linux514/515-tkg: Add an alternative fsync (FUTEX_WAIT_MULTIPLE) patchset that can be enabled alongside futex_waitv to have a fallback path on Proton/wine builds using the older Fsync implementation. Thanks to Kisak for reaching out to me and sharing the ideas and patches.

This commit is contained in:
Tk-Glitch
2021-11-03 18:56:32 +01:00
parent 2d603642da
commit caad3e526d
5 changed files with 1208 additions and 7 deletions

View File

@ -59,7 +59,7 @@ else
fi fi
pkgname=("${pkgbase}" "${pkgbase}-headers") pkgname=("${pkgbase}" "${pkgbase}-headers")
pkgver="${_basekernel}"."${_sub}" pkgver="${_basekernel}"."${_sub}"
pkgrel=212 pkgrel=213
pkgdesc='Linux-tkg' pkgdesc='Linux-tkg'
arch=('x86_64') # no i686 in here arch=('x86_64') # no i686 in here
url="http://www.kernel.org/" url="http://www.kernel.org/"
@ -561,6 +561,7 @@ case $_basever in
0005-glitched-pds.patch 0005-glitched-pds.patch
0006-add-acs-overrides_iommu.patch 0006-add-acs-overrides_iommu.patch
0007-v5.14-fsync.patch 0007-v5.14-fsync.patch
0007-v5.14-fsync-waitvcompat.patch
0007-v5.14-futex2_interface.patch 0007-v5.14-futex2_interface.patch
0007-v5.14-futex_waitv.patch 0007-v5.14-futex_waitv.patch
0007-v5.14-winesync.patch 0007-v5.14-winesync.patch
@ -588,6 +589,7 @@ case $_basever in
'fca63d15ca4502aebd73e76d7499b243d2c03db71ff5ab0bf5cf268b2e576320' 'fca63d15ca4502aebd73e76d7499b243d2c03db71ff5ab0bf5cf268b2e576320'
'19661ec0d39f9663452b34433214c755179894528bf73a42f6ba52ccf572832a' '19661ec0d39f9663452b34433214c755179894528bf73a42f6ba52ccf572832a'
'aa67e81a27d9062e463594acb91eca6dd13388f23cbe53ca56298f9dba61cc10' 'aa67e81a27d9062e463594acb91eca6dd13388f23cbe53ca56298f9dba61cc10'
'ee97a751e9b3d2b7ea1707b807ab775cd6f4c392a5efaa475d24d16202518a6a'
'efe5e21706fdf64559ead866c85a5d88c5c3f743d814410df3810ca61cc5b966' 'efe5e21706fdf64559ead866c85a5d88c5c3f743d814410df3810ca61cc5b966'
'5742277f41f22bf29fa9742562946b8a01377f8a22adb42ceed3607541c1d5b6' '5742277f41f22bf29fa9742562946b8a01377f8a22adb42ceed3607541c1d5b6'
'034d12a73b507133da2c69a34d61efd2f6b6618549650aa26d748142d22002e1' '034d12a73b507133da2c69a34d61efd2f6b6618549650aa26d748142d22002e1'
@ -617,7 +619,7 @@ case $_basever in
0005-glitched-pds.patch 0005-glitched-pds.patch
0006-add-acs-overrides_iommu.patch 0006-add-acs-overrides_iommu.patch
0007-v5.15-fsync.patch 0007-v5.15-fsync.patch
#0007-v5.15-futex2_interface.patch 0007-v5.15-fsync-waitvcompat.patch
0007-v5.15-futex_waitv.patch 0007-v5.15-futex_waitv.patch
0007-v5.15-winesync.patch 0007-v5.15-winesync.patch
#0008-5.14-bcachefs.patch #0008-5.14-bcachefs.patch
@ -644,7 +646,7 @@ case $_basever in
'fca63d15ca4502aebd73e76d7499b243d2c03db71ff5ab0bf5cf268b2e576320' 'fca63d15ca4502aebd73e76d7499b243d2c03db71ff5ab0bf5cf268b2e576320'
'19661ec0d39f9663452b34433214c755179894528bf73a42f6ba52ccf572832a' '19661ec0d39f9663452b34433214c755179894528bf73a42f6ba52ccf572832a'
'6c4f0099896f69e56ebd8c9eac266ac8ad993acecd50945e0e84ef6f95f9ddca' '6c4f0099896f69e56ebd8c9eac266ac8ad993acecd50945e0e84ef6f95f9ddca'
#'efe5e21706fdf64559ead866c85a5d88c5c3f743d814410df3810ca61cc5b966' 'd6753a5d95a422eeb443edc04dee67512de8408aac1231e8f2e8ff9e000ad48e'
'c8f7c50d9b1418ba22b5ca735c47111a162be416109714d26a674162e5b2cb97' 'c8f7c50d9b1418ba22b5ca735c47111a162be416109714d26a674162e5b2cb97'
'034d12a73b507133da2c69a34d61efd2f6b6618549650aa26d748142d22002e1' '034d12a73b507133da2c69a34d61efd2f6b6618549650aa26d748142d22002e1'
'9fad4a40449e09522899955762c8928ae17f4cdaa16e01239fd12592e9d58177' '9fad4a40449e09522899955762c8928ae17f4cdaa16e01239fd12592e9d58177'

View File

@ -148,6 +148,7 @@ _acs_override=""
_bcachefs="" _bcachefs=""
# Set to "true" to enable support for fsync, an experimental replacement for esync found in Valve Proton 4.11+ - https://steamcommunity.com/games/221410/announcements/detail/2957094910196249305 # Set to "true" to enable support for fsync, an experimental replacement for esync found in Valve Proton 4.11+ - https://steamcommunity.com/games/221410/announcements/detail/2957094910196249305
# Can be enabled alongside _futex_waitv on 5.14+ to use it as a fallback for older Proton builds
_fsync="true" _fsync="true"
# Set to "true" to enable support for futex2, an experimental interface that can be used by proton-tkg and proton 5.13 experimental through Fsync - Can be enabled alongside fsync to use it as a fallback # Set to "true" to enable support for futex2, an experimental interface that can be used by proton-tkg and proton 5.13 experimental through Fsync - Can be enabled alongside fsync to use it as a fallback
@ -155,7 +156,7 @@ _fsync="true"
_futex2="true" _futex2="true"
# Set to "true" to enable backported patches to add support for the futex_waitv() syscall, a new interface for fsync. It will appear in mainline at Linux 5.16 release and requires a wine/proton with builtin support for it. It's expected to be available in Valve Proton 6.3 stable soon - https://github.com/ValveSoftware/wine/pull/128 # Set to "true" to enable backported patches to add support for the futex_waitv() syscall, a new interface for fsync. It will appear in mainline at Linux 5.16 release and requires a wine/proton with builtin support for it. It's expected to be available in Valve Proton 6.3 stable soon - https://github.com/ValveSoftware/wine/pull/128
# !! Disables fsync/futex2 interfaces support !! # !! Disables futex2 interfaces support !!
# https://github.com/andrealmeid/futex_waitv_patches # https://github.com/andrealmeid/futex_waitv_patches
_futex_waitv="false" _futex_waitv="false"

View File

@ -1156,7 +1156,7 @@ CONFIG_DEBUG_INFO_BTF_MODULES=y\r
if [ -z "$_futex_waitv" ]; then if [ -z "$_futex_waitv" ]; then
plain "" plain ""
plain "Enable support for futex_waitv, backported patches for fsync from 5.16 Kernel" plain "Enable support for futex_waitv, backported patches for fsync from 5.16 Kernel"
plain "! Will disable fsync/futex2 patchsets !" plain "! Will disable futex2 patchset !"
