhelgaas: clear status on resume, add comments, commit log]
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
|
|
If the link is powered off during suspend, electrical noise may cause
errors that are logged via AER. If the AER interrupt is enabled and shares
an IRQ with PME, that causes a spurious wakeup during suspend.
Disable the AER interrupt during suspend to prevent this. Clear error
status before re-enabling IRQ interrupts during resume so we don't get an
interrupt for errors that occurred during the suspend/resume process.
Link: https://bugzilla.kernel.org/show_bug.cgi?id=209149
Link: https://bugzilla.kernel.org/show_bug.cgi?id=216295
Link: https://bugzilla.kernel.org/show_bug.cgi?id=218090
Link: https://lore.kernel.org/r/20240416043225.1462548-2-kai.heng.feng@canonical.com
Signed-off-by: Kai-Heng Feng <kai.heng.feng@canonical.com>
[bhelgaas: drop pci_ancestor_pr3_present() etc, commit log]
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
|
|
With ARCH=arm64, make allmodconfig && make W=1 C=1 reports:
WARNING: modpost: missing MODULE_DESCRIPTION() in drivers/pci/hotplug/acpiphp_ampere_altra.o
Add the missing MODULE_DESCRIPTION().
Link: https://lore.kernel.org/r/20240612-md-drivers-pci-hotplug-v1-1-2b30d14d783d@quicinc.com
Signed-off-by: Jeff Johnson <quic_jjohnson@quicinc.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
|
|
During remove & rescan cycle, PCI subsystem will recalculate and adjust
the bridge window sizing that was initially done by "BIOS". The size
calculation is based on the required alignment of the largest resource
among the downstream resources as per pbus_size_mem() (unimportant or
zero parameters marked with "..."):
min_align = calculate_mem_align(aligns, max_order);
size0 = calculate_memsize(size, ..., min_align);
inside calculate_memsize(), for the largest alignment:
min_align = align1 >> 1;
...
return min_align;
and then in calculate_memsize():
return ALIGN(max(size, ...), align);
If the original bridge window sizing tried to conserve space, this will
lead to massive increase of the required bridge window size when the
downstream has a large disparity in BAR sizes. E.g., with 16MiB and
16GiB BARs this results in 24GiB bridge window size even if 16MiB BAR
does not require gigabytes of space to fit.
When doing remove & rescan for a bus that contains such a PCI device, a
larger bridge window is suddenly required on rescan but when there is a
bridge window upstream that is already assigned based on the original
size, it cannot be enlarged to the new requirement. This causes the
allocation of the bridge window to fail (0x600000000 > 0x400ffffff):
pci 0000:02:01.0: PCI bridge to [bus 03]
pci 0000:02:01.0: bridge window [mem 0x40400000-0x405fffff]
pci 0000:02:01.0: bridge window [mem 0x6000000000-0x6400ffffff 64bit pref]
pci 0000:01:00.0: PCI bridge to [bus 02-04]
pci 0000:01:00.0: bridge window [mem 0x40400000-0x406fffff]
pci 0000:01:00.0: bridge window [mem 0x6000000000-0x6400ffffff 64bit pref]
pci 0000:03:00.0: device released
pci 0000:02:01.0: device released
pcieport 0000:01:00.0: scanning [bus 02-04] behind bridge, pass 0
pci 0000:02:01.0: PCI bridge to [bus 03]
pci 0000:02:01.0: bridge window [mem 0x40400000-0x405fffff]
pci 0000:02:01.0: bridge window [mem 0x6000000000-0x6400ffffff 64bit pref]
pci 0000:02:01.0: scanning [bus 03-03] behind bridge, pass 0
pci 0000:03:00.0: BAR 0 [mem 0x6400000000-0x6400ffffff 64bit pref]
pci 0000:03:00.0: BAR 2 [mem 0x6000000000-0x63ffffffff 64bit pref]
pci 0000:03:00.0: ROM [mem 0x40400000-0x405fffff pref]
pci 0000:02:01.0: PCI bridge to [bus 03]
pci 0000:02:01.0: scanning [bus 03-03] behind bridge, pass 1
pcieport 0000:01:00.0: scanning [bus 02-04] behind bridge, pass 1
pci 0000:02:01.0: bridge window [mem size 0x600000000 64bit pref]: can't assign; no space
pci 0000:02:01.0: bridge window [mem size 0x600000000 64bit pref]: failed to assign
pci 0000:02:01.0: bridge window [mem 0x40400000-0x405fffff]: assigned
pci 0000:03:00.0: BAR 2 [mem size 0x400000000 64bit pref]: can't assign; no space
pci 0000:03:00.0: BAR 2 [mem size 0x400000000 64bit pref]: failed to assign
pci 0000:03:00.0: BAR 0 [mem size 0x01000000 64bit pref]: can't assign; no space
pci 0000:03:00.0: BAR 0 [mem size 0x01000000 64bit pref]: failed to assign
pci 0000:03:00.0: ROM [mem 0x40400000-0x405fffff pref]: assigned
pci 0000:02:01.0: PCI bridge to [bus 03]
pci 0000:02:01.0: bridge window [mem 0x40400000-0x405fffff]
This is a major surprise for users who are suddenly left with a device that
was working fine with the original bridge window sizing.
