Device set same 256 bytes maximum read request size equal MAX_READ_REQ_SIZE
was find some issue with HDMI scrambled picture and nvme devices
at intensive writing...

[    4.798971] nvme 0000:01:00.0: fix MRRS from 512 to 256

This quirk setup same MRRS if we try solve this problem with
pci=pcie_bus_perf kernel command line param

Signed-off-by: Artem Lapkin <art@khadas.com>

- Annotate __iomem pointer (Bjorn Helgaas)

- Fix kernel-doc comment (Krzysztof Wilczyński)

* remotes/lorenzo/pci/xgene:
  PCI: xgene: Fix a non-compliant kernel-doc
  PCI: xgene: Annotate __iomem pointer

- Fix host init during resume (Vidya Sagar)

- Add missing MODULE_DEVICE_TABLE for driver loading (Zou Wei)

* remotes/lorenzo/pci/tegra:
  PCI: tegra: Add missing MODULE_DEVICE_TABLE
  PCI: tegra: Fix host initialization during resume

- Removed unused readl and writel functions (Krzysztof Wilczyński)

* remotes/lorenzo/pci/mobiveil:
  PCI: mobiveil: Remove unused readl and writel functions

- Make struct event_descs static (Krzysztof Wilczyński)

* remotes/lorenzo/pci/microchip:
  PCI: microchip: Make the struct event_descs static

- Drop redundant error message from mtk_pcie_subsys_powerup() (Zhen Lei)

* remotes/lorenzo/pci/mediatek:
  PCI: mediatek: Remove redundant error printing in mtk_pcie_subsys_powerup()

- Fix race when removing device (Long Li)

- Remove unused bus device removal refcount/functions (Long Li)

* remotes/lorenzo/pci/hv:
  PCI: hv: Remove bus device removal unused refcount/functions
  PCI: hv: Fix a race condition when removing the device

- Rename PCI_IOSIZE, etc in ftpci100 to avoid macro name collisions (Randy
  Dunlap)

* remotes/lorenzo/pci/ftpci100:
  PCI: ftpci100: Rename macro name collision

- Drop redundant error message from imx6_pcie_probe() (Zhen Lei)

* remotes/lorenzo/pci/dwc:
  PCI: dwc/imx6: Remove redundant error printing in imx6_pcie_probe()

- Clarify error message for unbound devices (Moritz Fischer)

* pci/virtualization:
  PCI/IOV: Clarify error message for unbound devices

- Fix dsm_label_utf16s_to_utf8s() buffer overrun (Krzysztof Wilczyński)

- Use sysfs_emit() and sysfs_emit_at() in "show" functions (Krzysztof
  Wilczyński)

- Fix 'resource_alignment' newline issues (Krzysztof Wilczyński)

- Add newline to 'devspec' sysfs file (Krzysztof Wilczyński)

* pci/sysfs:
  PCI/sysfs: Add 'devspec' newline
  PCI/sysfs: Fix 'resource_alignment' newline issues
  PCI/sysfs: Use sysfs_emit() and sysfs_emit_at() in "show" functions
  PCI/sysfs: Rely on lengths from scnprintf(), dsm_label_utf16s_to_utf8s()
  PCI/sysfs: Fix dsm_label_utf16s_to_utf8s() buffer overrun

- Coalesce host bridge apertures so we can allocate large BARs that cross
  contiguous apertures (Kai-Heng Feng)

* pci/resource:
  PCI: Coalesce host bridge contiguous apertures

- Leave Apple Thunderbolt controllers on for s2idle or standby so they work
  after resume (Konstantin Kharlamov)

* pci/pm:
  PCI: Leave Apple Thunderbolt controllers on for s2idle or standby

- Rename Rename upstream_bridge_distance() to calc_map_type_and_dist()
  (Logan Gunthorpe)

- Collect ACS list message in stack buffer to avoid sleeping (Logan
  Gunthorpe)

- Use correct calc_map_type_and_dist() return type (Logan Gunthorpe)

- Warn if host bridge not in whitelist (Logan Gunthorpe)

- Refactor pci_p2pdma_map_type() (Logan Gunthorpe)

- Avoid pci_get_slot(), which may sleep (Logan Gunthorpe)

- Simplify distance calculation in __calc_map_type_and_dist() and
  calc_map_type_and_dist_warn() (Christoph Hellwig)

