| Age | Commit message (Collapse) | Author |
|---|
|
Remove include <linux/export.h> from all files which do not contain an
EXPORT_SYMBOL().
See commit 7d95680d64ac ("scripts/misc-check: check unnecessary #include
<linux/export.h> when W=1") for more details.
Acked-by: Alexander Gordeev <agordeev@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Alexander Gordeev <agordeev@linux.ibm.com>
|
|
The fixed commit sets up dev.dma_range_map but missed that this is
supposed to be an array of struct bus_dma_region with a sentinel element
with the size field set to 0 at the end. This would lead to overruns in
e.g. dma_range_map_min(). It could also result in wrong translations
instead of DMA_MAPPING_ERROR in translate_phys_to_dma() if the paddr
were to not fit in the aperture. Fix this by using the
dma_direct_set_offset() helper which creates a sentinel for us.
Fixes: d236843a6964 ("s390/pci: store DMA offset in bus_dma_region")
Reviewed-by: Matthew Rosato <mjrosato@linux.ibm.com>
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Link: https://lore.kernel.org/r/20250312-fix_dma_map_alloc-v2-1-530108d9de21@linux.ibm.com
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
|
|
For non-VFs, zpci_bus_is_isolated_vf() should return false because they
aren't VFs. While zpci_iov_find_parent_pf() specifically checks if
a function is a VF, it then simply returns that there is no parent. The
simplistic check for a parent then leads to these functions being
confused with isolated VFs and isolating them on their own domain even
if sibling PFs should share the domain.
Fix this by explicitly checking if a function is not a VF. Note also
that at this point the case where RIDs are ignored is already handled
and in this case all PCI functions get isolated by being detected in
zpci_bus_is_multifunction_root().
Cc: stable@vger.kernel.org
Fixes: 2844ddbd540f ("s390/pci: Fix handling of isolated VFs")
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Reviewed-by: Halil Pasic <pasic@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
|
|
PCI devices on s390 have a DMA offset that is reported via CLP. In
preparation for allowing identity domains, setup the bus_dma_region
for all PCI devices using the reported CLP value.
Signed-off-by: Matthew Rosato <mjrosato@linux.ibm.com>
Reviewed-by: Niklas Schnelle <schnelle@linux.ibm.com>
Tested-by: Niklas Schnelle <schnelle@linux.ibm.com>
Link: https://lore.kernel.org/r/20250212213418.182902-3-mjrosato@linux.ibm.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
|
|
In contrast to the commit message of the fixed commit VFs whose parent
PF is not configured are not always isolated, that is put on their own
PCI domain. This is because for VFs to be added to an existing PCI
domain it is enough for that PCI domain to share the same topology ID or
PCHID. Such a matching PCI domain without a parent PF may exist when
a PF from the same PCI card created the domain with the VF being a child
of a different, non accessible, PF. While not causing technical issues
it makes the rules which VFs are isolated inconsistent.
Fix this by explicitly checking that the parent PF exists on the PCI
domain determined by the topology ID or PCHID before registering the VF.
This works because a parent PF which is under control of this Linux
instance must be enabled and configured at the point where its child VFs
appear because otherwise SR-IOV could not have been enabled on the
parent.
Fixes: 25f39d3dcb48 ("s390/pci: Ignore RID for isolated VFs")
Cc: stable@vger.kernel.org
Reviewed-by: Halil Pasic <pasic@linux.ibm.com>
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
|
|
Since commit 25f39d3dcb48 ("s390/pci: Ignore RID for isolated VFs") PFs
which are not initially configured but in standby are considered
isolated. That is they create only a single function PCI domain. Due to
the PCI domains being created on discovery, this means that even if they
are configured later on, sibling PFs and their child VFs will not be
added to their PCI domain breaking SR-IOV expectations.
The reason the referenced commit ignored standby PFs for the creation of
multi-function PCI subhierarchies, was to work around a PCI domain
renumbering scenario on reboot. The renumbering would occur after
removing a previously in standby PF, whose domain number is used for its
configured sibling PFs and their child VFs, but which itself remained in
standby. When this is followed by a reboot, the sibling PF is used
instead to determine the PCI domain number of it and its child VFs.
In principle it is not possible to know which standby PFs will be
configured later and which may be removed. The PCI domain and root bus
are pre-requisites for hotplug slots so the decision of which functions
belong to which domain can not be postponed. With the renumbering
occurring only in rare circumstances and being generally benign, accept
it as an oddity and fix SR-IOV for initially standby PFs simply by
allowing them to create PCI domains.