plain "https://github.com/andrealmeid/futex_waitv_patches" plain "https://github.com/andrealmeid/futex_waitv_patches"
plain "https://github.com/ValveSoftware/wine/pull/128" plain "https://github.com/ValveSoftware/wine/pull/128"
read -rp "`echo $' > N/y : '`" CONDITION9; read -rp "`echo $' > N/y : '`" CONDITION9;
@ -1164,18 +1164,24 @@ CONFIG_DEBUG_INFO_BTF_MODULES=y\r
if [[ "$CONDITION9" =~ [yY] ]] || [ "$_futex_waitv" = "true" ]; then if [[ "$CONDITION9" =~ [yY] ]] || [ "$_futex_waitv" = "true" ]; then
_msg="Patching futex_waitv support" _msg="Patching futex_waitv support"
_tkg_patcher _tkg_patcher
_fsync="false"
_futex2="false" _futex2="false"
fi fi
fi fi
# fsync support # fsync support
tkgpatch="$srcdir/0007-v${_basekernel}-fsync.patch" if [[ "$CONDITION9" =~ [yY] ]] || [ "$_futex_waitv" = "true" ]; then
tkgpatch="$srcdir/0007-v${_basekernel}-fsync-waitvcompat.patch"
else
tkgpatch="$srcdir/0007-v${_basekernel}-fsync.patch"
fi
if [ -e "$tkgpatch" ]; then if [ -e "$tkgpatch" ]; then
if [ -z "$_fsync" ]; then if [ -z "$_fsync" ]; then
plain "" plain ""
plain "Enable support for fsync, an experimental replacement for esync in Valve Proton 4.11+" plain "Enable support for fsync, an experimental replacement for esync in Valve Proton 4.11+"
plain "https://steamcommunity.com/games/221410/announcements/detail/2957094910196249305" plain "https://steamcommunity.com/games/221410/announcements/detail/2957094910196249305"
if [[ "$CONDITION9" =~ [yY] ]] || [ "$_futex_waitv" = "true" ]; then
plain "Will be used as a fallback to futex_waitv on older Proton builds if enabled"
fi
read -rp "`echo $' > N/y : '`" CONDITION10; read -rp "`echo $' > N/y : '`" CONDITION10;
fi fi
if [[ "$CONDITION10" =~ [yY] ]] || [ "$_fsync" = "true" ]; then if [[ "$CONDITION10" =~ [yY] ]] || [ "$_fsync" = "true" ]; then

View File

@ -0,0 +1,596 @@
From 7ca8de005bdf1940bf0b809fd94f25fb7c7bfa17 Mon Sep 17 00:00:00 2001
From: Gabriel Krisman Bertazi <krisman@collabora.com>
Date: Fri, 13 Dec 2019 11:08:02 -0300
Subject: [PATCH] futex: Implement mechanism to wait on any of several futexes
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
This is a new futex operation, called FUTEX_WAIT_MULTIPLE, which allows
a thread to wait on several futexes at the same time, and be awoken by
any of them. In a sense, it implements one of the features that was
supported by pooling on the old FUTEX_FD interface.
The use case lies in the Wine implementation of the Windows NT interface
WaitMultipleObjects. This Windows API function allows a thread to sleep
waiting on the first of a set of event sources (mutexes, timers, signal,
console input, etc) to signal. Considering this is a primitive
synchronization operation for Windows applications, being able to quickly
signal events on the producer side, and quickly go to sleep on the
consumer side is essential for good performance of those running over Wine.
Wine developers have an implementation that uses eventfd, but it suffers
from FD exhaustion (there is applications that go to the order of
multi-milion FDs), and higher CPU utilization than this new operation.
The futex list is passed as an array of `struct futex_wait_block`
(pointer, value, bitset) to the kernel, which will enqueue all of them
and sleep if none was already triggered. It returns a hint of which
futex caused the wake up event to userspace, but the hint doesn't
guarantee that is the only futex triggered. Before calling the syscall
again, userspace should traverse the list, trying to re-acquire any of
the other futexes, to prevent an immediate -EWOULDBLOCK return code from
the kernel.
This was tested using three mechanisms:
1) By reimplementing FUTEX_WAIT in terms of FUTEX_WAIT_MULTIPLE and
running the unmodified tools/testing/selftests/futex and a full linux
distro on top of this kernel.
2) By an example code that exercises the FUTEX_WAIT_MULTIPLE path on a
multi-threaded, event-handling setup.
3) By running the Wine fsync implementation and executing multi-threaded
applications, in particular modern games, on top of this implementation.
Changes were tested for the following ABIs: x86_64, i386 and x32.
Support for x32 applications is not implemented since it would
take a major rework adding a new entry point and splitting the current
futex 64 entry point in two and we can't change the current x32 syscall
number without breaking user space compatibility.