* pci/p2pdma:
  PCI/P2PDMA: Simplify distance calculation
  PCI/P2PDMA: Avoid pci_get_slot(), which may sleep
  PCI/P2PDMA: Refactor pci_p2pdma_map_type()
  PCI/P2PDMA: Warn if host bridge not in whitelist
  PCI/P2PDMA: Use correct calc_map_type_and_dist() return type
  PCI/P2PDMA: Collect acs list in stack buffer to avoid sleeping
  PCI/P2PDMA: Rename upstream_bridge_distance() and rework doc

- Return true/false (not 1/0) from bool functions (Yang Li)

* pci/misc:
  x86/pci: Return true/false (not 1/0) from bool functions

- Ignore pciehp Link Down/Up caused by DPC so device remains bound to
  driver (Lukas Wunner)

* pci/hotplug:
  PCI: pciehp: Ignore Link Down/Up caused by DPC

- Make domain/bus/dev/fn format in AER messages match pci_name() format
  (Yicong Yang)

* pci/error:
  Documentation: PCI: Fix typo in pci-error-recovery.rst
  PCI/AER: Use consistent format when printing PCI device

- Add a debug message on PCI device release (Niklas Schnelle)

- Dynamically map ECAM regions to reduce vmalloc usage (Russell King)

* pci/enumeration:
  PCI: Dynamically map ECAM regions
  PCI: Print a debug message on PCI device release

Attempting to boot 32-bit ARM kernels under QEMU's 3.x virt models fails
when we have more than 512M of RAM in the model as we run out of vmalloc
space for the PCI ECAM regions. This failure will be silent when running
libvirt, as the console in that situation is a PCI device.

In this configuration, the kernel maps the whole ECAM, which QEMU sets up
for 256 buses, even when maybe only seven buses are in use.  Each bus uses
1M of ECAM space, and ioremap() adds an additional guard page between
allocations. The kernel vmap allocator will align these regions to 512K,
resulting in each mapping eating 1.5M of vmalloc space. This means we need
384M of vmalloc space just to map all of these, which is very wasteful of
resources.

Fix this by only mapping the ECAM for buses we are going to be using.  In
my setups, this is around seven buses in most guests, which is 10.5M of
vmalloc space - way smaller than the 384M that would otherwise be required.
This also means that the kernel can boot without forcing extra RAM into
highmem with the vmalloc= argument, or decreasing the virtual RAM available
to the guest.

Suggested-by: Arnd Bergmann <arnd@arndb.de>
Link: https://lore.kernel.org/r/E1lhCAV-0002yb-50@rmk-PC.armlinux.org.uk
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Arnd Bergmann <arnd@arndb.de>

Downstream Port Containment (PCIe r5.0, sec. 6.2.10) disables the link upon
an error and attempts to re-enable it when instructed by the DPC driver.

A slot which is both DPC- and hotplug-capable is currently powered off by
pciehp once DPC is triggered (due to the link change) and powered back up
on successful recovery.  That's undesirable, the slot should remain powered
so the hotplugged device remains bound to its driver.  DPC notifies the
driver of the error and of successful recovery in pcie_do_recovery() and
the driver may then restore the device to working state.

Moreover, Sinan points out that turning off slot power by pciehp may foil
recovery by DPC:  Power off/on is a cold reset concurrently to DPC's warm
reset.  Sathyanarayanan reports extended delays or failure in link
retraining by DPC if pciehp brings down the slot.

Fix by detecting whether a Link Down event is caused by DPC and awaiting
recovery if so.  On successful recovery, ignore both the Link Down and the
subsequent Link Up event.

Afterwards, check whether the link is down to detect surprise-removal or
another DPC event immediately after DPC recovery.  Ensure that the
corresponding DLLSC event is not ignored by synthesizing it and invoking
irq_wake_thread() to trigger a re-run of pciehp_ist().

The IRQ threads of the hotplug and DPC drivers, pciehp_ist() and
dpc_handler(), race against each other.  If pciehp is faster than DPC, it
will wait until DPC recovery completes.

Recovery consists of two steps:  The first step (waiting for link
disablement) is recognizable by pciehp through a set DPC Trigger Status
bit.  The second step (waiting for link retraining) is recognizable through
a newly introduced PCI_DPC_RECOVERING flag.

