Add the ability to handle USB wake events from USB devices when
in S2 mode. Typically there is some additional configuration
needed to tell the USB device to generate the wake event when
suspended but this varies with the different USB device classes.
For example, on USB Ethernet dongles, ethtool should be used to
enable the magic packet wake functionality in the dongle.
NOTE:  This requires that the "power/wakeup" sysfs entry for
the USB device generating the wakeup be set to "enabled".

This functionality requires a special hardware sideband path that
will trigger the AON_PM_L2 interrupt needed to wake the system from
S2 even though the USB host controllers are in IDDQ (low power state)
and most USB related clocks are shut off. For the sideband signaling
to work we need to leave the usbx_freerun clock running, but this
clock consumes very little power by design. There's a bug in the
XHCI wake hardware so only EHCI/OHCI wake is currently supported.
Signed-off-by: Al Cooper <alcooperx@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>

Currently the Phy driver will put the USB phys into the max
power saving mode (IDDQ) when there is no corresponding XHCI, EHCI
or OHCI client (through rmmod, unbind or if the driver is not
builtin). This change will also put the Phys into IDDQ mode
on suspend so that S2 will get the additional power savings.
Signed-off-by: Al Cooper <alcooperx@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>

The BRCM USB Phy, ohci, ehci and xhci drivers all use the USB clocks
but not all drivers use the clk_prepare_enable/clk_disable_unprepare
versions to enable/disable the clocks. This change gets all drivers
using the prepare version.
Signed-off-by: Al Cooper <alcooperx@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>

When the EHCI controller received a 512 byte USB packet that
had to be broken into 2 256 byte bursts across the SCB bus AND
there was a following 512 byte USB packet, the second burst of
data from the first packet was sometimes being lost. If the
burst size was changed to 128 bytes via the EBR_SCB_SIZE field
in the USB_CTRL_EBRIDGE register we'd see the 4th 128 byte burst
of the first packet being lost. This problem became much worse
if other threads were running that accessed memory, like a memcpy
test. Setting the EBR_SCB_SIZE to 512, which prevents breaking
the EHCI USB packet (max size of 512 bytes) into bursts, fixed
the problem.
Signed-off-by: Al Cooper <alcooperx@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>

For V4 QMP UFS Phy, we need to assert reset bits, configure the phy and
then deassert it, so add the QPHY_SW_RESET register which does this.
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Reviewed-by: Manu Gautam <mgautam@codeaurora.org>
Reviewed-by: Can Guo <cang@codeaurora.org>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>

SM8150 QMPY phy for UFS and onwards the PHY_SW_RESET is present in PHY's
PCS register so we should not mark no_pcs_sw_reset for sm8150 and
onwards
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Reviewed-by: Manu Gautam <mgautam@codeaurora.org>
Reviewed-by: Can Guo <cang@codeaurora.org>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>

We already write to QPHY_POWER_DOWN_CONTROL in qcom_qmp_phy_com_init()
before invoking qcom_qmp_phy_configure() so remove the duplicate write.
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Reviewed-by: Manu Gautam <mgautam@codeaurora.org>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>

We already define register offsets so use them in register layout.
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Reviewed-by: Manu Gautam <mgautam@codeaurora.org>
Reviewed-by: Can Guo <cang@codeaurora.org>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>

7216 is a 16nm process chip with a slightly different version of the PHY
SerdDeS/AFE that requires a specific tuning sequence. Key on the
compatible string to perform that initialization.
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>

Define "brcm,bcm7216-sata-phy" as a new compatible string for the
Broadcom SATA3 PHY.
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>

During review it came to light that exposing the pll clock outside is
not the right approach and struct phy_configure_opts_mipi_dphy exists
just for that reason to transfer parameters to the phy.

So drop the exposed clock and rely on the phy configure options
to bring in the correct rate. That way we can also just drop the
open coded timing struct and default values function.

Fixes: b7535a3b ("phy/rockchip: Add support for Innosilicon MIPI/LVDS/TTL PHY")
Signed-off-by: Heiko Stuebner <heiko.stuebner@theobroma-systems.com>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>

Further review of the dsi components for the px30 revealed that the
phy shouldn't expose the pll as clock but instead handle settings
via phy parameters.