CC: Steven Rostedt <rostedt@goodmis.org>
Cc: Richard Yao <ryao@gentoo.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Co-developed-by: Zebediah Figura <z.figura12@gmail.com>
Signed-off-by: Zebediah Figura <z.figura12@gmail.com>
Co-developed-by: Steven Noonan <steven@valvesoftware.com>
Signed-off-by: Steven Noonan <steven@valvesoftware.com>
Co-developed-by: Pierre-Loup A. Griffais <pgriffais@valvesoftware.com>
Signed-off-by: Pierre-Loup A. Griffais <pgriffais@valvesoftware.com>
Signed-off-by: Gabriel Krisman Bertazi <krisman@collabora.com>
[Added compatibility code]
Co-developed-by: André Almeida <andrealmeid@collabora.com>
Signed-off-by: André Almeida <andrealmeid@collabora.com>
---
Changes since v2:
- Loop counters are now unsigned
- Add ifdef around `in_x32_syscall()`, so this function is only compiled
in architectures that declare it
Changes since RFC:
- Limit waitlist to 128 futexes
- Simplify wait loop
- Document functions
- Reduce allocated space
- Return hint if a futex was awoken during setup
- Check if any futex was awoken prior to sleep
- Drop relative timer logic
- Add compatibility struct and entry points
- Add selftests
PPA Note: This patch has been modified to build against 5.13
---
include/uapi/linux/futex.h | 20 +++
kernel/futex.c | 351 ++++++++++++++++++++++++++++++++++++-
2 files changed, 369 insertions(+), 2 deletions(-)
diff --git a/include/uapi/linux/futex.h b/include/uapi/linux/futex.h
index a89eb0accd5e..580001e89c6c 100644
--- a/include/uapi/linux/futex.h
+++ b/include/uapi/linux/futex.h
@@ -21,6 +21,7 @@
#define FUTEX_WAIT_REQUEUE_PI 11
#define FUTEX_CMP_REQUEUE_PI 12
#define FUTEX_LOCK_PI2 13
+#define FUTEX_WAIT_MULTIPLE 14
#define FUTEX_PRIVATE_FLAG 128
#define FUTEX_CLOCK_REALTIME 256
@@ -40,6 +41,8 @@
FUTEX_PRIVATE_FLAG)
#define FUTEX_CMP_REQUEUE_PI_PRIVATE (FUTEX_CMP_REQUEUE_PI | \
FUTEX_PRIVATE_FLAG)
+#define FUTEX_WAIT_MULTIPLE_PRIVATE (FUTEX_WAIT_MULTIPLE | \
+ FUTEX_PRIVATE_FLAG)
/*
* Support for robust futexes: the kernel cleans up held futexes at
@@ -150,4 +153,21 @@ struct robust_list_head {
(((op & 0xf) << 28) | ((cmp & 0xf) << 24) \
| ((oparg & 0xfff) << 12) | (cmparg & 0xfff))
+/*
+ * Maximum number of multiple futexes to wait for
+ */
+#define FUTEX_MULTIPLE_MAX_COUNT 128
+
+/**
+ * struct futex_wait_block - Block of futexes to be waited for
+ * @uaddr: User address of the futex
+ * @val: Futex value expected by userspace
+ * @bitset: Bitset for the optional bitmasked wakeup
+ */
+struct futex_wait_block {
+ __u32 __user *uaddr;
+ __u32 val;
+ __u32 bitset;
+};
+
#endif /* _UAPI_LINUX_FUTEX_H */
diff --git a/kernel/futex.c b/kernel/futex.c
index 45a13eb8894e..b28a6abbdefd 100644
--- a/kernel/futex.c
+++ b/kernel/futex.c
@@ -198,6 +198,8 @@ struct futex_pi_state {
* @rt_waiter: rt_waiter storage for use with requeue_pi
* @requeue_pi_key: the requeue_pi target futex key
* @bitset: bitset for the optional bitmasked wakeup
+ * @uaddr: userspace address of futex
+ * @uval: expected futex's value
*
* We use this hashed waitqueue, instead of a normal wait_queue_entry_t, so
* we can wake only the relevant ones (hashed queues may be shared).
@@ -220,6 +222,8 @@ struct futex_q {
struct rt_mutex_waiter *rt_waiter;
union futex_key *requeue_pi_key;
u32 bitset;
+ u32 __user *uaddr;
+ u32 uval;
} __randomize_layout;
static const struct futex_q futex_q_init = {
@@ -2313,6 +2317,29 @@ static int unqueue_me(struct futex_q *q)
return ret;
}
+/**
+ * unqueue_multiple2() - Remove several futexes from their futex_hash_bucket
+ * @q: The list of futexes to unqueue
+ * @count: Number of futexes in the list
+ *
+ * Helper to unqueue a list of futexes. This can't fail.
+ *
+ * Return:
+ * - >=0 - Index of the last futex that was awoken;
+ * - -1 - If no futex was awoken
+ */
+static int unqueue_multiple2(struct futex_q *q, int count)
+{
+ int ret = -1;
+ int i;
+
+ for (i = 0; i < count; i++) {
+ if (!unqueue_me(&q[i]))
+ ret = i;
+ }
+ return ret;
+}
+
/*
* PI futexes can not be requeued and must remove themself from the
* hash bucket. The hash bucket lock (i.e. lock_ptr) is held on entry
@@ -2680,6 +2707,204 @@ static int futex_wait_setup(u32 __user *uaddr, u32 val, unsigned int flags,
return ret;
}
+/**
+ * futex_wait_multiple_setup2() - Prepare to wait and enqueue multiple futexes
+ * @qs: The corresponding futex list
+ * @count: The size of the lists
+ * @flags: Futex flags (FLAGS_SHARED, etc.)
+ * @awaken: Index of the last awoken futex
+ *
+ * Prepare multiple futexes in a single step and enqueue them. This may fail if
+ * the futex list is invalid or if any futex was already awoken. On success the
+ * task is ready to interruptible sleep.
+ *
+ * Return:
+ * - 1 - One of the futexes was awaken by another thread
+ * - 0 - Success
+ * - <0 - -EFAULT, -EWOULDBLOCK or -EINVAL
+ */
+static int futex_wait_multiple_setup2(struct futex_q *qs, int count,
+ unsigned int flags, int *awaken)
+{
+ struct futex_hash_bucket *hb;
+ int ret, i;
+ u32 uval;
+
+ /*
+ * Enqueuing multiple futexes is tricky, because we need to
+ * enqueue each futex in the list before dealing with the next
+ * one to avoid deadlocking on the hash bucket. But, before
+ * enqueuing, we need to make sure that current->state is
+ * TASK_INTERRUPTIBLE, so we don't absorb any awake events, which
+ * cannot be done before the get_futex_key of the next key,
+ * because it calls get_user_pages, which can sleep. Thus, we
+ * fetch the list of futexes keys in two steps, by first pinning
+ * all the memory keys in the futex key, and only then we read
+ * each key and queue the corresponding futex.
+ */
+retry:
+ for (i = 0; i < count; i++) {
+ qs[i].key = FUTEX_KEY_INIT;
+ ret = get_futex_key(qs[i].uaddr, flags & FLAGS_SHARED,
+ &qs[i].key, FUTEX_READ);
+ if (unlikely(ret))
+ return ret;
+ }
+
+ set_current_state(TASK_INTERRUPTIBLE);
+
+ for (i = 0; i < count; i++) {
+ struct futex_q *q = &qs[i];
+
+ hb = queue_lock(q);
+
+ ret = get_futex_value_locked(&uval, q->uaddr);
+ if (ret) {
+ /*
+ * We need to try to handle the fault, which
+ * cannot be done without sleep, so we need to
+ * undo all the work already done, to make sure
+ * we don't miss any wake ups. Therefore, clean
+ * up, handle the fault and retry from the
+ * beginning.