If DPC is faster than pciehp, neither of the two flags will be set and
pciehp may glean the recovery status from the new PCI_DPC_RECOVERED flag.
The flag is zero if DPC didn't occur at all, hence DLLSC events are not
ignored by default.

pciehp waits up to 4 seconds before assuming that DPC recovery failed and
bringing down the slot.  This timeout is not taken from the spec (it
doesn't mandate one) but based on a report from Yicong Yang that DPC may
take a bit more than 3 seconds on HiSilicon's Kunpeng platform.

The timeout is necessary because the DPC Trigger Status bit may never
clear:  On Root Ports which support RP Extensions for DPC, the DPC driver
polls the DPC RP Busy bit for up to 1 second before giving up on DPC
recovery.  Without the timeout, pciehp would then wait indefinitely for DPC
to complete.

This commit draws inspiration from previous attempts to synchronize DPC
with pciehp:

By Sinan Kaya, August 2018:
https://lore.kernel.org/linux-pci/20180818065126.77912-1-okaya@kernel.org/

By Ethan Zhao, October 2020:
https://lore.kernel.org/linux-pci/20201007113158.48933-1-haifeng.zhao@intel.com/

By Kuppuswamy Sathyanarayanan, March 2021:
https://lore.kernel.org/linux-pci/59cb30f5e5ac6d65427ceaadf1012b2ba8dbf66c.1615606143.git.sathyanarayanan.kuppuswamy@linux.intel.com/

Link: https://lore.kernel.org/r/0be565d97438fe2a6d57354b3aa4e8626952a00b.1619857124.git.lukas@wunner.de
Reported-by: Sinan Kaya <okaya@kernel.org>
Reported-by: Ethan Zhao <haifeng.zhao@intel.com>
Reported-by: Kuppuswamy Sathyanarayanan <sathyanarayanan.kuppuswamy@linux.intel.com>
Tested-by: Kuppuswamy Sathyanarayanan <sathyanarayanan.kuppuswamy@linux.intel.com>
Tested-by: Yicong Yang <yangyicong@hisilicon.com>
Signed-off-by: Lukas Wunner <lukas@wunner.de>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Kuppuswamy Sathyanarayanan <sathyanarayanan.kuppuswamy@linux.intel.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Ashok Raj <ashok.raj@intel.com>
Cc: Keith Busch <kbusch@kernel.org>

Merge __calc_map_type_and_dist() and calc_map_type_and_dist_warn() into
calc_map_type_and_dist() to simplify the code a bit.  This now means we add
the devfn strings to the acs_buf unconditionally even if the buffer is not
printed, but that is not a lot of overhead and keeps the code much simpler.

Link: https://lore.kernel.org/r/20210614055310.3960791-1-hch@lst.de
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Logan Gunthorpe <logang@deltatee.com>

In order to use upstream_bridge_distance_warn() from a dma_map function, it
must not sleep. However, pci_get_slot() takes the pci_bus_sem so it might
sleep.

In order to avoid this, try to get the host bridge's device from the first
element in the device list. It should be impossible for the host bridge's
device to go away while references are held on child devices, so the first
element should not be able to change and, thus, this should be safe.

Introduce a static function called pci_host_bridge_dev() to obtain the host
bridge's root device.

Link: https://lore.kernel.org/r/20210610160609.28447-7-logang@deltatee.com
Signed-off-by: Logan Gunthorpe <logang@deltatee.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>

All callers of pci_p2pdma_map_type() have a struct dev_pgmap and a struct
device (of the client doing the DMA transfer). Thus move the conversion to
struct pci_devs for the provider and client into this function.

Link: https://lore.kernel.org/r/20210610160609.28447-6-logang@deltatee.com
Signed-off-by: Logan Gunthorpe <logang@deltatee.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>

If the host bridge is not in the whitelist print a warning in the
calc_map_type_and_dist_warn() path detailing the vendor and device IDs that
would need to be added to the whitelist.

Suggested-by: Don Dutile <ddutile@redhat.com>
Link: https://lore.kernel.org/r/20210610160609.28447-5-logang@deltatee.com
Signed-off-by: Logan Gunthorpe <logang@deltatee.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>

Instead of using an int for the return value of this function, use the
correct enum pci_p2pdma_map_type.

Link: https://lore.kernel.org/r/20210610160609.28447-4-logang@deltatee.com
Signed-off-by: Logan Gunthorpe <logang@deltatee.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>

In order to call the calc_map_type_and_dist_warn() function from a dma_map
operation, the function must not sleep. The only reason it sleeps is to
allocate memory for the seq_buf to print a verbose warning telling the user
how to disable ACS for that path.