As the phy binding is new and not used anywhere yet, just drop them
so they don't get used.

Fixes: 3817c796 ("dt-bindings: phy: add yaml binding for rockchip,px30-dsi-dphy")
Signed-off-by: Heiko Stuebner <heiko.stuebner@theobroma-systems.com>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>

Fixes coccicheck warning:

drivers/phy/lantiq/phy-lantiq-vrx200-pcie.c:389:2-3: Unneeded semicolon

Fixes: e52a6321 ("phy: lantiq: vrx200-pcie: add a driver for the Lantiq VRX200 PCIe PHY")
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Ma Feng <mafeng.ma@huawei.com>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>

Clang warns:

../drivers/phy/qualcomm/phy-qcom-apq8064-sata.c:83:4: warning:
misleading indentation; statement is not part of the previous 'if'
[-Wmisleading-indentation]
                 usleep_range(DELAY_INTERVAL_US, DELAY_INTERVAL_US + 50);
                 ^
../drivers/phy/qualcomm/phy-qcom-apq8064-sata.c:80:3: note: previous
statement is here
                if (readl_relaxed(addr) & mask)
                ^
1 warning generated.

This warning occurs because there is a space after the tab on this line.
Remove it so that the indentation is consistent with the Linux kernel
coding style and clang no longer warns.

Fixes: 1de990d8 ("phy: qcom: Add driver for QCOM APQ8064 SATA PHY")
Link: https://github.com/ClangBuiltLinux/linux/issues/816Signed-off-by: Nathan Chancellor <natechancellor@gmail.com>
Reviewed-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>

In order to enforce suspend/resume ordering, this commit creates link
between phy consumers and phy devices. This link avoids to suspend phy
before phy consumers.
Signed-off-by: Alexandre Torgue <alexandre.torgue@st.com>
[jonathanh@nvidia.com: Fix an abort when of_phy_get() returns error]
Signed-off-by: Jonathan Hunter <jonathanh@nvidia.com>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>

The driver calls clk_prepare_enable in probe but the corresponding
clk_disable_unprepare() is in ti_pipe3_disable_clocks().
Move clk_disable_unprepare() to remove to make them symmetric.
Signed-off-by: Chuhong Yuan <hslester96@gmail.com>
Acked-by: Roger Quadros <rogerq@ti.com>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>

This patch corrects the SPDX License Identifier style in
header files related to PHY drivers for Qualcomm platforms.
For C header files Documentation/process/license-rules.rst
mandates C-like comments (opposed to C source files where
C++ style should be used).

Changes made by using a script provided by Joe Perches here:
https://lkml.org/lkml/2019/2/7/46.
Suggested-by: Joe Perches <joe@perches.com>
Signed-off-by: Nishad Kamdar <nishadkamdar@gmail.com>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>

Adjust indentation from spaces to tab (+optional two spaces) as in
coding style with command like:
	$ sed -e 's/^        /\t/' -i */Kconfig
Signed-off-by: Krzysztof Kozlowski <krzk@kernel.org>
Signed-off-by: Kishon Vijay Abraham I <kishon@ti.com>

Pull networking fixes from David Miller:

 1) More jumbo frame fixes in r8169, from Heiner Kallweit.

 2) Fix bpf build in minimal configuration, from Alexei Starovoitov.

 3) Use after free in slcan driver, from Jouni Hogander.

 4) Flower classifier port ranges don't work properly in the HW offload
    case, from Yoshiki Komachi.

 5) Use after free in hns3_nic_maybe_stop_tx(), from Yunsheng Lin.

 6) Out of bounds access in mqprio_dump(), from Vladyslav Tarasiuk.

 7) Fix flow dissection in dsa TX path, from Alexander Lobakin.

 8) Stale syncookie timestampe fixes from Guillaume Nault.