+ */
+ queue_unlock(hb);
+
+ /*
+ * Keys 0..(i-1) are implicitly put
+ * on unqueue_multiple2.
+ */
+ *awaken = unqueue_multiple2(qs, i);
+
+ __set_current_state(TASK_RUNNING);
+
+ /*
+ * On a real fault, prioritize the error even if
+ * some other futex was awoken. Userspace gave
+ * us a bad address, -EFAULT them.
+ */
+ ret = get_user(uval, q->uaddr);
+ if (ret)
+ return ret;
+
+ /*
+ * Even if the page fault was handled, If
+ * something was already awaken, we can safely
+ * give up and succeed to give a hint for userspace to
+ * acquire the right futex faster.
+ */
+ if (*awaken >= 0)
+ return 1;
+
+ goto retry;
+ }
+
+ if (uval != q->uval) {
+ queue_unlock(hb);
+
+ /*
+ * If something was already awaken, we can
+ * safely ignore the error and succeed.
+ */
+ *awaken = unqueue_multiple2(qs, i);
+ __set_current_state(TASK_RUNNING);
+ if (*awaken >= 0)
+ return 1;
+
+ return -EWOULDBLOCK;
+ }
+
+ /*
+ * The bucket lock can't be held while dealing with the
+ * next futex. Queue each futex at this moment so hb can
+ * be unlocked.
+ */
+ queue_me(&qs[i], hb);
+ }
+ return 0;
+}
+
+/**
+ * futex_wait_multiple2() - Prepare to wait on and enqueue several futexes
+ * @qs: The list of futexes to wait on
+ * @op: Operation code from futex's syscall
+ * @count: The number of objects
+ * @abs_time: Timeout before giving up and returning to userspace
+ *
+ * Entry point for the FUTEX_WAIT_MULTIPLE futex operation, this function
+ * sleeps on a group of futexes and returns on the first futex that
+ * triggered, or after the timeout has elapsed.
+ *
+ * Return:
+ * - >=0 - Hint to the futex that was awoken
+ * - <0 - On error
+ */
+static int futex_wait_multiple2(struct futex_q *qs, int op,
+ u32 count, ktime_t *abs_time)
+{
+ struct hrtimer_sleeper timeout, *to;
+ int ret, flags = 0, hint = 0;
+ unsigned int i;
+
+ if (!(op & FUTEX_PRIVATE_FLAG))
+ flags |= FLAGS_SHARED;
+
+ if (op & FUTEX_CLOCK_REALTIME)
+ flags |= FLAGS_CLOCKRT;
+
+ to = futex_setup_timer(abs_time, &timeout, flags, 0);
+ while (1) {
+ ret = futex_wait_multiple_setup2(qs, count, flags, &hint);
+ if (ret) {
+ if (ret > 0) {
+ /* A futex was awaken during setup */
+ ret = hint;
+ }
+ break;
+ }
+
+ if (to)
+ hrtimer_start_expires(&to->timer, HRTIMER_MODE_ABS);
+
+ /*
+ * Avoid sleeping if another thread already tried to
+ * wake us.
+ */
+ for (i = 0; i < count; i++) {
+ if (plist_node_empty(&qs[i].list))
+ break;
+ }
+
+ if (i == count && (!to || to->task))
+ freezable_schedule();
+
+ ret = unqueue_multiple2(qs, count);
+
+ __set_current_state(TASK_RUNNING);
+
+ if (ret >= 0)
+ break;
+ if (to && !to->task) {
+ ret = -ETIMEDOUT;
+ break;
+ } else if (signal_pending(current)) {
+ ret = -ERESTARTSYS;
+ break;
+ }
+ /*
+ * The final case is a spurious wakeup, for
+ * which just retry.
+ */
+ }
+
+ if (to) {
+ hrtimer_cancel(&to->timer);
+ destroy_hrtimer_on_stack(&to->timer);
+ }
+
+ return ret;
+}
+
static int futex_wait(u32 __user *uaddr, unsigned int flags, u32 val,
ktime_t *abs_time, u32 bitset)
{
@@ -3761,6 +3986,43 @@ long do_futex(u32 __user *uaddr, int op, u32 val, ktime_t *timeout,
return -ENOSYS;
}
+/**
+ * futex_read_wait_block - Read an array of futex_wait_block from userspace
+ * @uaddr: Userspace address of the block
+ * @count: Number of blocks to be read
+ *
+ * This function creates and allocate an array of futex_q (we zero it to
+ * initialize the fields) and then, for each futex_wait_block element from
+ * userspace, fill a futex_q element with proper values.
+ */
+inline struct futex_q *futex_read_wait_block(u32 __user *uaddr, u32 count)
+{
+ unsigned int i;
+ struct futex_q *qs;
+ struct futex_wait_block fwb;
+ struct futex_wait_block __user *entry =
+ (struct futex_wait_block __user *)uaddr;
+
+ if (!count || count > FUTEX_MULTIPLE_MAX_COUNT)
+ return ERR_PTR(-EINVAL);
+
+ qs = kcalloc(count, sizeof(*qs), GFP_KERNEL);
+ if (!qs)
+ return ERR_PTR(-ENOMEM);
+
+ for (i = 0; i < count; i++) {
+ if (copy_from_user(&fwb, &entry[i], sizeof(fwb))) {
+ kfree(qs);
+ return ERR_PTR(-EFAULT);
+ }
+
+ qs[i].uaddr = fwb.uaddr;
+ qs[i].uval = fwb.val;
+ qs[i].bitset = fwb.bitset;
+ }
+
+ return qs;
+}
SYSCALL_DEFINE6(futex, u32 __user *, uaddr, int, op, u32, val,
struct __kernel_timespec __user *, utime, u32 __user *, uaddr2,
@@ -4025,6 +4025,7 @@
case FUTEX_LOCK_PI:
case FUTEX_WAIT_BITSET:
case FUTEX_WAIT_REQUEUE_PI:
+ case FUTEX_WAIT_MULTIPLE:
return true;
}
return false;
@@ -3796,6 +4059,25 @@ SYSCALL_DEFINE6(futex, u32 __user *, uaddr, int, op, u32, val,
tp = &t;
}
+ if (cmd == FUTEX_WAIT_MULTIPLE) {
+ int ret;
+ struct futex_q *qs;
+
+#ifdef CONFIG_X86_X32
+ if (unlikely(in_x32_syscall()))
+ return -ENOSYS;
+#endif
+ qs = futex_read_wait_block(uaddr, val);
+
+ if (IS_ERR(qs))
+ return PTR_ERR(qs);
+
+ ret = futex_wait_multiple2(qs, op, val, tp);
+ kfree(qs);
+
+ return ret;
+ }
+
return do_futex(uaddr, op, val, tp, uaddr2, (unsigned long)utime, val3);
}
@@ -3958,6 +4240,57 @@ COMPAT_SYSCALL_DEFINE3(get_robust_list, int, pid,
#endif /* CONFIG_COMPAT */
#ifdef CONFIG_COMPAT_32BIT_TIME
+/**
+ * struct compat_futex_wait_block - Block of futexes to be waited for
+ * @uaddr: User address of the futex (compatible pointer)
+ * @val: Futex value expected by userspace
+ * @bitset: Bitset for the optional bitmasked wakeup
+ */
+struct compat_futex_wait_block {
+ compat_uptr_t uaddr;
+ __u32 val;
+ __u32 bitset;
+};
+
+/**
+ * compat_futex_read_wait_block - Read an array of futex_wait_block from
+ * userspace
+ * @uaddr: Userspace address of the block
+ * @count: Number of blocks to be read
+ *
+ * This function does the same as futex_read_wait_block(), except that it
+ * converts the pointer to the futex from the compat version to the regular one.
