When doing a function/slot/bus reset PCI grabs the device_lock for each
device to block things like suspend and driver probes, but call paths exist
where this lock may already be held. This creates an opportunity for
deadlock. For instance, vfio allows userspace to issue resets so long as
it owns the device(s). If a driver unbind .remove callback races with
userspace issuing a reset, we have a deadlock as userspace gets stuck
waiting on device_lock while another thread has device_lock and waits for
.remove to complete. To resolve this, we can make a version of the reset
interfaces which use trylock. With this, we can safely attempt a reset and
return error to userspace if there is contention.
[bhelgaas: the deadlock happens when A (userspace) has a file descriptor for
the device, and B waits in this path:
driver_detach
device_lock # take device_lock
__device_release_driver
pci_device_remove # pci_bus_type.remove
vfio_pci_remove # pci_driver .remove
vfio_del_group_dev
wait_event(vfio.release_q, !vfio_dev_present) # wait (holding device_lock)
Now B is stuck until A gives up the file descriptor. If A tries to acquire
device_lock for any reason, we deadlock because A is waiting for B to release
the lock, and B is waiting for A to release the file descriptor.]
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Using 'make namespacecheck' identify code which should be declared static.
Checked for users in other driver/archs as well. Compile tested only.
This stops exporting the following interfaces to modules:
pci_target_state()
pci_load_saved_state()
[bhelgaas: retained pci_find_next_ext_capability() and pci_cfg_space_size()]
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
My philosophy is unused code is dead code. And dead code is subject to bit
rot and is a likely source of bugs. Use it or lose it.
This reverts b48d4425b6 ("PCI: add ID-based ordering enable/disable
support"), removing these interfaces:
pci_enable_ido()
pci_disable_ido()
[bhelgaas: split to separate patch, also remove prototypes from pci.h]
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
CC: Jesse Barnes <jbarnes@virtuousgeek.org>
My philosophy is unused code is dead code. And dead code is subject to bit
rot and is a likely source of bugs. Use it or lose it.
This reverts 48a92a8179 ("PCI: add OBFF enable/disable support"),
removing these interfaces:
pci_enable_obff()
pci_disable_obff()
[bhelgaas: split to separate patch, also remove prototypes from pci.h]
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
CC: Jesse Barnes <jbarnes@virtuousgeek.org>
My philosophy is unused code is dead code. And dead code is subject to bit
rot and is a likely source of bugs. Use it or lose it.
This reverts 51c2e0a7e5 ("PCI: add latency tolerance reporting
enable/disable support"), removing these interfaces:
pci_enable_ltr()
pci_disable_ltr()
pci_set_ltr()
[bhelgaas: split to separate patch, also remove prototypes from pci.h]
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
CC: Jesse Barnes <jbarnes@virtuousgeek.org>
While we don't really have any infrastructure for making use of VC
support, the system BIOS can configure the topology to non-default
VC values prior to boot. This may be due to silicon bugs, desire to
reserve traffic classes, or perhaps just BIOS bugs. When we reset
devices, the VC configuration may return to default values, which can
be incompatible with devices upstream. For instance, Nvidia GRID
cards provide a PCIe switch and some number of GPUs, all supporting
VC. The power-on default for VC is to support TC0-7 across VC0,
however some platforms will only enable TC0/VC0 mapping across the
topology. When we do a secondary bus reset on the downstream switch
port, the GPU is reset to a TC0-7/VC0 mapping while the opposite end
of the link only enables TC0/VC0. If the GPU attempts to use TC1-7,
it fails.
This patch attempts to provide complete support for VC save/restore,
even beyond the minimally required use case above. This includes
save/restore and reload of the arbitration table, save/restore and
reload of the port arbitration tables, and re-enabling of the
channels for VC, VC9, and MFVC capabilities.
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Current save/restore is specific to standard capabilities.
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
We currently have two instance of this loop which waits for a pending bit
to clear in a status dword. Generalize the function for future users.
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Modify tg3_chip_reset() and tg3_close() to check if the PCI network
adapter device is accessible at all in order to skip poking it or
trying to handle a carrier loss in vain when that's not the case.
Introduce a special PCI helper function pci_device_is_present()
for this purpose.
Of course, this uncovers the lack of the appropriate RTNL locking
in tg3_suspend() and tg3_resume(), so add that locking in there
too.
These changes prevent tg3 from burning a CPU at 100% load level for
solid several seconds after the Thunderbolt link is disconnected from
a Matrox DS1 docking station.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix whitespace, capitalization, and spelling errors. No functional change.