[ Did an evil merge to silence a warning introduced by this pull - Linus ]

* git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (84 commits)
  r8169: fix rtl_hw_jumbo_disable for RTL8168evl
  net_sched: validate TCA_KIND attribute in tc_chain_tmplt_add()
  r8169: add missing RX enabling for WoL on RTL8125
  vhost/vsock: accept only packets with the right dst_cid
  net: phy: dp83867: fix hfs boot in rgmii mode
  net: ethernet: ti: cpsw: fix extra rx interrupt
  inet: protect against too small mtu values.
  gre: refetch erspan header from skb->data after pskb_may_pull()
  pppoe: remove redundant BUG_ON() check in pppoe_pernet
  tcp: Protect accesses to .ts_recent_stamp with {READ,WRITE}_ONCE()
  tcp: tighten acceptance of ACKs not matching a child socket
  tcp: fix rejected syncookies due to stale timestamps
  lpc_eth: kernel BUG on remove
  tcp: md5: fix potential overestimation of TCP option space
  net: sched: allow indirect blocks to bind to clsact in TC
  net: core: rename indirect block ingress cb function
  net-sysfs: Call dev_hold always in netdev_queue_add_kobject
  net: dsa: fix flow dissection on Tx path
  net/tls: Fix return values to avoid ENOTSUPP
  net: avoid an indirect call in ____sys_recvmsg()
  ...

Pull more SCSI updates from James Bottomley:
 "Eleven patches, all in drivers (no core changes) that are either minor
  cleanups or small fixes.

  They were late arriving, but still safe for -rc1"

* tag 'scsi-misc' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
  scsi: MAINTAINERS: Add the linux-scsi mailing list to the ISCSI entry
  scsi: megaraid_sas: Make poll_aen_lock static
  scsi: sd_zbc: Improve report zones error printout
  scsi: qla2xxx: Fix qla2x00_request_irqs() for MSI
  scsi: qla2xxx: unregister ports after GPN_FT failure
  scsi: qla2xxx: fix rports not being mark as lost in sync fabric scan
  scsi: pm80xx: Remove unused include of linux/version.h
  scsi: pm80xx: fix logic to break out of loop when register value is 2 or 3
  scsi: scsi_transport_sas: Fix memory leak when removing devices
  scsi: lpfc: size cpu map by last cpu id set
  scsi: ibmvscsi_tgt: Remove unneeded variable rc

Pull cifs fixes from Steve French:
 "Nine cifs/smb3 fixes:

   - one fix for stable (oops during oplock break)

   - two timestamp fixes including important one for updating mtime at
     close to avoid stale metadata caching issue on dirty files (also
     improves perf by using SMB2_CLOSE_FLAG_POSTQUERY_ATTRIB over the
     wire)

   - two fixes for "modefromsid" mount option for file create (now
     allows mode bits to be set more atomically and accurately on create
     by adding "sd_context" on create when modefromsid specified on
     mount)

   - two fixes for multichannel found in testing this week against
     different servers

   - two small cleanup patches"

* tag '5.5-rc-smb3-fixes-part2' of git://git.samba.org/sfrench/cifs-2.6:
  smb3: improve check for when we send the security descriptor context on create
  smb3: fix mode passed in on create for modetosid mount option
  cifs: fix possible uninitialized access and race on iface_list
  cifs: Fix lookup of SMB connections on multichannel
  smb3: query attributes on file close
  smb3: remove unused flag passed into close functions
  cifs: remove redundant assignment to pointer pneg_ctxt
  fs: cifs: Fix atime update check vs mtime
  CIFS: Fix NULL-pointer dereference in smb2_push_mandatory_locks

Pull misc vfs cleanups from Al Viro:
 "No common topic, just three cleanups".

* 'work.misc' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
  make __d_alloc() static
  fs/namespace: add __user to open_tree and move_mount syscalls
  fs/fnctl: fix missing __user in fcntl_rw_hint()

Pull NTB update from Jon Mason:
 "Just a simple patch to add a new Hygon Device ID to the AMD NTB device
  driver"

* tag 'ntb-5.5' of git://github.com/jonmason/ntb:
  NTB: Add Hygon Device ID

Pull more input updates from Dmitry Torokhov:

 - fixups for Synaptics RMI4 driver

 - a quirk for Goodinx touchscreen on Teclast tablet

 - a new keycode definition for activating privacy screen feature found
   on a few "enterprise" laptops