+ */
+inline struct futex_q *compat_futex_read_wait_block(u32 __user *uaddr,
+ u32 count)
+{
+ unsigned int i;
+ struct futex_q *qs;
+ struct compat_futex_wait_block fwb;
+ struct compat_futex_wait_block __user *entry =
+ (struct compat_futex_wait_block __user *)uaddr;
+
+ if (!count || count > FUTEX_MULTIPLE_MAX_COUNT)
+ return ERR_PTR(-EINVAL);
+
+ qs = kcalloc(count, sizeof(*qs), GFP_KERNEL);
+ if (!qs)
+ return ERR_PTR(-ENOMEM);
+
+ for (i = 0; i < count; i++) {
+ if (copy_from_user(&fwb, &entry[i], sizeof(fwb))) {
+ kfree(qs);
+ return ERR_PTR(-EFAULT);
+ }
+
+ qs[i].uaddr = compat_ptr(fwb.uaddr);
+ qs[i].uval = fwb.val;
+ qs[i].bitset = fwb.bitset;
+ }
+
+ return qs;
+}
+
SYSCALL_DEFINE6(futex_time32, u32 __user *, uaddr, int, op, u32, val,
struct old_timespec32 __user *, utime, u32 __user *, uaddr2,
u32, val3)
@@ -3986,6 +4320,19 @@ SYSCALL_DEFINE6(futex_time32, u32 __user *, uaddr, int, op, u32, val,
tp = &t;
}
+ if (cmd == FUTEX_WAIT_MULTIPLE) {
+ int ret;
+ struct futex_q *qs = compat_futex_read_wait_block(uaddr, val);
+
+ if (IS_ERR(qs))
+ return PTR_ERR(qs);
+
+ ret = futex_wait_multiple2(qs, op, val, tp);
+ kfree(qs);
+
+ return ret;
+ }
+
return do_futex(uaddr, op, val, tp, uaddr2, (unsigned long)utime, val3);
}
#endif /* CONFIG_COMPAT_32BIT_TIME */
--
GitLab
From cdaa060153c4adf0a64d7bbf7e3ebba727c7e335 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Andr=C3=A9=20Almeida?= <andrealmeid@collabora.com>
Date: Fri, 7 Feb 2020 23:28:02 -0300
Subject: [PATCH] futex: Add Proton compatibility code
PPA Note: This patch has been modified to build against 5.13
---
include/uapi/linux/futex.h | 2 +-
kernel/futex.c | 5 +++--
2 files changed, 4 insertions(+), 3 deletions(-)
diff --git a/include/uapi/linux/futex.h b/include/uapi/linux/futex.h
index 580001e89c6c..a3e760886b8e 100644
--- a/include/uapi/linux/futex.h
+++ b/include/uapi/linux/futex.h
@@ -21,7 +21,7 @@
#define FUTEX_WAIT_REQUEUE_PI 11
#define FUTEX_CMP_REQUEUE_PI 12
#define FUTEX_LOCK_PI2 13
-#define FUTEX_WAIT_MULTIPLE 14
+#define FUTEX_WAIT_MULTIPLE 31
#define FUTEX_PRIVATE_FLAG 128
#define FUTEX_CLOCK_REALTIME 256
diff --git a/kernel/futex.c b/kernel/futex.c
index b28a6abbdefd..5b547c911b92 100644
--- a/kernel/futex.c
+++ b/kernel/futex.c
@@ -4038,7 +4038,7 @@
return -EINVAL;
*t = timespec64_to_ktime(*ts);
- if (cmd == FUTEX_WAIT)
+ if (cmd == FUTEX_WAIT || cmd == FUTEX_WAIT_MULTIPLE)
*t = ktime_add_safe(ktime_get(), *t);
else if (cmd != FUTEX_LOCK_PI && !(op & FUTEX_CLOCK_REALTIME))
*t = timens_ktime_to_host(CLOCK_MONOTONIC, *t);
@@ -4248,6 +4248,7 @@ COMPAT_SYSCALL_DEFINE3(get_robust_list, int, pid,
*/
struct compat_futex_wait_block {
compat_uptr_t uaddr;
+ __u32 pad;
__u32 val;
__u32 bitset;
};
--
GitLab

View File

@ -0,0 +1,596 @@
From 7ca8de005bdf1940bf0b809fd94f25fb7c7bfa17 Mon Sep 17 00:00:00 2001
From: Gabriel Krisman Bertazi <krisman@collabora.com>
Date: Fri, 13 Dec 2019 11:08:02 -0300
Subject: [PATCH] futex: Implement mechanism to wait on any of several futexes
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
This is a new futex operation, called FUTEX_WAIT_MULTIPLE, which allows
a thread to wait on several futexes at the same time, and be awoken by
any of them. In a sense, it implements one of the features that was
supported by pooling on the old FUTEX_FD interface.
The use case lies in the Wine implementation of the Windows NT interface
WaitMultipleObjects. This Windows API function allows a thread to sleep
waiting on the first of a set of event sources (mutexes, timers, signal,
console input, etc) to signal. Considering this is a primitive
synchronization operation for Windows applications, being able to quickly
signal events on the producer side, and quickly go to sleep on the
consumer side is essential for good performance of those running over Wine.
Wine developers have an implementation that uses eventfd, but it suffers
from FD exhaustion (there is applications that go to the order of
multi-milion FDs), and higher CPU utilization than this new operation.
The futex list is passed as an array of `struct futex_wait_block`
(pointer, value, bitset) to the kernel, which will enqueue all of them
and sleep if none was already triggered. It returns a hint of which
futex caused the wake up event to userspace, but the hint doesn't
guarantee that is the only futex triggered. Before calling the syscall
again, userspace should traverse the list, trying to re-acquire any of
the other futexes, to prevent an immediate -EWOULDBLOCK return code from
the kernel.