I know "busses" is not an error, but "buses" was more common, so I used it
consistently.
Signed-off-by: Marta Rybczynska <rybczynska@gmail.com> (pci_reset_bridge_secondary_bus())
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
* pci/misc:
PCI: Enable upstream bridges even for VFs on virtual buses
PCI: Add pci_upstream_bridge()
PCI: Add x86_msi.msi_mask_irq() and msix_mask_irq()
Previously we enabled the upstream PCI-to-PCI bridge only when
"dev->bus->self != NULL". In the case of a VF on a virtual bus, where
"bus->self == NULL", we didn't enable the upstream bridge.
This fixes that by enabling the upstream bridge of the PF corresponding to
the VF.
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Acked-by: Yinghai Lu <yinghai@kernel.org>
* pci/misc:
PCI: Warn on driver probe return value greater than zero
PCI: Drop warning about drivers that don't use pci_set_master()
PCI: Workaround missing pci_set_master in pci drivers
PCI: Update pcie_ports 'auto' behavior for non-ACPI platforms
f41f064cf4 ("PCI: Workaround missing pci_set_master in pci drivers") made
pci_enable_bridge() turn on bus mastering if the driver hadn't done so
already. It also added a warning in this case. But there's no reason to
warn about it unless it's actually a problem to enable bus mastering here.
This patch drops the warning because I'm not aware of any such problem.
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
CC: Paul Bolle <pebolle@tiscali.nl>
Ben Herrenschmidt found that commit 928bea9648 ("PCI: Delay enabling
bridges until they're needed") breaks PCI in some powerpc environments.
The reason is that the PCIe port driver will call pci_enable_device() on
the bridge, so the device is enabled, but skips pci_set_master because
pcie_port_auto and no acpi on powerpc.
Because of that, pci_enable_bridge() later on (called as a result of the
child device driver doing pci_enable_device) will see the bridge as
already enabled and will not call pci_set_master() on it.
Fixed by add checking in pci_enable_bridge, and call pci_set_master
if driver skip that.
That will make the code more robot and wade off problem for missing
pci_set_master in drivers.
Reported-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
* pci/yijing-mps-v1:
drm/radeon: use pcie_get_readrq() and pcie_set_readrq() to simplify code
staging: et131x: Use pci_dev->pcie_mpss and pcie_set_readrq() to simplify code
IB/qib: Drop qib_tune_pcie_caps() and qib_tune_pcie_coalesce() return values
IB/qib: Use pcie_set_mps() and pcie_get_mps() to simplify code
IB/qib: Use pci_is_root_bus() to check whether it is a root bus
tile/PCI: use cached pci_dev->pcie_mpss to simplify code
PCI: Export pcie_set_mps() and pcie_get_mps()
Previously, if kmalloc() failed, we claimed "PME# enabled" in dmesg,
even though we didn't add the device to the pci_pme_list. This prints
a more correct warning.
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Export pcie_get_mps() and pcie_set_mps() functions so drivers can use
them to simplify code.
Signed-off-by: Yijing Wang <wangyijing@huawei.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Pull networking changes from David Miller:
"Noteworthy changes this time around:
1) Multicast rejoin support for team driver, from Jiri Pirko.
2) Centralize and simplify TCP RTT measurement handling in order to
reduce the impact of bad RTO seeding from SYN/ACKs. Also, when
both timestamps and local RTT measurements are available prefer
the later because there are broken middleware devices which
scramble the timestamp.
From Yuchung Cheng.
3) Add TCP_NOTSENT_LOWAT socket option to limit the amount of kernel
memory consumed to queue up unsend user data. From Eric Dumazet.
4) Add a "physical port ID" abstraction for network devices, from
Jiri Pirko.
5) Add a "suppress" operation to influence fib_rules lookups, from
Stefan Tomanek.
6) Add a networking development FAQ, from Paul Gortmaker.
7) Extend the information provided by tcp_probe and add ipv6 support,
from Daniel Borkmann.
8) Use RCU locking more extensively in openvswitch data paths, from
Pravin B Shelar.
9) Add SCTP support to openvswitch, from Joe Stringer.
10) Add EF10 chip support to SFC driver, from Ben Hutchings.
11) Add new SYNPROXY netfilter target, from Patrick McHardy.
12) Compute a rate approximation for sending in TCP sockets, and use
this to more intelligently coalesce TSO frames. Furthermore, add
a new packet scheduler which takes advantage of this estimate when
available. From Eric Dumazet.
13) Allow AF_PACKET fanouts with random selection, from Daniel
Borkmann.
14) Add ipv6 support to vxlan driver, from Cong Wang"
Resolved conflicts as per discussion.