 - updates to snvs_pwrkey driver

 - polling uinput device for writing (which is always allowed) now works

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input:
  Input: synaptics-rmi4 - don't increment rmiaddr for SMBus transfers
  Input: synaptics-rmi4 - re-enable IRQs in f34v7_do_reflash
  Input: goodix - add upside-down quirk for Teclast X89 tablet
  Input: add privacy screen toggle keycode
  Input: uinput - fix returning EPOLLOUT from uinput_poll
  Input: snvs_pwrkey - remove gratuitous NULL initializers
  Input: snvs_pwrkey - send key events for i.MX6 S, DL and Q

Pull iomap fixes from Darrick Wong:
 "Fix a race condition and a use-after-free error:

   - Fix a UAF when reporting writeback errors

   - Fix a race condition when handling page uptodate on fragmented file
     with blocksize < pagesize"

* tag 'iomap-5.5-merge-14' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux:
  iomap: stop using ioend after it's been freed in iomap_finish_ioend()
  iomap: fix sub-page uptodate handling

Pull xfs fixes from Darrick Wong:
 "Fix a couple of resource management errors and a hang:

   - fix a crash in the log setup code when log mounting fails

   - fix a hang when allocating space on the realtime device

   - fix a block leak when freeing space on the realtime device"

* tag 'xfs-5.5-merge-17' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux:
  xfs: fix mount failure crash on invalid iclog memory access
  xfs: don't check for AG deadlock for realtime files in bunmapi
  xfs: fix realtime file data space leak

Pull orangefs update from Mike Marshall:
 "orangefs: posix open permission checking...

  Orangefs has no open, and orangefs checks file permissions on each
  file access. Posix requires that file permissions be checked on open
  and nowhere else. Orangefs-through-the-kernel needs to seem posix
  compliant.

  The VFS opens files, even if the filesystem provides no method. We can
  see if a file was successfully opened for read and or for write by
  looking at file->f_mode.

  When writes are flowing from the page cache, file is no longer
  available. We can trust the VFS to have checked file->f_mode before
  writing to the page cache.

  The mode of a file might change between when it is opened and IO
  commences, or it might be created with an arbitrary mode.

  We'll make sure we don't hit EACCES during the IO stage by using
  UID 0"

[ This is "posixish", but not a great solution in the long run, since a
  proper secure network server shouldn't really trust the client like this.
  But proper and secure POSIX behavior requires an open method and a
  resulting cookie for IO of some kind, or similar.    - Linus ]

* tag 'for-linus-5.5-ofs1' of git://git.kernel.org/pub/scm/linux/kernel/git/hubcap/linux:
  orangefs: posix open permission checking...

Pull nfsd updates from Bruce Fields:
 "This is a relatively quiet cycle for nfsd, mainly various bugfixes.

  Possibly most interesting is Trond's fixes for some callback races
  that were due to my incomplete understanding of rpc client shutdown.
  Unfortunately at the last minute I've started noticing a new
  intermittent failure to send callbacks. As the logic seems basically
  correct, I'm leaving Trond's patches in for now, and hope to find a
  fix in the next week so I don't have to revert those patches"

* tag 'nfsd-5.5' of git://linux-nfs.org/~bfields/linux: (24 commits)
  nfsd: depend on CRYPTO_MD5 for legacy client tracking
  NFSD fixing possible null pointer derefering in copy offload
  nfsd: check for EBUSY from vfs_rmdir/vfs_unink.
  nfsd: Ensure CLONE persists data and metadata changes to the target file
  SUNRPC: Fix backchannel latency metrics
  nfsd: restore NFSv3 ACL support
  nfsd: v4 support requires CRYPTO_SHA256
  nfsd: Fix cld_net->cn_tfm initialization
  lockd: remove __KERNEL__ ifdefs
  sunrpc: remove __KERNEL__ ifdefs
  race in exportfs_decode_fh()
  nfsd: Drop LIST_HEAD where the variable it declares is never used.
  nfsd: document callback_wq serialization of callback code
  nfsd: mark cb path down on unknown errors
  nfsd: Fix races between nfsd4_cb_release() and nfsd4_shutdown_callback()
  nfsd: minor 4.1 callback cleanup
  SUNRPC: Fix svcauth_gss_proxy_init()
  SUNRPC: Trace gssproxy upcall results
  sunrpc: fix crash when cache_head become valid before update
  nfsd: remove private bin2hex implementation
  ...