This was tested using three mechanisms:
1) By reimplementing FUTEX_WAIT in terms of FUTEX_WAIT_MULTIPLE and
running the unmodified tools/testing/selftests/futex and a full linux
distro on top of this kernel.
2) By an example code that exercises the FUTEX_WAIT_MULTIPLE path on a
multi-threaded, event-handling setup.
3) By running the Wine fsync implementation and executing multi-threaded
applications, in particular modern games, on top of this implementation.
Changes were tested for the following ABIs: x86_64, i386 and x32.
Support for x32 applications is not implemented since it would
take a major rework adding a new entry point and splitting the current
futex 64 entry point in two and we can't change the current x32 syscall
number without breaking user space compatibility.
CC: Steven Rostedt <rostedt@goodmis.org>
Cc: Richard Yao <ryao@gentoo.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Co-developed-by: Zebediah Figura <z.figura12@gmail.com>
Signed-off-by: Zebediah Figura <z.figura12@gmail.com>
Co-developed-by: Steven Noonan <steven@valvesoftware.com>
Signed-off-by: Steven Noonan <steven@valvesoftware.com>
Co-developed-by: Pierre-Loup A. Griffais <pgriffais@valvesoftware.com>
Signed-off-by: Pierre-Loup A. Griffais <pgriffais@valvesoftware.com>
Signed-off-by: Gabriel Krisman Bertazi <krisman@collabora.com>
[Added compatibility code]
Co-developed-by: André Almeida <andrealmeid@collabora.com>
Signed-off-by: André Almeida <andrealmeid@collabora.com>
---
Changes since v2:
- Loop counters are now unsigned
- Add ifdef around `in_x32_syscall()`, so this function is only compiled
in architectures that declare it
Changes since RFC:
- Limit waitlist to 128 futexes
- Simplify wait loop
- Document functions
- Reduce allocated space
- Return hint if a futex was awoken during setup
- Check if any futex was awoken prior to sleep
- Drop relative timer logic
- Add compatibility struct and entry points
- Add selftests
PPA Note: This patch has been modified to build against 5.13
---
include/uapi/linux/futex.h | 20 +++
kernel/futex.c | 351 ++++++++++++++++++++++++++++++++++++-
2 files changed, 369 insertions(+), 2 deletions(-)
diff --git a/include/uapi/linux/futex.h b/include/uapi/linux/futex.h
index a89eb0accd5e..580001e89c6c 100644
--- a/include/uapi/linux/futex.h
+++ b/include/uapi/linux/futex.h
@@ -21,6 +21,7 @@
#define FUTEX_WAIT_REQUEUE_PI 11
#define FUTEX_CMP_REQUEUE_PI 12
#define FUTEX_LOCK_PI2 13
+#define FUTEX_WAIT_MULTIPLE 14
#define FUTEX_PRIVATE_FLAG 128
#define FUTEX_CLOCK_REALTIME 256
@@ -40,6 +41,8 @@
FUTEX_PRIVATE_FLAG)
#define FUTEX_CMP_REQUEUE_PI_PRIVATE (FUTEX_CMP_REQUEUE_PI | \
FUTEX_PRIVATE_FLAG)
+#define FUTEX_WAIT_MULTIPLE_PRIVATE (FUTEX_WAIT_MULTIPLE | \
+ FUTEX_PRIVATE_FLAG)
/*
* Support for robust futexes: the kernel cleans up held futexes at
@@ -150,4 +153,21 @@ struct robust_list_head {
(((op & 0xf) << 28) | ((cmp & 0xf) << 24) \
| ((oparg & 0xfff) << 12) | (cmparg & 0xfff))
+/*
+ * Maximum number of multiple futexes to wait for
+ */
+#define FUTEX_MULTIPLE_MAX_COUNT 128
+
+/**
+ * struct futex_wait_block - Block of futexes to be waited for
+ * @uaddr: User address of the futex
+ * @val: Futex value expected by userspace
+ * @bitset: Bitset for the optional bitmasked wakeup
+ */
+struct futex_wait_block {
+ __u32 __user *uaddr;
+ __u32 val;
+ __u32 bitset;
+};
+
#endif /* _UAPI_LINUX_FUTEX_H */
diff --git a/kernel/futex.c b/kernel/futex.c
index 45a13eb8894e..b28a6abbdefd 100644
--- a/kernel/futex.c
+++ b/kernel/futex.c
@@ -198,6 +198,8 @@ struct futex_pi_state {
* @bitset: bitset for the optional bitmasked wakeup
* @requeue_state: State field for futex_requeue_pi()
* @requeue_wait: RCU wait for futex_requeue_pi() (RT only)
+ * @uaddr: userspace address of futex
+ * @uval: expected futex's value
*
* We use this hashed waitqueue, instead of a normal wait_queue_entry_t, so
* we can wake only the relevant ones (hashed queues may be shared).
@@ -220,6 +222,8 @@ struct futex_q {
union futex_key *requeue_pi_key;
u32 bitset;
atomic_t requeue_state;
+ u32 __user *uaddr;
+ u32 uval;
#ifdef CONFIG_PREEMPT_RT
struct rcuwait requeue_wait;
#endif
@@ -2313,6 +2317,29 @@ static int unqueue_me(struct futex_q *q)
return ret;
}
+/**
+ * unqueue_multiple2() - Remove several futexes from their futex_hash_bucket
+ * @q: The list of futexes to unqueue
+ * @count: Number of futexes in the list
+ *
+ * Helper to unqueue a list of futexes. This can't fail.
+ *
+ * Return:
+ * - >=0 - Index of the last futex that was awoken;
+ * - -1 - If no futex was awoken
+ */
+static int unqueue_multiple2(struct futex_q *q, int count)
+{
+ int ret = -1;
+ int i;
+
+ for (i = 0; i < count; i++) {
+ if (!unqueue_me(&q[i]))
+ ret = i;
+ }
+ return ret;
+}
+
/*
* PI futexes can not be requeued and must remove themself from the
* hash bucket. The hash bucket lock (i.e. lock_ptr) is held on entry
@@ -2680,6 +2707,204 @@ static int futex_wait_setup(u32 __user *uaddr, u32 val, unsigned int flags,
return ret;
}
+/**
+ * futex_wait_multiple_setup2() - Prepare to wait and enqueue multiple futexes
+ * @qs: The corresponding futex list
+ * @count: The size of the lists
+ * @flags: Futex flags (FLAGS_SHARED, etc.)
+ * @awaken: Index of the last awoken futex
+ *
+ * Prepare multiple futexes in a single step and enqueue them. This may fail if
+ * the futex list is invalid or if any futex was already awoken. On success the
+ * task is ready to interruptible sleep.