Cc: stable@vger.kernel.org
Reviewed-by: Gerd Bayer <gbayer@linux.ibm.com>
Fixes: 25f39d3dcb48 ("s390/pci: Ignore RID for isolated VFs")
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Signed-off-by: Alexander Gordeev <agordeev@linux.ibm.com>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/pci/pci
Pull PCI updates from Bjorn Helgaas:
"Enumeration:
- Make pci_stop_dev() and pci_destroy_dev() safe so concurrent
callers can't stop a device multiple times, even as we migrate from
the global pci_rescan_remove_lock to finer-grained locking (Keith
Busch)
- Improve pci_walk_bus() implementation by making it recursive and
moving locking up to avoid need for a 'locked' parameter (Keith
Busch)
- Unexport pci_walk_bus_locked(), which is only used internally by
the PCI core (Keith Busch)
- Detect some Thunderbolt chips that are built-in and hence
'trustworthy' by a heuristic since the 'ExternalFacingPort' and
'usb4-host-interface' ACPI properties are not quite enough (Esther
Shimanovich)
Resource management:
- Use PCI bus addresses (not CPU addresses) in 'ranges' properties
when building dynamic DT nodes so systems where PCI and CPU
addresses differ work correctly (Andrea della Porta)
- Tidy resource sizing and assignment with helpers to reduce
redundancy (Ilpo Järvinen)
- Improve pdev_sort_resources() 'bogus alignment' warning to be more
specific (Ilpo Järvinen)
Driver binding:
- Convert driver .remove_new() callbacks to .remove() again to finish
the conversion from returning 'int' to being 'void' (Sergio
Paracuellos)
- Export pcim_request_all_regions(), a managed interface to request
all BARs (Philipp Stanner)
- Replace pcim_iomap_regions_request_all() with
pcim_request_all_regions(), and pcim_iomap_table()[n] with
pcim_iomap(n), in the following drivers: ahci, crypto qat, crypto
octeontx2, intel_th, iwlwifi, ntb idt, serial rp2, ALSA korg1212
(Philipp Stanner)
- Remove the now unused pcim_iomap_regions_request_all() (Philipp
Stanner)
- Export pcim_iounmap_region(), a managed interface to unmap and
release a PCI BAR (Philipp Stanner)
- Replace pcim_iomap_regions(mask) with pcim_iomap_region(n), and
pcim_iounmap_regions(mask) with pcim_iounmap_region(n), in the
following drivers: fpga dfl-pci, block mtip32xx, gpio-merrifield,
cavium (Philipp Stanner)
Error handling:
- Add sysfs 'reset_subordinate' to reset the entire hierarchy below a
bridge; previously Secondary Bus Reset could only be used when
there was a single device below a bridge (Keith Busch)
- Warn if we reset a running device where the driver didn't register
pci_error_handlers notification callbacks (Keith Busch)
ASPM:
- Disable ASPM L1 before touching L1 PM Substates to follow the spec
closer and avoid a CPU load timeout on some platforms (Ajay
Agarwal)
- Set devices below Intel VMD to D0 before enabling ASPM L1 Substates
as required per spec for all L1 Substates changes (Jian-Hong Pan)
Power management:
- Enable starfive controller runtime PM before probing host bridge
(Mayank Rana)
- Enable runtime power management for host bridges (Krishna chaitanya
chundru)
Power control:
- Use of_platform_device_create() instead of of_platform_populate()
to create pwrctl platform devices so we can control it based on the
child nodes (Manivannan Sadhasivam)
- Create pwrctrl platform devices only if there's a relevant power
supply property (Manivannan Sadhasivam)
- Add device link from the pwrctl supplier to the PCI dev to ensure
pwrctl drivers are probed before the PCI dev driver; this avoids a
race where pwrctl could change device power state while the PCI
driver was active (Manivannan Sadhasivam)
- Find pwrctl device for removal with of_find_device_by_node()
instead of searching all children of the parent (Manivannan
Sadhasivam)
- Rename 'pwrctl' to 'pwrctrl' to match new bandwidth controller
('bwctrl') and hotplug files (Bjorn Helgaas)
Bandwidth control:
- Add read/modify/write locking for Link Control 2, which is used to
manage Link speed (Ilpo Järvinen)
- Extract Link Bandwidth Management Status check into
pcie_lbms_seen(), where it can be shared between the bandwidth
controller and quirks that use it to help retrain failed links
(Ilpo Järvinen)
- Re-add Link Bandwidth notification support with updates to address
the reasons it was previously reverted (Alexandru Gagniuc, Ilpo
Järvinen)
- Add pcie_set_target_speed() and related functionality so drivers
can manage PCIe Link speed based on thermal or other constraints
(Ilpo Järvinen)
- Add a thermal cooling driver to throttle PCIe Links via the
existing thermal management framework (Ilpo Järvinen)
- Add a userspace selftest for the PCIe bandwidth controller (Ilpo
Järvinen)
PCI device hotplug:
- Add hotplug controller driver for Marvell OCTEON multi-function
device where function 0 has a management console interface to
enable/disable and provision various personalities for the other