Pull NFS client updates from Trond Myklebust:
 "Highlights include:

  Features:

   - NFSv4.2 now supports cross device offloaded copy (i.e. offloaded
     copy of a file from one source server to a different target
     server).

   - New RDMA tracepoints for debugging congestion control and Local
     Invalidate WRs.

  Bugfixes and cleanups

   - Drop the NFSv4.1 session slot if nfs4_delegreturn_prepare waits for
     layoutreturn

   - Handle bad/dead sessions correctly in nfs41_sequence_process()

   - Various bugfixes to the delegation return operation.

   - Various bugfixes pertaining to delegations that have been revoked.

   - Cleanups to the NFS timespec code to avoid unnecessary conversions
     between timespec and timespec64.

   - Fix unstable RDMA connections after a reconnect

   - Close race between waking an RDMA sender and posting a receive

   - Wake pending RDMA tasks if connection fails

   - Fix MR list corruption, and clean up MR usage

   - Fix another RPCSEC_GSS issue with MIC buffer space"

* tag 'nfs-for-5.5-1' of git://git.linux-nfs.org/projects/trondmy/linux-nfs: (79 commits)
  SUNRPC: Capture completion of all RPC tasks
  SUNRPC: Fix another issue with MIC buffer space
  NFS4: Trace lock reclaims
  NFS4: Trace state recovery operation
  NFSv4.2 fix memory leak in nfs42_ssc_open
  NFSv4.2 fix kfree in __nfs42_copy_file_range
  NFS: remove duplicated include from nfs4file.c
  NFSv4: Make _nfs42_proc_copy_notify() static
  NFS: Fallocate should use the nfs4_fattr_bitmap
  NFS: Return -ETXTBSY when attempting to write to a swapfile
  fs: nfs: sysfs: Remove NULL check before kfree
  NFS: remove unneeded semicolon
  NFSv4: add declaration of current_stateid
  NFSv4.x: Drop the slot if nfs4_delegreturn_prepare waits for layoutreturn
  NFSv4.x: Handle bad/dead sessions correctly in nfs41_sequence_process()
  nfsv4: Move NFSPROC4_CLNT_COPY_NOTIFY to end of list
  SUNRPC: Avoid RPC delays when exiting suspend
  NFS: Add a tracepoint in nfs_fh_to_dentry()
  NFSv4: Don't retry the GETATTR on old stateid in nfs4_delegreturn_done()
  NFSv4: Handle NFS4ERR_OLD_STATEID in delegreturn
  ...

We had cases in the previous patch where we were sending the security
descriptor context on SMB3 open (file create) in cases when we hadn't
mounted with with "modefromsid" mount option.

Add check for that mount flag before calling ad_sd_context in
open init.
Signed-off-by: Steve French <stfrench@microsoft.com>
Reviewed-by: Pavel Shilovsky <pshilov@microsoft.com>

Pull VFIO updates from Alex Williamson:

 - Remove hugepage checks for reserved pfns (Ben Luo)

 - Fix irq-bypass unregister ordering (Jiang Yi)

* tag 'vfio-v5.5-rc1' of git://github.com/awilliam/linux-vfio:
  vfio/pci: call irq_bypass_unregister_producer() before freeing irq
  vfio/type1: remove hugepage checks in is_invalid_reserved_pfn()

Pull more xen updates from Juergen Gross:

 - a patch to fix a build warning

 - a cleanup of no longer needed code in the Xen event handling

 - a small series for the Xen grant driver avoiding high order
   allocations and replacing an insane global limit by a per-call one

 - a small series fixing Xen frontend/backend module referencing

* tag 'for-linus-5.5b-rc1-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip:
  xen-blkback: allow module to be cleanly unloaded
  xen/xenbus: reference count registered modules
  xen/gntdev: switch from kcalloc() to kvcalloc()
  xen/gntdev: replace global limit of mapped pages by limit per call
  xen/gntdev: remove redundant non-zero check on ret
  xen/events: remove event handling recursion detection

Merge misc Kconfig updates from Andrew Morton:
 "A number of changes to Kconfig files under lib/ from Changbin Du and
  Krzysztof Kozlowski"