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next: (1218 commits)
openvswitch: Fix alignment of struct sw_flow_key.
netfilter: Fix build errors with xt_socket.c
tcp: Add missing braces to do_tcp_setsockopt
caif: Add missing braces to multiline if in cfctrl_linkup_request
bnx2x: Add missing braces in bnx2x:bnx2x_link_initialize
vxlan: Fix kernel panic on device delete.
net: mvneta: implement ->ndo_do_ioctl() to support PHY ioctls
net: mvneta: properly disable HW PHY polling and ensure adjust_link() works
icplus: Use netif_running to determine device state
ethernet/arc/arc_emac: Fix huge delays in large file copies
tuntap: orphan frags before trying to set tx timestamp
tuntap: purge socket error queue on detach
qlcnic: use standard NAPI weights
ipv6:introduce function to find route for redirect
bnx2x: VF RSS support - VF side
bnx2x: VF RSS support - PF side
vxlan: Notify drivers for listening UDP port changes
net: usbnet: update addr_assign_type if appropriate
driver/net: enic: update enic maintainers and driver
driver/net: enic: Exposing symbols for Cisco's low latency driver
...
* pci/misc:
PCI: Remove pcie_cap_has_devctl()
PCI: Support PCIe Capability Slot registers only for ports with slots
PCI: Remove PCIe Capability version checks
PCI: Allow PCIe Capability link-related register access for switches
PCI: Add offsets of PCIe capability registers
PCI: Tidy bitmasks and spacing of PCIe capability definitions
PCI: Remove obsolete comment reference to pci_pcie_cap2()
PCI: Clarify PCI_EXP_TYPE_PCI_BRIDGE comment
PCI: Rename PCIe capability definitions to follow convention
PCI: Disable decoding for BAR sizing only when it was actually enabled
PCI: Add comment about needing pci_msi_off() even when CONFIG_PCI_MSI=n
PCI: Add pcibios_pm_ops for optional arch-specific hibernate functionality
All other PCIe capability register fields include "PCI_EXP" + <reg-name> +
<field-name>. This renames PCI_EXP_OBFF_MASK, PCI_EXP_IDO_REQ_EN,
PCI_EXP_LTR_EN, and related fields using the same convention.
No functional change.
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Acked-by: Samuel Ortiz <sameo@linux.intel.com> # for MFD driver
* pci/yinghai-assign-unassigned-v6:
PCI: Assign resources for hot-added host bridge more aggressively
PCI: Move resource reallocation code to non-__init
PCI: Delay enabling bridges until they're needed
PCI: Assign resources on a per-bus basis
PCI: Enable unassigned resource reallocation on per-bus basis
PCI: Turn on reallocation for unassigned resources with host bridge offset
PCI: Look for unassigned resources on per-bus basis
PCI: Drop temporary variable in pci_assign_unassigned_resources()
Per f5f2b13129 ("msi: sanely support hardware level msi disabling"), we
want pci_msi_off() to work even if MSI support is not compiled into the
kernel, and there are existing callers that use it when CONFIG_PCI_MSI=n.
This adds a comment to that effect.
No functional change.
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
After 59875ae489 ("PCI/core: Use PCI Express Capability accessors"),
pcie_get_mps() never returns an error, so don't bother to check for it.
No functional change.
[bhelgaas: changelog, fix pcie_get_mps() doc]
Signed-off-by: Yijing Wang <wangyijing@huawei.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Users of pci_reset_bus() and pci_reset_slot() need a way to probe
whether the bus or slot supports reset. Add trivial helper functions
and export them as vfio-pci will make use of these.
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
The PCI spec indicates that with stable power, reset needs to be
asserted for a minimum of 1ms (Trst). We should be able to assume
stable power for a Hot Reset, but we add another millisecond as
a fudge factor to make sure the reset is seen on the bus for at least
a full 1ms.
After reset is de-asserted we must wait for devices to complete
initialization. The specs refer to this as "recovery time" (Trhfa).
For PCI this is 2^25 clock cycles or 2^26 for PCI-X. For minimum
bus speeds, both of those come to 1s. PCIe "softens" this
requirement with the Configuration Request Retry Status (CRS)
completion status. Theoretically we could use CRS to shorten the
wait time. We don't make use of that here, using a fixed 1s delay
to allow devices to re-initialize.
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Devices come out of reset in D0. Restoring a device to a different
post-reset state takes more smarts than our simple config space
restore, which can leave devices in an inconsistent state. For
example, if a device is reset in D3, but the restore doesn't
successfully return the device to D3, then the actual state of the
device and dev->current_state are contradictory. Put everything
in D0 going into the reset, then we don't need to do anything
special on the way out.