functions (Shijith Thotton)
- Retain a reference to the pci_bus for the lifetime of a pci_slot to
avoid a use-after-free when the thunderbolt driver resets USB4 host
routers on boot, causing hotplug remove/add of downstream docks or
other devices (Lukas Wunner)
- Remove unused cpcihp struct cpci_hp_controller_ops.hardware_test
(Guilherme Giacomo Simoes)
- Remove unused cpqphp struct ctrl_dbg.ctrl (Christophe JAILLET)
- Use pci_bus_read_dev_vendor_id() instead of hand-coded presence
detection in cpqphp (Ilpo Järvinen)
- Simplify cpqphp enumeration, which is already simple-minded and
doesn't handle devices below hot-added bridges (Ilpo Järvinen)
Virtualization:
- Add ACS quirk for Wangxun FF5xxx NICs, which don't advertise an ACS
capability but do isolate functions as though PCI_ACS_RR and
PCI_ACS_CR were set, so the functions can be in independent IOMMU
groups (Mengyuan Lou)
TLP Processing Hints (TPH):
- Add and document TLP Processing Hints (TPH) support so drivers can
enable and disable TPH and the kernel can save/restore TPH
configuration (Wei Huang)
- Add TPH Steering Tag support so drivers can retrieve Steering Tag
values associated with specific CPUs via an ACPI _DSM to improve
performance by directing DMA writes closer to their consumers (Wei
Huang)
Data Object Exchange (DOE):
- Wait up to 1 second for DOE Busy bit to clear before writing a
request to the mailbox to avoid failures if the mailbox is still
busy from a previous transfer (Gregory Price)
Endpoint framework:
- Skip attempts to allocate from endpoint controller memory window if
the requested size is larger than the window (Damien Le Moal)
- Add and document pci_epc_mem_map() and pci_epc_mem_unmap() to
handle controller-specific size and alignment constraints, and add
test cases to the endpoint test driver (Damien Le Moal)
- Implement dwc pci_epc_ops.align_addr() so pci_epc_mem_map() can
observe DWC-specific alignment requirements (Damien Le Moal)
- Synchronously cancel command handler work in endpoint test before
cleaning up DMA and BARs (Damien Le Moal)
- Respect endpoint page size in dw_pcie_ep_align_addr() (Niklas
Cassel)
- Use dw_pcie_ep_align_addr() in dw_pcie_ep_raise_msi_irq() and
dw_pcie_ep_raise_msix_irq() instead of open coding the equivalent
(Niklas Cassel)
- Avoid NULL dereference if Modem Host Interface Endpoint lacks
'mmio' DT property (Zhongqiu Han)
- Release PCI domain ID of Endpoint controller parent (not controller
itself) and before unregistering the controller, to avoid
use-after-free (Zijun Hu)
- Clear secondary (not primary) EPC in pci_epc_remove_epf() when
removing the secondary controller associated with an NTB (Zijun Hu)
Cadence PCIe controller driver:
- Lower severity of 'phy-names' message (Bartosz Wawrzyniak)
Freescale i.MX6 PCIe controller driver:
- Fix suspend/resume support on i.MX6QDL, which has a hardware
erratum that prevents use of L2 (Stefan Eichenberger)
Intel VMD host bridge driver:
- Add 0xb60b and 0xb06f Device IDs for client SKUs (Nirmal Patel)
MediaTek PCIe Gen3 controller driver:
- Update mediatek-gen3 DT binding to require the exact number of
clocks for each SoC (Fei Shao)
- Add support for DT 'max-link-speed' and 'num-lanes' properties to
restrict the link speed and width (AngeloGioacchino Del Regno)
Microchip PolarFlare PCIe controller driver:
- Add DT and driver support for using either of the two PolarFire
Root Ports (Conor Dooley)
NVIDIA Tegra194 PCIe controller driver:
- Move endpoint controller cleanups that depend on refclk from the
host to the notifier that tells us the host has deasserted PERST#,
when refclk should be valid (Manivannan Sadhasivam)
Qualcomm PCIe controller driver:
- Add qcom SAR2130P DT binding with an additional clock (Dmitry
Baryshkov)
- Enable MSI interrupts if 'global' IRQ is supported, since a
previous commit unintentionally masked them (Manivannan Sadhasivam)
- Move endpoint controller cleanups that depend on refclk from the
host to the notifier that tells us the host has deasserted PERST#,
when refclk should be valid (Manivannan Sadhasivam)
- Add DT binding and driver support for IPQ9574, with Synopsys IP
v5.80a and Qcom IP 1.27.0 (devi priya)
- Move the OPP "operating-points-v2" table from the
qcom,pcie-sm8450.yaml DT binding to qcom,pcie-common.yaml, where it
can be used by other Qcom platforms (Qiang Yu)
- Add 'global' SPI interrupt for events like link-up, link-down to
qcom,pcie-x1e80100 DT binding so we can start enumeration when the
link comes up (Qiang Yu)
- Disable ASPM L0s for qcom,pcie-x1e80100 since the PHY is not tuned
to support this (Qiang Yu)
- Add ops_1_21_0 for SC8280X family SoC, which doesn't use the
'iommu-map' DT property and doesn't need BDF-to-SID translation
(Qiang Yu)
Rockchip PCIe controller driver:
- Define ROCKCHIP_PCIE_AT_SIZE_ALIGN to replace magic 256 endpoint
.align value (Damien Le Moal)