* emailed patches from Andrew Morton <akpm@linux-foundation.org>:
  lib/: fix Kconfig indentation
  kernel-hacking: move DEBUG_FS to 'Generic Kernel Debugging Instruments'
  kernel-hacking: move DEBUG_BUGVERBOSE to 'printk and dmesg options'
  kernel-hacking: create a submenu for scheduler debugging options
  kernel-hacking: move SCHED_STACK_END_CHECK after DEBUG_STACK_USAGE
  kernel-hacking: move Oops into 'Lockups and Hangs'
  kernel-hacking: move kernel testing and coverage options to same submenu
  kernel-hacking: group kernel data structures debugging together
  kernel-hacking: create submenu for arch special debugging options
  kernel-hacking: group sysrq/kgdb/ubsan into 'Generic Kernel Debugging Instruments'

In referenced fix we removed the RTL8168e-specific jumbo config for
RTL8168evl in rtl_hw_jumbo_enable(). We have to do the same in
rtl_hw_jumbo_disable().

v2: fix referenced commit id

Fixes: 14012c9f ("r8169: fix jumbo configuration for RTL8168evl")
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

pipe_wait() may be simple, but since it relies on the pipe lock, it
means that we have to do the wakeup while holding the lock.  That's
unfortunate, because the very first thing the waked entity will want to
do is to get the pipe lock for itself.

So get rid of the pipe_wait() usage by simply releasing the pipe lock,
doing the wakeup (if required) and then using wait_event_interruptible()
to wait on the right condition instead.

wait_event_interruptible() handles races on its own by comparing the
wakeup condition before and after adding itself to the wait queue, so
you can use an optimistic unlocked condition for it.

Cc: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Signed-off-by: Jiasen Lin <linjiasen@hygon.cn>
Signed-off-by: Jon Mason <jdmason@kudzu.us>

This code is ancient, and goes back to when we only had a single page
for the pipe buffers.  The exact history is hidden in the mists of time
(ie "before git", and in fact predates the BK repository too).

At that long-ago point in time, it actually helped to try to merge big
back-and-forth pipe reads and writes, and not limit pipe reads to the
single pipe buffer in length just because that was all we had at a time.

However, since then we've expanded the pipe buffers to multiple pages,
and this logic really doesn't seem to make sense.  And a lot of it is
somewhat questionable (ie "hmm, the user asked for a non-blocking read,
but we see that there's a writer pending, so let's wait anyway to get
the extra data that the writer will have").

But more importantly, it makes the "go to sleep" logic much less
obvious, and considering the wakeup issues we've had, I want to make for
less of those kinds of things.

Cc: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

This is the read side version of the previous commit: it simplifies the
logic to only wake up waiting writers when necessary, and makes sure to
use a synchronous wakeup.  This time not so much for GNU make jobserver
reasons (that pipe never fills up), but simply to get the writer going
quickly again.

A bit less verbose commentary this time, if only because I assume that
the write side commentary isn't going to be ignored if you touch this
code.

Cc: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

The pipe rework ends up having been extra painful, partly becaused of
actual bugs with ordering and caching of the pipe state, but also
because of subtle performance issues.

In particular, the pipe rework caused the kernel build to inexplicably
slow down.

The reason turns out to be that the GNU make jobserver (which limits the
parallelism of the build) uses a pipe to implement a "token" system: a
parallel submake will read a character from the pipe to get the job
token before starting a new job, and will write a character back to the
pipe when it is done.  The overall job limit is thus easily controlled
by just writing the appropriate number of initial token characters into
the pipe.

But to work well, that really means that the old behavior of write
wakeups being synchronous (WF_SYNC) is very important - when the pipe
writer wakes up a reader, we want the reader to actually get scheduled
immediately.  Otherwise you lose the parallelism of the build.

The pipe rework lost that synchronous wakeup on write, and we had
clearly all forgotten the reasons and rules for it.

This rewrites the pipe write wakeup logic to do the required Wsync
wakeups, but also clarifies the logic and avoids extraneous wakeups.

It also ends up addign a number of comments about what oit does and why,
so that we hopefully don't end up forgetting about this next time we
change this code.

Cc: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>