+ *
+ * Return:
+ * - 1 - One of the futexes was awaken by another thread
+ * - 0 - Success
+ * - <0 - -EFAULT, -EWOULDBLOCK or -EINVAL
+ */
+static int futex_wait_multiple_setup2(struct futex_q *qs, int count,
+ unsigned int flags, int *awaken)
+{
+ struct futex_hash_bucket *hb;
+ int ret, i;
+ u32 uval;
+
+ /*
+ * Enqueuing multiple futexes is tricky, because we need to
+ * enqueue each futex in the list before dealing with the next
+ * one to avoid deadlocking on the hash bucket. But, before
+ * enqueuing, we need to make sure that current->state is
+ * TASK_INTERRUPTIBLE, so we don't absorb any awake events, which
+ * cannot be done before the get_futex_key of the next key,
+ * because it calls get_user_pages, which can sleep. Thus, we
+ * fetch the list of futexes keys in two steps, by first pinning
+ * all the memory keys in the futex key, and only then we read
+ * each key and queue the corresponding futex.
+ */
+retry:
+ for (i = 0; i < count; i++) {
+ qs[i].key = FUTEX_KEY_INIT;
+ ret = get_futex_key(qs[i].uaddr, flags & FLAGS_SHARED,
+ &qs[i].key, FUTEX_READ);
+ if (unlikely(ret))
+ return ret;
+ }
+
+ set_current_state(TASK_INTERRUPTIBLE);
+
+ for (i = 0; i < count; i++) {
+ struct futex_q *q = &qs[i];
+
+ hb = queue_lock(q);
+
+ ret = get_futex_value_locked(&uval, q->uaddr);
+ if (ret) {
+ /*
+ * We need to try to handle the fault, which
+ * cannot be done without sleep, so we need to
+ * undo all the work already done, to make sure
+ * we don't miss any wake ups. Therefore, clean
+ * up, handle the fault and retry from the
+ * beginning.
+ */
+ queue_unlock(hb);
+
+ /*
+ * Keys 0..(i-1) are implicitly put
+ * on unqueue_multiple2.
+ */
+ *awaken = unqueue_multiple2(qs, i);
+
+ __set_current_state(TASK_RUNNING);
+
+ /*
+ * On a real fault, prioritize the error even if
+ * some other futex was awoken. Userspace gave
+ * us a bad address, -EFAULT them.
+ */
+ ret = get_user(uval, q->uaddr);
+ if (ret)
+ return ret;
+
+ /*
+ * Even if the page fault was handled, If
+ * something was already awaken, we can safely
+ * give up and succeed to give a hint for userspace to
+ * acquire the right futex faster.
+ */
+ if (*awaken >= 0)
+ return 1;
+
+ goto retry;
+ }
+
+ if (uval != q->uval) {
+ queue_unlock(hb);
+
+ /*
+ * If something was already awaken, we can
+ * safely ignore the error and succeed.
+ */
+ *awaken = unqueue_multiple2(qs, i);
+ __set_current_state(TASK_RUNNING);
+ if (*awaken >= 0)
+ return 1;
+
+ return -EWOULDBLOCK;
+ }
+
+ /*
+ * The bucket lock can't be held while dealing with the
+ * next futex. Queue each futex at this moment so hb can
+ * be unlocked.
+ */
+ queue_me(&qs[i], hb);
+ }
+ return 0;
+}
+
+/**
+ * futex_wait_multiple2() - Prepare to wait on and enqueue several futexes
+ * @qs: The list of futexes to wait on
+ * @op: Operation code from futex's syscall
+ * @count: The number of objects
+ * @abs_time: Timeout before giving up and returning to userspace
+ *
+ * Entry point for the FUTEX_WAIT_MULTIPLE futex operation, this function
+ * sleeps on a group of futexes and returns on the first futex that
+ * triggered, or after the timeout has elapsed.
+ *
+ * Return:
+ * - >=0 - Hint to the futex that was awoken
+ * - <0 - On error
+ */
+static int futex_wait_multiple2(struct futex_q *qs, int op,
+ u32 count, ktime_t *abs_time)
+{
+ struct hrtimer_sleeper timeout, *to;
+ int ret, flags = 0, hint = 0;
+ unsigned int i;
+
+ if (!(op & FUTEX_PRIVATE_FLAG))
+ flags |= FLAGS_SHARED;
+
+ if (op & FUTEX_CLOCK_REALTIME)
+ flags |= FLAGS_CLOCKRT;
+
+ to = futex_setup_timer(abs_time, &timeout, flags, 0);
+ while (1) {
+ ret = futex_wait_multiple_setup2(qs, count, flags, &hint);
+ if (ret) {
+ if (ret > 0) {
+ /* A futex was awaken during setup */
+ ret = hint;
+ }
+ break;
+ }
+
+ if (to)
+ hrtimer_start_expires(&to->timer, HRTIMER_MODE_ABS);
+
+ /*
+ * Avoid sleeping if another thread already tried to
+ * wake us.
+ */
+ for (i = 0; i < count; i++) {
+ if (plist_node_empty(&qs[i].list))
+ break;
+ }
+
+ if (i == count && (!to || to->task))
+ freezable_schedule();
+
+ ret = unqueue_multiple2(qs, count);
+
+ __set_current_state(TASK_RUNNING);
+
+ if (ret >= 0)
+ break;
+ if (to && !to->task) {
+ ret = -ETIMEDOUT;
+ break;
+ } else if (signal_pending(current)) {
+ ret = -ERESTARTSYS;
+ break;
+ }
+ /*
+ * The final case is a spurious wakeup, for
+ * which just retry.
+ */
+ }
+
+ if (to) {
+ hrtimer_cancel(&to->timer);
+ destroy_hrtimer_on_stack(&to->timer);
+ }
+
+ return ret;
+}
+
static int futex_wait(u32 __user *uaddr, unsigned int flags, u32 val,
ktime_t *abs_time, u32 bitset)
{
@@ -3761,6 +3986,43 @@ long do_futex(u32 __user *uaddr, int op, u32 val, ktime_t *timeout,
return -ENOSYS;
}
+/**
+ * futex_read_wait_block - Read an array of futex_wait_block from userspace
+ * @uaddr: Userspace address of the block
+ * @count: Number of blocks to be read
+ *
+ * This function creates and allocate an array of futex_q (we zero it to
+ * initialize the fields) and then, for each futex_wait_block element from
+ * userspace, fill a futex_q element with proper values.