- When unmapping an endpoint window, compute the region index instead
of searching for it, and verify that the address was mapped (Damien
Le Moal)
- When mapping an endpoint window, verify that the address hasn't
been mapped already (Damien Le Moal)
- Implement pci_epc_ops.align_addr() for rockchip-ep (Damien Le Moal)
- Fix MSI IRQ data mapping to observe the alignment constraint, which
fixes intermittent page faults in memcpy_toio() and memcpy_fromio()
(Damien Le Moal)
- Rename rockchip_pcie_parse_ep_dt() to
rockchip_pcie_ep_get_resources() for consistency with similar DT
interfaces (Damien Le Moal)
- Skip the unnecessary link train in rockchip_pcie_ep_probe() and do
it only in the endpoint start operation (Damien Le Moal)
- Implement pci_epc_ops.stop_link() to disable link training and
controller configuration (Damien Le Moal)
- Attempt link training at 5 GT/s when both partners support it
(Damien Le Moal)
- Add a handler for PERST# signal so we can detect host-initiated
resets and start link training after PERST# is deasserted (Damien
Le Moal)
Synopsys DesignWare PCIe controller driver:
- Clear outbound address on unmap so dw_pcie_find_index() won't match
an ATU index that was already unmapped (Damien Le Moal)
- Use of_property_present() instead of of_property_read_bool() when
testing for presence of non-boolean DT properties (Rob Herring)
- Advertise 1MB size if endpoint supports Resizable BARs, which was
inadvertently lost in v6.11 (Niklas Cassel)
TI J721E PCIe driver:
- Add PCIe support for J722S SoC (Siddharth Vadapalli)
- Delay PCIE_T_PVPERL_MS (100 ms), not just PCIE_T_PERST_CLK_US (100
us), before deasserting PERST# to ensure power and refclk are
stable (Siddharth Vadapalli)
TI Keystone PCIe controller driver:
- Set the 'ti,keystone-pcie' mode so v3.65a devices work in Root
Complex mode (Kishon Vijay Abraham I)
- Try to avoid unrecoverable SError for attempts to issue config
transactions when the link is down; this is racy but the best we
can do (Kishon Vijay Abraham I)
Miscellaneous:
- Reorganize kerneldoc parameter names to match order in function
signature (Julia Lawall)
- Fix sysfs reset_method_store() memory leak (Todd Kjos)
- Simplify pci_create_slot() (Ilpo Järvinen)
- Fix incorrect printf format specifiers in pcitest (Luo Yifan)"
* tag 'pci-v6.13-changes' of git://git.kernel.org/pub/scm/linux/kernel/git/pci/pci: (127 commits)
PCI: rockchip-ep: Handle PERST# signal in EP mode
PCI: rockchip-ep: Improve link training
PCI: rockship-ep: Implement the pci_epc_ops::stop_link() operation
PCI: rockchip-ep: Refactor endpoint link training enable
PCI: rockchip-ep: Refactor rockchip_pcie_ep_probe() MSI-X hiding
PCI: rockchip-ep: Refactor rockchip_pcie_ep_probe() memory allocations
PCI: rockchip-ep: Rename rockchip_pcie_parse_ep_dt()
PCI: rockchip-ep: Fix MSI IRQ data mapping
PCI: rockchip-ep: Implement the pci_epc_ops::align_addr() operation
PCI: rockchip-ep: Improve rockchip_pcie_ep_map_addr()
PCI: rockchip-ep: Improve rockchip_pcie_ep_unmap_addr()
PCI: rockchip-ep: Use a macro to define EP controller .align feature
PCI: rockchip-ep: Fix address translation unit programming
PCI/pwrctrl: Rename pwrctrl functions and structures
PCI/pwrctrl: Rename pwrctl files to pwrctrl
PCI/pwrctl: Remove pwrctl device without iterating over all children of pwrctl parent
PCI/pwrctl: Ensure that pwrctl drivers are probed before PCI client drivers
PCI/pwrctl: Create pwrctl device only if at least one power supply is present
PCI/pwrctl: Use of_platform_device_create() to create pwrctl devices
tools: PCI: Fix incorrect printf format specifiers
...
|
|
2fe2abf896c1 ("PCI: augment bus resource table with a list") added
PCI_SUBTRACTIVE_DECODE which is put into the struct pci_bus_resource flags
field but is never read. There seems to never have been users for it.
Remove both PCI_SUBTRACTIVE_DECODE and the flags field from the struct
pci_bus_resource.
Link: https://lore.kernel.org/r/20241017141111.44612-1-ilpo.jarvinen@linux.intel.com
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
|
|
Ensure that VFs used in isolation, that is with their parent PF not
visible to the configuration but with their RID exposed, are treated
compatibly with existing isolated VF use cases without exposed RID
including RoCE Express VFs. This allows creating configurations where
one LPAR manages PFs while their child VFs are used by other LPARs. This
gives the LPAR managing the PFs a role analogous to that of the
hypervisor in a typical use case of passing child VFs to guests.
Instead of creating a multifunction struct zpci_bus whenever a PCI
function with RID exposed is discovered only create such a bus for
configured physical functions and only consider multifunction busses
when searching for an existing bus. Additionally only set zdev->devfn to
the devfn part of the RID once the function is added to a multifunction
bus.