Even if the already assigned bridge window could be enlarged by
reallocation in some cases (something the current code does not attempt
to do), it is not possible in general case and the large amount of
wasted space at the tail of the bridge window may lead to other
resource exhaustion problems on Root Complex level (think of multiple
PCIe cards with VFs and BAR size disparity in a single system).
PCI BARs only need natural alignment (PCIe r6.1, sec 7.5.1.2.1) and bridge
memory windows need 1MiB (sec 7.5.1.3). The current bridge window tail
alignment rule was introduced in the commit 5d0a8965aea9 ("[PATCH] 2.5.14:
New PCI allocation code (alpha, arm, parisc) [2/2]") that only states:
"pbus_size_mem: core stuff; tested with randomly generated sets of
resources". It does not explain the motivation for the extra tail space
allocated that is not truly needed by the downstream resources. As such, it
is far from clear if it ever has been required by any HW.
To prevent devices with BAR size disparity from becoming unusable after
remove & rescan cycle, attempt to do a truly minimal allocation for memory
resources if needed. First check if the normally calculated bridge window
will not fit into an already assigned upstream resource. In such case, try
with relaxed bridge window tail sizing rules instead where no extra tail
space is requested beyond what the downstream resources require. Only
enforce the alignment requirement of the bridge window itself (normally
1MiB).
With this patch, the resources are successfully allocated:
pci 0000:02:01.0: PCI bridge to [bus 03]
pci 0000:02:01.0: scanning [bus 03-03] behind bridge, pass 1
pcieport 0000:01:00.0: scanning [bus 02-04] behind bridge, pass 1
pcieport 0000:01:00.0: Assigned bridge window [mem 0x6000000000-0x6400ffffff 64bit pref] to [bus 02-04] cannot fit 0x600000000 required for 0000:02:01.0 bridging to [bus 03]
pci 0000:02:01.0: bridge window [mem 0x6000000000-0x6400ffffff 64bit pref] to [bus 03] requires relaxed alignment rules
pcieport 0000:01:00.0: Assigned bridge window [mem 0x40400000-0x406fffff] to [bus 02-04] free space at [mem 0x40400000-0x405fffff]
pci 0000:02:01.0: bridge window [mem 0x6000000000-0x6400ffffff 64bit pref]: assigned
pci 0000:02:01.0: bridge window [mem 0x40400000-0x405fffff]: assigned
pci 0000:03:00.0: BAR 2 [mem 0x6000000000-0x63ffffffff 64bit pref]: assigned
pci 0000:03:00.0: BAR 0 [mem 0x6400000000-0x6400ffffff 64bit pref]: assigned
pci 0000:03:00.0: ROM [mem 0x40400000-0x405fffff pref]: assigned
pci 0000:02:01.0: PCI bridge to [bus 03]
pci 0000:02:01.0: bridge window [mem 0x40400000-0x405fffff]
pci 0000:02:01.0: bridge window [mem 0x6000000000-0x6400ffffff 64bit pref]
This patch draws inspiration from the initial investigations and work by
Mika Westerberg.
Closes: https://bugzilla.kernel.org/show_bug.cgi?id=216795
Link: https://lore.kernel.org/linux-pci/20190812144144.2646-1-mika.westerberg@linux.intel.com/
Fixes: 5d0a8965aea9 ("[PATCH] 2.5.14: New PCI allocation code (alpha, arm, parisc) [2/2]")
Link: https://lore.kernel.org/r/20240507102523.57320-9-ilpo.jarvinen@linux.intel.com
Tested-by: Lidong Wang <lidong.wang@intel.com>
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Mika Westerberg <mika.westerberg@linux.intel.com>
|
|
Calculations related to bridge window size contain literal 20 that is the
minimum alignment for a bridge window. Make the code more obvious by
converting the literal 20 to __ffs(SZ_1M).
Link: https://lore.kernel.org/r/20240507102523.57320-8-ilpo.jarvinen@linux.intel.com
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
[bhelgaas: squash https://lore.kernel.org/r/20240612093250.17544-1-ilpo.jarvinen@linux.intel.com]
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Mika Westerberg <mika.westerberg@linux.intel.com>
|
|
The recent adventure with adding lockdep tracking for cfg_access_lock,
while it yielded many false positives [1], did catch a true positive in the
pci_reset_bus() path [2].
So, while lockdep is difficult to deploy, open coding a check that
cfg_access_lock is held during the reset is feasible.
While this does not offer a full backtrace, it should be sufficient to
implicate the caller of pci_bridge_secondary_bus_reset() as a path that
needs investigation.
Link: https://lore.kernel.org/r/171711746953.1628941.4692125082286867825.stgit@dwillia2-xfh.jf.intel.com
Link: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_134186v1/shard-dg2-1/igt@device_reset@unbind-reset-rebind.html [1]
Link: http://lore.kernel.org/r/cfb50601-5d2a-4676-a958-1bd3f1b06654@intel.com [2]
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Tested-by: Hans de Goede <hdegoede@redhat.com>
Tested-by: Kalle Valo <kvalo@kernel.org>
Reviewed-by: Dave Jiang <dave.jiang@intel.com>
|