+ */
+inline struct futex_q *futex_read_wait_block(u32 __user *uaddr, u32 count)
+{
+ unsigned int i;
+ struct futex_q *qs;
+ struct futex_wait_block fwb;
+ struct futex_wait_block __user *entry =
+ (struct futex_wait_block __user *)uaddr;
+
+ if (!count || count > FUTEX_MULTIPLE_MAX_COUNT)
+ return ERR_PTR(-EINVAL);
+
+ qs = kcalloc(count, sizeof(*qs), GFP_KERNEL);
+ if (!qs)
+ return ERR_PTR(-ENOMEM);
+
+ for (i = 0; i < count; i++) {
+ if (copy_from_user(&fwb, &entry[i], sizeof(fwb))) {
+ kfree(qs);
+ return ERR_PTR(-EFAULT);
+ }
+
+ qs[i].uaddr = fwb.uaddr;
+ qs[i].uval = fwb.val;
+ qs[i].bitset = fwb.bitset;
+ }
+
+ return qs;
+}
SYSCALL_DEFINE6(futex, u32 __user *, uaddr, int, op, u32, val,
struct __kernel_timespec __user *, utime, u32 __user *, uaddr2,
@@ -4025,6 +4025,7 @@
case FUTEX_LOCK_PI:
case FUTEX_WAIT_BITSET:
case FUTEX_WAIT_REQUEUE_PI:
+ case FUTEX_WAIT_MULTIPLE:
return true;
}
return false;
@@ -3796,6 +4059,25 @@ SYSCALL_DEFINE6(futex, u32 __user *, uaddr, int, op, u32, val,
tp = &t;
}
+ if (cmd == FUTEX_WAIT_MULTIPLE) {
+ int ret;
+ struct futex_q *qs;
+
+#ifdef CONFIG_X86_X32
+ if (unlikely(in_x32_syscall()))
+ return -ENOSYS;
+#endif
+ qs = futex_read_wait_block(uaddr, val);
+
+ if (IS_ERR(qs))
+ return PTR_ERR(qs);
+
+ ret = futex_wait_multiple2(qs, op, val, tp);
+ kfree(qs);
+
+ return ret;
+ }
+
return do_futex(uaddr, op, val, tp, uaddr2, (unsigned long)utime, val3);
}
@@ -3958,6 +4240,57 @@ COMPAT_SYSCALL_DEFINE3(get_robust_list, int, pid,
#endif /* CONFIG_COMPAT */
#ifdef CONFIG_COMPAT_32BIT_TIME
+/**
+ * struct compat_futex_wait_block - Block of futexes to be waited for
+ * @uaddr: User address of the futex (compatible pointer)
+ * @val: Futex value expected by userspace
+ * @bitset: Bitset for the optional bitmasked wakeup
+ */
+struct compat_futex_wait_block {
+ compat_uptr_t uaddr;
+ __u32 val;
+ __u32 bitset;
+};
+
+/**
+ * compat_futex_read_wait_block - Read an array of futex_wait_block from
+ * userspace
+ * @uaddr: Userspace address of the block
+ * @count: Number of blocks to be read
+ *
+ * This function does the same as futex_read_wait_block(), except that it
+ * converts the pointer to the futex from the compat version to the regular one.
+ */
+inline struct futex_q *compat_futex_read_wait_block(u32 __user *uaddr,
+ u32 count)
+{
+ unsigned int i;
+ struct futex_q *qs;
+ struct compat_futex_wait_block fwb;
+ struct compat_futex_wait_block __user *entry =
+ (struct compat_futex_wait_block __user *)uaddr;
+
+ if (!count || count > FUTEX_MULTIPLE_MAX_COUNT)
+ return ERR_PTR(-EINVAL);
+
+ qs = kcalloc(count, sizeof(*qs), GFP_KERNEL);
+ if (!qs)
+ return ERR_PTR(-ENOMEM);
+
+ for (i = 0; i < count; i++) {
+ if (copy_from_user(&fwb, &entry[i], sizeof(fwb))) {
+ kfree(qs);
+ return ERR_PTR(-EFAULT);
+ }
+
+ qs[i].uaddr = compat_ptr(fwb.uaddr);
+ qs[i].uval = fwb.val;
+ qs[i].bitset = fwb.bitset;
+ }
+
+ return qs;
+}
+
SYSCALL_DEFINE6(futex_time32, u32 __user *, uaddr, int, op, u32, val,
struct old_timespec32 __user *, utime, u32 __user *, uaddr2,
u32, val3)
@@ -3986,6 +4320,19 @@ SYSCALL_DEFINE6(futex_time32, u32 __user *, uaddr, int, op, u32, val,
tp = &t;
}
+ if (cmd == FUTEX_WAIT_MULTIPLE) {
+ int ret;
+ struct futex_q *qs = compat_futex_read_wait_block(uaddr, val);
+
+ if (IS_ERR(qs))
+ return PTR_ERR(qs);
+
+ ret = futex_wait_multiple2(qs, op, val, tp);
+ kfree(qs);
+
+ return ret;
+ }
+
return do_futex(uaddr, op, val, tp, uaddr2, (unsigned long)utime, val3);
}
#endif /* CONFIG_COMPAT_32BIT_TIME */
--
GitLab
From cdaa060153c4adf0a64d7bbf7e3ebba727c7e335 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Andr=C3=A9=20Almeida?= <andrealmeid@collabora.com>
Date: Fri, 7 Feb 2020 23:28:02 -0300
Subject: [PATCH] futex: Add Proton compatibility code
PPA Note: This patch has been modified to build against 5.13
---
include/uapi/linux/futex.h | 2 +-
kernel/futex.c | 5 +++--
2 files changed, 4 insertions(+), 3 deletions(-)
diff --git a/include/uapi/linux/futex.h b/include/uapi/linux/futex.h
index 580001e89c6c..a3e760886b8e 100644
--- a/include/uapi/linux/futex.h
+++ b/include/uapi/linux/futex.h
@@ -21,7 +21,7 @@
#define FUTEX_WAIT_REQUEUE_PI 11
#define FUTEX_CMP_REQUEUE_PI 12
#define FUTEX_LOCK_PI2 13
-#define FUTEX_WAIT_MULTIPLE 14
+#define FUTEX_WAIT_MULTIPLE 31
#define FUTEX_PRIVATE_FLAG 128
#define FUTEX_CLOCK_REALTIME 256
diff --git a/kernel/futex.c b/kernel/futex.c
index b28a6abbdefd..5b547c911b92 100644
--- a/kernel/futex.c
+++ b/kernel/futex.c
@@ -4038,7 +4038,7 @@
return -EINVAL;
*t = timespec64_to_ktime(*ts);
- if (cmd == FUTEX_WAIT)
+ if (cmd == FUTEX_WAIT || cmd == FUTEX_WAIT_MULTIPLE)
*t = ktime_add_safe(ktime_get(), *t);
else if (cmd != FUTEX_LOCK_PI && !(op & FUTEX_CLOCK_REALTIME))
*t = timens_ktime_to_host(CLOCK_MONOTONIC, *t);
@@ -4248,6 +4248,7 @@ COMPAT_SYSCALL_DEFINE3(get_robust_list, int, pid,
*/
struct compat_futex_wait_block {
compat_uptr_t uaddr;
+ __u32 pad;
__u32 val;
__u32 bitset;
};
--
GitLab