This also fixes probing of more than 7 such isolated VFs from the same
physical bus. This is because common PCI code in pci_scan_slot() only
looks for more functions when pdev->multifunction is set which somewhat
counter intutively is not the case for VFs.
Note that PFs are looked at before their child VFs is guaranteed because
we sort the zpci_list by RID ascending.
Reviewed-by: Gerd Bayer <gbayer@linux.ibm.com>
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
|
|
The newly introduced topology ID (TID) field in the CLP Query PCI
Function explicitly identifies groups of PCI functions whose RIDs belong
to the same (sub-)topology. When available use the TID instead of the
PCHID to match zPCI busses/domains for multi-function devices. Note that
currently only a single PCI bus per TID is supported. This change is
required because in future machines the PCHID will not identify a PCI
card but a specific port in the case of some multi-port NICs while from
a PCI point of view the entire card is a subtopology.
Reviewed-by: Gerd Bayer <gbayer@linux.ibm.com>
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
|
|
While s390 already has a standard IOMMU driver and previous changes have
added I/O TLB flushing operations this driver is currently only used for
user-space PCI access such as vfio-pci. For the DMA API s390 instead
utilizes its own implementation in arch/s390/pci/pci_dma.c which drives
the same hardware and shares some code but requires a complex and
fragile hand over between DMA API and IOMMU API use of a device and
despite code sharing still leads to significant duplication and
maintenance effort. Let's utilize the common code DMAP API
implementation from drivers/iommu/dma-iommu.c instead allowing us to
get rid of arch/s390/pci/pci_dma.c.
Reviewed-by: Matthew Rosato <mjrosato@linux.ibm.com>
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Link: https://lore.kernel.org/r/20230928-dma_iommu-v13-3-9e5fc4dacc36@linux.ibm.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
|
|
Heiko Carstens says:
===================
There are a couple of oddities within the s390 uaccess library
functions. Therefore cleanup the whole uaccess.c file.
There is no functional change, only improved readability. The output
of "objdump -Dr" was always compared before/after each patch to make
sure that the generated object file is identical, if that could be
expected. Therefore the series also includes more patches than really
required to cleanup the code.
Furthermore the kunit usercopy tests also still pass.
===================
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
|
|
Prior to commit 960ac3626487 ("s390/pci: allow zPCI zbus without
a function zero") enabling and scanning a PCI function had to
potentially be postponed until the function with devfn zero on that bus
was plugged. While the commit removed the waiting itself extra code to
scan all functions on the PCI bus once function zero appeared was
missed. Remove that code and the outdated comments about waiting for
function zero.
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Reviewed-by: Matthew Rosato <mjrosato@linux.ibm.com>
Link: https://lore.kernel.org/r/20230306151014.60913-5-schnelle@linux.ibm.com
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
|
|
The pci_bus_add_devices() call in zpci_bus_create_pci_bus() is without
function since at this point no device could have been added to the
freshly created PCI bus.
Suggested-by: Bjorn Helgaas <helgaas@kernel.org>
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Reviewed-by: Matthew Rosato <mjrosato@linux.ibm.com>
Link: https://lore.kernel.org/r/20230306151014.60913-4-schnelle@linux.ibm.com
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
|
|
As the name suggests zpci_bus_scan_device() is used to scan a specific
device and thus pci_bus_add_device() for that device is sufficient.
Furthermore move this call inside the rescan/remove locking.
Suggested-by: Bjorn Helgaas <helgaas@kernel.org>
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Reviewed-by: Matthew Rosato <mjrosato@linux.ibm.com>
Link: https://lore.kernel.org/r/20230306151014.60913-3-schnelle@linux.ibm.com
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
|
|
On s390 PCI functions may be hotplugged individually even when they
belong to a multi-function device. In particular on an SR-IOV device VFs
may be removed and later re-added.
In commit a50297cf8235 ("s390/pci: separate zbus creation from
scanning") it was missed however that struct pci_bus and struct
zpci_bus's resource list retained a reference to the PCI functions MMIO
resources even though those resources are released and freed on
hot-unplug. These stale resources may subsequently be claimed when the
PCI function re-appears resulting in use-after-free.
One idea of fixing this use-after-free in s390 specific code that was
investigated was to simply keep resources around from the moment a PCI
function first appeared until the whole virtual PCI bus created for
a multi-function device disappears. The problem with this however is
that due to the requirement of artificial MMIO addreesses (address
cookies) extra logic is then needed to keep the address cookies
compatible on re-plug. At the same time the MMIO resources semantically
belong to the PCI function so tying their lifecycle to the function
seems more logical.
Instead a simpler approach is to remove the resources of an individually
hot-unplugged PCI function from the PCI bus's resource list while
keeping the resources of other PCI functions on the PCI bus untouched.