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Sometimes pci_reset_function() is not sufficient. We have cases where
devices do not support any kind of reset, but there might be multiple
functions on the bus preventing pci_reset_function() from doing a
secondary bus reset. We also have cases where a device will advertise
that it supports a PM reset, but really does nothing on D3hot->D0
(graphics cards are notorious for this). These devices often also
have more than one function, so even blacklisting PM reset for them
wouldn't allow a secondary bus reset through pci_reset_function().
If a driver supports multiple devices it should have the ability to
induce a bus reset when it needs to. This patch provides that ability
through pci_reset_slot() and pci_reset_bus(). It's the caller's
responsibility when using these interfaces to understand that all of
the devices in or below the slot (or on or below the bus) will be
reset and therefore should be under control of the caller. PCI state
of all the affected devices is saved and restored around these resets,
but internal state of all of the affected devices is reset (which
should be the intention).
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Only cosmetic code changes to existing paths. Expand the comment in
the new pci_dev_save_and_disable() function since there's a lot
hidden in that Command register write.
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
If the hotplug controller provides a way to reset a slot, use that
before a direct parent bus reset. Like the bus reset option, this is
only available when a single pci_dev occupies the slot.
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
* pci/vipul-chelsio-reset-v2:
PCI: Use pci_wait_for_pending_transaction() instead of for loop
bnx2x: Use pci_wait_for_pending_transaction() instead of for loop
PCI: Chelsio quirk: Enable Bus Master during Function-Level Reset
PCI: Add pci_wait_for_pending_transaction()
New routine to avoid duplication of code to wait for pending PCI
transactions to complete.
Signed-off-by: Casey Leedom <leedom@chelsio.com>
Signed-off-by: Vipul Pandya <vipul@chelsio.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Move the secondary bus reset code from pci_parent_bus_reset() into its own
function. Export it as we'll later be calling it from hotplug controllers
and elsewhere.
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
* pci/misc:
PCI: Fix comment typo for pci_add_cap_save_buffer()
PCI: Return -ENOSYS for SR-IOV operations on non-SR-IOV devices
PCI: Update NumVFs register when disabling SR-IOV
x86/PCI: MMCONFIG: Check earlier for MMCONFIG region at address zero
PCI: Convert class code to use dev_groups
frv/PCI: Mark pcibios_fixup_bus() as non-init
x86/pci/mrst: Cleanup checkpatch.pl warnings
PCI: Rename "PCI Express support" kconfig title
PCI: Fix comment typo in iov.c
A PCI Express device can potentially report a link width and speed which it will
not properly fulfill due to being plugged into a slower link higher in the
chain. This function walks up the PCI bus chain and calculates the minimum link
width and speed of this entire chain. This can be useful to enable a device to
determine if it has enough bandwidth for optimum functionality.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
We currently enable PCI bridges after scanning a bus and assigning
resources. This is often done in arch code.
This patch changes this so we don't enable a bridge until necessary, i.e.,
until we enable a PCI device behind the bridge. We do this in the generic
pci_enable_device() path, so this also removes the arch-specific code to
enable bridges.
[bhelgaas: changelog]
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
We currently misinterpret that in order for an ACS feature to be
enabled it must be set in the control field. In reality, this means
that the feature is not only enabled, but controllable. Many of the
ACS capability bits are not required if the device behaves by default
in the way specified when both the capability and control bit are set
and does not support or allow the alternate mode. We therefore need
to check the capabilities and mask out flags that are enabled but not
controllable. Egress control seems to be the only flag which is
purely optional.
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Acked-by: Donald Dutile <ddutile@redhat.com>
The multifunction ACS rules do not apply to downstream ports. Those
should be tested regardless of whether they are single function or
multifunction. The PCIe spec also fully specifies which PCIe types
are subject to the multifunction rules and excludes event collectors
and PCIe-to-PCI bridges entirely. Document each rule to the section
of the PCIe spec and provide overall documentation of the function.
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Acked-by: Donald Dutile <ddutile@redhat.com>
PCI PM cap register offset has been saved in pci_pm_init(),
so we can use pdev->pm_cap instead of using pci_find_capability(..)
here.
Signed-off-by: Yijing Wang <wangyijing@huawei.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Platforms may want to provide architecture-specific functionality when
a PCI device is released. Add a pcibios_release_device() call that
architectures can override to do so.
Signed-off-by: Sebastian Ott <sebott@linux.vnet.ibm.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Linus Torvalds