Instead of allocating the memory with kmalloc(), allocate a smaller buffer
on the stack. A 128 byte buffer is enough to print 10 PCI device names. A
system with 10 bridge ports between two devices that have ACS enabled would
be unusually large, so this should still be a reasonable limit.

This also cleans up the awkward (and broken) return with -ENOMEM which
contradicts the return type and the caller was not prepared for.

Link: https://lore.kernel.org/r/20210610160609.28447-3-logang@deltatee.com
Signed-off-by: Logan Gunthorpe <logang@deltatee.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>

The function upstream_bridge_distance() has evolved such that its name is
no longer entirely reflective of what the function does.  It not only
calculates the distance between two peers but also calculates how the DMA
addresses for those two peers should be mapped.

Rename it to calc_map_type_and_dist() and rework the documentation to
better describe the two pieces of information the function returns.

[bhelgaas: tweak comment wording]
Link: https://lore.kernel.org/r/20210610160609.28447-2-logang@deltatee.com
Signed-off-by: Logan Gunthorpe <logang@deltatee.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>

Be more verbose to disambiguate the error case when trying to configure
SR-IOV with no driver bound vs. a driver that does not implement the
.sriov_configure() callback.

Link: https://lore.kernel.org/r/20210327175140.682708-1-mdf@kernel.org
Reported-by: Brian Foley <bpfoley@google.com>
Signed-off-by: Moritz Fischer <mdf@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Krzysztof Wilczyński <kw@linux.com>

On Macbook 2013, resuming from suspend-to-idle or standby resulted in the
external monitor no longer being detected, a stacktrace, and errors like
this in dmesg:

  pcieport 0000:06:00.0: can't change power state from D3hot to D0 (config space inaccessible)

The reason is that we know how to turn power to the Thunderbolt controller
*off* via the SXIO/SXFP/SXLF methods, but we don't know how to turn power
back on.  We have to rely on firmware to turn the power back on.

When going to the "suspend-to-idle" or "standby" system sleep states,
firmware is not involved either on the suspend side or the resume side, so
we can't use SXIO/SXFP/SXLF to turn the power off.

Skip SXIO/SXFP/SXLF when firmware isn't involved in suspend, e.g., when
we're going to the "suspend-to-idle" or "standby" system sleep states.

Fixes: 1df5172 ("PCI: Suspend/resume quirks for Apple thunderbolt")
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=212767
Link: https://lore.kernel.org/r/20210520235501.917397-1-Hi-Angel@yandex.ru
Signed-off-by: Konstantin Kharlamov <Hi-Angel@yandex.ru>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Lukas Wunner <lukas@wunner.de>
Cc: stable@vger.kernel.org

Built-in graphics on HP EliteDesk 805 G6 doesn't work because graphics
can't get the BAR it needs:

  pci_bus 0000:00: root bus resource [mem 0x10020200000-0x100303fffff window]
  pci_bus 0000:00: root bus resource [mem 0x10030400000-0x100401fffff window]

  pci 0000:00:08.1:   bridge window [mem 0xd2000000-0xd23fffff]
  pci 0000:00:08.1:   bridge window [mem 0x10030000000-0x100401fffff 64bit pref]
  pci 0000:00:08.1: can't claim BAR 15 [mem 0x10030000000-0x100401fffff 64bit pref]: no compatible bridge window
  pci 0000:00:08.1: [mem 0x10030000000-0x100401fffff 64bit pref] clipped to [mem 0x10030000000-0x100303fffff 64bit pref]
  pci 0000:00:08.1:   bridge window [mem 0x10030000000-0x100303fffff 64bit pref]
  pci 0000:07:00.0: can't claim BAR 0 [mem 0x10030000000-0x1003fffffff 64bit pref]: no compatible bridge window
  pci 0000:07:00.0: can't claim BAR 2 [mem 0x10040000000-0x100401fffff 64bit pref]: no compatible bridge window

However, the root bus has two contiguous apertures that can contain the
child resource requested.

Coalesce contiguous apertures so we can allocate from the entire contiguous
region.