This is done by introducing pci_bus_remove_resource() to remove an
individual resource. Similarly the resource also needs to be removed
from the struct zpci_bus's resource list. It turns out however, that
there is really no need to add the MMIO resources to the struct
zpci_bus's resource list at all and instead we can simply use the
zpci_bar_struct's resource pointer directly.
Fixes: a50297cf8235 ("s390/pci: separate zbus creation from scanning")
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Reviewed-by: Matthew Rosato <mjrosato@linux.ibm.com>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Link: https://lore.kernel.org/r/20230306151014.60913-2-schnelle@linux.ibm.com
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
|
|
Currently the zPCI code blocks PCI bus creation and probing of a zPCI zbus
unless there is a PCI function with devfn 0. This is always the case for
the PCI functions with hidden RID, but may keep PCI functions from a
multi-function PCI device with RID information invisible until the function
0 becomes visible. Worse, as a PCI bus is necessary to even present a PCI
hotplug slot, even that remains invisible.
With the probing of these so-called isolated PCI functions enabled for s390
in common code, this restriction is no longer necessary. On network cards
with multiple ports and a PF per port this also allows using each port on
its own while still providing the physical PCI topology information in the
devfn needed to associate VFs with their parent PF.
Link: https://lore.kernel.org/r/20220628143100.3228092-6-schnelle@linux.ibm.com
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Pierre Morel <pmorel@linux.ibm.com>
|
|
Currently zpci_dma_init_device()/zpci_dma_exit_device() is called as
part of zpci_enable_device()/zpci_disable_device() and errors for
zpci_dma_exit_device() are always ignored even if we could abort.
Improve upon this by moving zpci_dma_exit_device() out of
zpci_disable_device() and check for errors whenever we have a way to
abort the current operation. Note that for example in
zpci_event_hard_deconfigured() the device is expected to be gone so we
really can't abort and proceed even in case of error.
Similarly move the cc == 3 special case out of zpci_unregister_ioat()
and into the callers allowing to abort when finding an already disabled
devices precludes proceeding with the operation.
While we are at it log IOAT register/unregister errors in the s390
debugfs log,
Reviewed-by: Matthew Rosato <mjrosato@linux.ibm.com>
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
|
|
On failure to register a struct zpci_dev with a struct zpci_bus we left
a dangling pointer in zdev->zbus. As zpci_create_device() bails if
zpci_bus_device_register() fails this is of no consequence but still bad
practice.
Reviewed-by: Matthew Rosato <mjrosato@linux.ibm.com>
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
|
|
Now that the zbus can be created without being scanned we can go one
step further and make registering a device to a zbus independent from
scanning it. This way the zbus handling becomes much more natural
in that functions can be registered on the zbus to be scanned later more
closely resembling the handling of both real PCI hardware and other
virtual PCI busses like Hyper-V's virtual PCI bus (see for example
drivers/pci/controller/pci-hyperv.c:create_root_hv_pci_bus()).
Having zbus registration separate from scanning allows us to return
fully initialized but still disabled zdevs from zpci_create_device()
which can then be configured just as we would configure a zdev from
standby (minus the SCLP Configure already done by the platform). There
is still the exception that a PCI function with non-zero devfn can be
plugged before its PCI bus, which depends on the function with zero
devfn, is created. In this case the zdev returend from
zpci_create_device() is still missing its bus, hotplug slot, and
resources which need to be created later but at least it doesn't wait in
the enabled state and can otherwise be treated as initialized.
With this we also separate the initial PCI scan using CLP List PCI
Functions into two phases. In the CLP loop's callback we only register
each function with a virtual zbus creating the latter as needed. Then,
after we have built this virtual PCI topology based on our list of
zbusses, we can make use of the common code functionality to scan each
complete zbus as a separate child bus.
Reviewed-by: Matthew Rosato <mjrosato@linux.ibm.com>
Acked-by: Pierre Morel <pmorel@linux.ibm.com>
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
|
|
In a later change we will first collect all PCI functions from the CLP
List PCI functions call, then register them to/creating the relevant
zbus. Then only after we've created our virtual bus structure will we
scan all zbusses iterating over the zbus list. Since scanning is
relatively slow a spinlock is a bad fit for protecting the
loop over the devices on the zbus. Furthermore doing the probing on the
bus we need to use pci_lock_rescan_remove() as devices are added to
the PCI subsystem and that is a mutex which can't be locked nested
inside a spinlock section. Note that the contention of this lock should
be very low either way as zbusses are only added/removed concurrently on
hotplug events.
Reviewed-by: Matthew Rosato <mjrosato@linux.ibm.com>
Reviewed-by: Pierre Morel <pmorel@linux.ibm.com>
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
|
|
In the existing code the creation of the PCI bus and the scanning of
function zero all happens in zpci_scan_bus(). This in turn requires
functions to be enabled and their resources to be available before the
PCI bus is even created.
This not only means that functions are enabled long before they are
actually made available to the common PCI subsystem. In case of
functions with non-zero devfn which appeared before the function with
devfn zero they can wait arbitrarily long in this enabled but not
scanned state.
Fix this by separating the creation of the PCI bus from scanning it and
only prepare, that is enable and setup MMIO bus resources, functions
just before they are scanned. As they may be scanned multiple times
track if we already created resources in the zdev.
Reviewed-by: Matthew Rosato <mjrosato@linux.ibm.com>
Acked-by: Pierre Morel <pmorel@linux.ibm.com>
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
|
|
Pull setting the maximum bus speed and multifunction attribute into
zpci_bus_scan() in preparation for handling bus creation separately
from scanning the bus.
Reviewed-by: Matthew Rosato <mjrosato@linux.ibm.com>
Acked-by: Pierre Morel <pmorel@linux.ibm.com>
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
|
|
To match zpci_bus_scan_device() and the PCI common code terminology and
to remove some code duplication, we pull the multiple uses of
pci_scan_single_device() into a function. For now this has the side
effect of adding each device to the PCI bus separately and locking and
unlocking the rescan/remove lock for each instead of just once per bus.
This is clearly less efficient but provides a correct intermediate
behavior until a follow on change does both the adding and scanning only
once per bus.
Reviewed-by: Matthew Rosato <mjrosato@linux.ibm.com>
Acked-by: Pierre Morel <pmorel@linux.ibm.com>
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
|
|
The zpci_remove_device() function removes the device from the PCI common
code core which is an operation dealing primarily with the zbus and PCI
bus code. With that and to match an upcoming refactoring of the
symmetric scanning part move it to the bus code.
Reviewed-by: Matthew Rosato <mjrosato@linux.ibm.com>
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
|
|
Pull VFIO updates from Alex Williamson:
- New fsl-mc vfio bus driver supporting userspace drivers of objects
within NXP's DPAA2 architecture (Diana Craciun)
- Support for exposing zPCI information on s390 (Matthew Rosato)
- Fixes for "detached" VFs on s390 (Matthew Rosato)
- Fixes for pin-pages and dma-rw accesses (Yan Zhao)
- Cleanups and optimize vconfig regen (Zenghui Yu)
- Fix duplicate irq-bypass token registration (Alex Williamson)
* tag 'vfio-v5.10-rc1' of git://github.com/awilliam/linux-vfio: (30 commits)
vfio iommu type1: Fix memory leak in vfio_iommu_type1_pin_pages
vfio/pci: Clear token on bypass registration failure
vfio/fsl-mc: fix the return of the uninitialized variable ret
vfio/fsl-mc: Fix the dead code in vfio_fsl_mc_set_irq_trigger
vfio/fsl-mc: Fixed vfio-fsl-mc driver compilation on 32 bit
MAINTAINERS: Add entry for s390 vfio-pci
vfio-pci/zdev: Add zPCI capabilities to VFIO_DEVICE_GET_INFO
vfio/fsl-mc: Add support for device reset
vfio/fsl-mc: Add read/write support for fsl-mc devices
vfio/fsl-mc: trigger an interrupt via eventfd
vfio/fsl-mc: Add irq infrastructure for fsl-mc devices
vfio/fsl-mc: Added lock support in preparation for interrupt handling
vfio/fsl-mc: Allow userspace to MMAP fsl-mc device MMIO regions
vfio/fsl-mc: Implement VFIO_DEVICE_GET_REGION_INFO ioctl call
vfio/fsl-mc: Implement VFIO_DEVICE_GET_INFO ioctl
vfio/fsl-mc: Scan DPRC objects on vfio-fsl-mc driver bind
vfio: Introduce capability definitions for VFIO_DEVICE_GET_INFO
s390/pci: track whether util_str is valid in the zpci_dev
s390/pci: stash version in the zpci_dev
vfio/fsl-mc: Add VFIO framework skeleton for fsl-mc devices
...
|
|
For s390 we can have VFs that are passed-through without the associated
PF. Firmware provides an emulation layer to allow these devices to
operate independently, but is missing emulation of the Memory Space
Enable bit. For these as well as linked VFs, set no_command_memory
which specifies these devices do not implement PCI_COMMAND_MEMORY.
Fixes: abafbc551fdd ("vfio-pci: Invalidate mmaps and block MMIO access on disabled memory")
Signed-off-by: Matthew Rosato <mjrosato@linux.ibm.com>
Reviewed-by: Niklas Schnelle <schnelle@linux.ibm.com>
Reviewed-by: Pierre Morel <pmorel@linux.ibm.com>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
|
|
currently we have multiple #ifdef CONFIG_PCI_IOV blocks spread over
different compliation units and headers, all dealing with SR-IOV
specific behavior.
This violates the style guide which discourages conditionally compiled
code blocks and hinders maintainability by speading SR-IOV functionality
over many files.
Let's move all of this into a conditionally compiled pci_iov.c file and
local header and prefix SR-IOV specific functions with zpci_iov_*.
Reviewed-by: Matthew Rosato <mjrosato@linux.ibm.com>
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
|
|
Currently there are four places in which a PCI function is scanned
and made available to drivers:
1. In pci_scan_root_bus() as part of the initial zbus
creation.
2. In zpci_bus_add_devices() when registering
a device in configured state on a zbus that has already been
scanned.