[bhelgaas: fold in https://lore.kernel.org/r/20210528170242.1564038-1-kai.heng.feng@canonical.com]
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=212013
Suggested-by: Bjorn Helgaas <bhelgaas@google.com>
Link: https://lore.kernel.org/r/20210401131252.531935-1-kai.heng.feng@canonical.com
Signed-off-by: Kai-Heng Feng <kai.heng.feng@canonical.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>

Previously, when the value of the "devspec" sysfs attribute was read from
the user space there was no newline present, and utilities such as "cat"
wouldn't display the result of the read correctly.

Append a newline character in the show() function to match other "devspec"
attributes.

Link: https://lore.kernel.org/r/20210603000112.703037-5-kw@linux.com
Signed-off-by: Krzysztof Wilczyński <kw@linux.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Logan Gunthorpe <logang@deltatee.com>

The value of the "resource_alignment" can be specified using a kernel
command-line argument ("pci=resource_alignment=") or through the
corresponding sysfs attribute under the /sys/bus/pci path.

Previously, when the value was set via the kernel command-line argument,
and then subsequently accessed through sysfs attribute, the value read back
was not correct:

  # grep -oE 'pci=resource_alignment.+' /proc/cmdline
  pci=resource_alignment=20@00:1f.2
  # cat /sys/bus/pci/resource_alignment
  20@00:1f.

This was also true when the value was set through the sysfs attribute
without including a trailing newline:

  # echo -n 20@00:1f.2 > /sys/bus/pci/resource_alignment
  # cat /sys/bus/pci/resource_alignment
  20@00:1f.

When it was set through the sysfs attribute *including* a newline,
reading it back worked as intended:

  # echo 20@00:1f.2 > /sys/bus/pci/resource_alignment
  # cat /sys/bus/pci/resource_alignment
  20@00:1f.2

To fix this inconsistency, append a trailing newline in the show() function
and strip the trailing line in the store() function if one is present.

Also, allow for the value previously set using either a command-line
argument or through the sysfs object to be cleared at run-time.

[bhelgaas: fold in kfree fix from
https://lore.kernel.org/linux-pci/20210604133230.983956-4-kw@linux.com]
Fixes: e499081 ("PCI: Force trailing new line to resource_alignment_param in sysfs")
Link: https://lore.kernel.org/r/20210603000112.703037-4-kw@linux.com
Signed-off-by: Krzysztof Wilczyński <kw@linux.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Logan Gunthorpe <logang@deltatee.com>

The sysfs_emit() and sysfs_emit_at() functions were introduced to make
it less ambiguous which function is preferred when writing to the output
buffer in a device attribute's "show" callback [1].

Convert the PCI sysfs object "show" functions from sprintf(), snprintf()
and scnprintf() to sysfs_emit() and sysfs_emit_at() accordingly, as the
latter is aware of the PAGE_SIZE buffer and correctly returns the number
of bytes written into the buffer.

No functional change intended.

[1] Documentation/filesystems/sysfs.rst

Related commit: ad025f8 ("PCI/sysfs: Use sysfs_emit() and
sysfs_emit_at() in "show" functions").

Link: https://lore.kernel.org/r/20210603000112.703037-2-kw@linux.com
Signed-off-by: Krzysztof Wilczyński <kw@linux.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Logan Gunthorpe <logang@deltatee.com>

scnprintf() returns the number of bytes written into the buffer.  Change
dsm_label_utf16s_to_utf8s() to do the same.  Rely on those values instead
of using strlen() to compute the buffer length.

No functional change intended.

[bhelgaas: reorder patch in series, len++ to include newline added by
dsm_label_utf16s_to_utf8s(), commit log]
Link: https://lore.kernel.org/r/20210603000112.703037-3-kw@linux.com
Signed-off-by: Krzysztof Wilczyński <kw@linux.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Logan Gunthorpe <logang@deltatee.com>

"utf16s_to_utf8s(..., buf, PAGE_SIZE)" puts up to PAGE_SIZE bytes into
"buf" and returns the number of bytes it actually put there.  If it wrote
PAGE_SIZE bytes, the newline added by dsm_label_utf16s_to_utf8s() would
overrun "buf".

Reduce the size available for utf16s_to_utf8s() to use so there is always
space for the newline.

[bhelgaas: reorder patch in series, commit log]
Fixes: 6058989 ("PCI: Export ACPI _DSM provided firmware instance number and string name to sysfs")
Link: https://lore.kernel.org/r/20210603000112.703037-7-kw@linux.com
Reported-by: Joe Perches <joe@perches.com>
Signed-off-by: Krzysztof Wilczyński <kw@linux.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>