3. When a function is already known to zPCI (in reserved/standby state)
and configuration is triggered through firmware by PEC 0x301.
4. When a device is already known to zPCI (in standby/reserved state)
and configuration is triggered from within Linux using
enable_slot().
The PF/VF linking step and setting of pdev->is_virtfn introduced with
commit e5794cf1a270 ("s390/pci: create links between PFs and VFs") was
only triggered for the second case, which is where VFs created through
sriov_numvfs usually land. However unlike some other platforms but like
POWER VFs can be individually enabled/disabled through
/sys/bus/pci/slots.
Fix this by doing VF setup as part of pcibios_bus_add_device() which is
called in all of the above cases.
Finally to remove the PF/VF links call the common code
pci_iov_remove_virtfn() function to remove linked VFs.
This takes care of the necessary sysfs cleanup.
Fixes: e5794cf1a270 ("s390/pci: create links between PFs and VFs")
Cc: <stable@vger.kernel.org> # 5.8: 2f0230b2f2d5: s390/pci: re-introduce zpci_remove_device()
Cc: <stable@vger.kernel.org> # 5.8
Acked-by: Pierre Morel <pmorel@linux.ibm.com>
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
|
|
We were missing the pci_dev_put() for candidate PFs. Furhtermore in
discussion with upstream it turns out that somewhat counterintuitively
some common code, in particular the vfio-pci driver, assumes that
pdev->is_virtfn always implies that pdev->physfn is set, i.e. that VFs
are always linked.
While POWER does seem to set pdev->is_virtfn even for unlinked functions
(see comments in arch/powerpc/kernel/eeh.c:eeh_debugfs_break_device())
for now just be safe and only set pdev->is_virtfn on linking.
Also make sure that we only search for parent PFs if the zbus is
multifunction and we thus know the devfn values supplied by firmware
come from the RID.
Fixes: e5794cf1a270 ("s390/pci: create links between PFs and VFs")
Cc: <stable@vger.kernel.org> # 5.8
Reviewed-by: Pierre Morel <pmorel@linux.ibm.com>
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
|
|
On s390 PCI Virtual Functions (VFs) are scanned by firmware and are made
available to Linux via the hot-plug interface. As such the common code
path of doing the scan directly using the parent Physical Function (PF)
is not used and fenced off with the no_vf_scan attribute.
Even if the partition created the VFs itself e.g. using the sriov_numvfs
attribute of a PF, the PF/VF links thus need to be established after the
fact. To do this when a VF is plugged we scan through all functions on
the same zbus and test whether they are the parent PF in which case we
establish the necessary links.
With these links established there is now no more need to fence off
pci_iov_remove_virtfn() for pdev->no_vf_scan as the common code now
works fine.
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Pierre Morel <pmorel@linux.ibm.com>
Link: https://lore.kernel.org/r/20200506154139.90609-3-schnelle@linux.ibm.com
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
|
|
The assignment of the PCI device multifunction attribute
is set during the PCI device probe.
There is no need to set it here.
Let's do it right and remove this assignment.
Signed-off-by: Pierre Morel <pmorel@linux.ibm.com>
Reviewed-by: Niklas Schnelle <schnelle@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
|
|
We allow multiple functions on a single bus.
We suppress the ZPCI_DEVFN definition and replace its
occurences with zpci->devfn.
We verify the number of device during the registration.
There can never be more domains in use than existing
devices, so we do not need to verify the count of domain
after having verified the count of devices.
Signed-off-by: Pierre Morel <pmorel@linux.ibm.com>
Reviewed-by: Niklas Schnelle <schnelle@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
|
|
The current PCI implementation do not provide a bus resource.
This leads to a notice being print at boot.
Let's do it more nicely and provide the bus resource.
Signed-off-by: Pierre Morel <pmorel@linux.ibm.com>
Reviewed-by: Niklas Schnelle <schnelle@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
|
|
The zPCI bus is in charge to handle common zPCI resources for
zPCI devices.
Creating the zPCI bus, the PCI bus, the zPCI devices and the
PCI devices and hotplug slots
done in a specific order:
- PCI hotplug slot creation needs a PCI bus
- PCI bus needs a PCI domain
which is reported by the pci_domain_nr() when setting up the
host bridge
- PCI domain is set from the zPCI with devfn 0
this is necessary to have a reproducible enumeration
Therefore we can not create devices or hotplug slots for any PCI
device associated with a zPCI device before having discovered
the function zero of the bus.
The discovery and initialization of devices can be done at several
points in the code:
- On Events, serialized in a thread context
- On initialization, in the kernel init thread context
- When powering on the hotplug slot, in a user thread context
The removal of devices and their parent bus may also be done on
events or for devices when powering down the slot.
To guarantee the existence of the bus and devices until they are
no more needed we use kref in zPCI bus and introduce a reference
count in the zPCI devices.
In this patch the zPCI bus still only accept a device with
a devfn 0.
Signed-off-by: Pierre Morel <pmorel@linux.ibm.com>
Reviewed-by: Niklas Schnelle <schnelle@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
|
