CC-ing stable(a)vger.kernel.org and adding some more explanations.
On Tue, 5 May 2020 10:10:33 +0200 SeongJae Park <sjpark(a)amazon.com> wrote:
> From: SeongJae Park <sjpark(a)amazon.de>
>
> The commit 6d7855c54e1e ("sockfs: switch to ->free_inode()") made the
> deallocation of 'socket_alloc' to be done asynchronously using RCU, as
> same to 'sock.wq'. And the following commit 333f7909a857 ("coallocate
> socket_sq with socket itself") made those to have same life cycle.
>
> The changes made the code much more simple, but also made 'socket_alloc'
> live longer than before. For the reason, user programs intensively
> repeating allocations and deallocations of sockets could cause memory
> pressure on recent kernels.
I found this problem on a production virtual machine utilizing 4GB memory while
running lebench[1]. The 'poll big' test of lebench opens 1000 sockets, polls
and closes those. This test is repeated 10,000 times. Therefore it should
consume only 1000 'socket_alloc' objects at once. As size of socket_alloc is
about 800 Bytes, it's only 800 KiB. However, on the recent kernels, it could
consume up to 10,000,000 objects (about 8 GiB). On the test machine, I
confirmed it consuming about 4GB of the system memory and results in OOM.
[1] https://github.com/LinuxPerfStudy/LEBench
>
> To avoid the problem, this commit reverts the changes.
I also tried to make fixup rather than reverts, but I couldn't easily find
simple fixup. As the commits 6d7855c54e1e and 333f7909a857 were for code
refactoring rather than performance optimization, I thought introducing complex
fixup for this problem would make no sense. Meanwhile, the memory pressure
regression could affect real machines. To this end, I decided to quickly
revert the commits first and consider better refactoring later.
Thanks,
SeongJae Park
>
> SeongJae Park (2):
> Revert "coallocate socket_wq with socket itself"
> Revert "sockfs: switch to ->free_inode()"
>
> drivers/net/tap.c | 5 +++--
> drivers/net/tun.c | 8 +++++---
> include/linux/if_tap.h | 1 +
> include/linux/net.h | 4 ++--
> include/net/sock.h | 4 ++--
> net/core/sock.c | 2 +-
> net/socket.c | 23 ++++++++++++++++-------
> 7 files changed, 30 insertions(+), 17 deletions(-)
>
> --
> 2.17.1
We recorded the dependencies for WAIT_FOR_SUBMIT in order that we could
correctly perform priority inheritance from the parallel branches to the
common trunk. However, for the purpose of timeslicing and reset
handling, the dependency is weak -- as we the pair of requests are
allowed to run in parallel and not in strict succession. So for example
we do need to suspend one if the other hangs.
The real significance though is that this allows us to rearrange
groups of WAIT_FOR_SUBMIT linked requests along the single engine, and
so can resolve user level inter-batch scheduling dependencies from user
semaphores.
Fixes: c81471f5e95c ("drm/i915: Copy across scheduler behaviour flags across submit fences")
Testcase: igt/gem_exec_fence/submit
Signed-off-by: Chris Wilson <chris(a)chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin(a)intel.com>
Cc: <stable(a)vger.kernel.org> # v5.6+
---
drivers/gpu/drm/i915/gt/intel_lrc.c | 9 +++++++++
drivers/gpu/drm/i915/i915_request.c | 8 ++++++--
drivers/gpu/drm/i915/i915_scheduler.c | 6 +++---
drivers/gpu/drm/i915/i915_scheduler.h | 3 ++-
drivers/gpu/drm/i915/i915_scheduler_types.h | 1 +
5 files changed, 21 insertions(+), 6 deletions(-)
diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
index dc3f2ee7136d..10109f661bcb 100644
--- a/drivers/gpu/drm/i915/gt/intel_lrc.c
+++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
@@ -1880,6 +1880,9 @@ static void defer_request(struct i915_request *rq, struct list_head * const pl)
struct i915_request *w =
container_of(p->waiter, typeof(*w), sched);
+ if (p->flags & I915_DEPENDENCY_WEAK)
+ continue;
+
/* Leave semaphores spinning on the other engines */
if (w->engine != rq->engine)
continue;
@@ -2726,6 +2729,9 @@ static void __execlists_hold(struct i915_request *rq)
struct i915_request *w =
container_of(p->waiter, typeof(*w), sched);
+ if (p->flags & I915_DEPENDENCY_WEAK)
+ continue;
+
/* Leave semaphores spinning on the other engines */
if (w->engine != rq->engine)
continue;
@@ -2850,6 +2856,9 @@ static void __execlists_unhold(struct i915_request *rq)
struct i915_request *w =
container_of(p->waiter, typeof(*w), sched);
+ if (p->flags & I915_DEPENDENCY_WEAK)
+ continue;
+
/* Propagate any change in error status */
if (rq->fence.error)
i915_request_set_error_once(w, rq->fence.error);
diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c
index 22635bbabf06..02a5644ae105 100644
--- a/drivers/gpu/drm/i915/i915_request.c
+++ b/drivers/gpu/drm/i915/i915_request.c
@@ -1038,7 +1038,9 @@ i915_request_await_request(struct i915_request *to, struct i915_request *from)
return 0;
if (to->engine->schedule) {
- ret = i915_sched_node_add_dependency(&to->sched, &from->sched);
+ ret = i915_sched_node_add_dependency(&to->sched,
+ &from->sched,
+ I915_DEPENDENCY_EXTERNAL);
if (ret < 0)
return ret;
}
@@ -1200,7 +1202,9 @@ __i915_request_await_execution(struct i915_request *to,
/* Couple the dependency tree for PI on this exposed to->fence */
if (to->engine->schedule) {
- err = i915_sched_node_add_dependency(&to->sched, &from->sched);
+ err = i915_sched_node_add_dependency(&to->sched,
+ &from->sched,
+ I915_DEPENDENCY_WEAK);
if (err < 0)
return err;
}
diff --git a/drivers/gpu/drm/i915/i915_scheduler.c b/drivers/gpu/drm/i915/i915_scheduler.c
index 37cfcf5b321b..6e2d4190099f 100644
--- a/drivers/gpu/drm/i915/i915_scheduler.c
+++ b/drivers/gpu/drm/i915/i915_scheduler.c
@@ -462,7 +462,8 @@ bool __i915_sched_node_add_dependency(struct i915_sched_node *node,
}
int i915_sched_node_add_dependency(struct i915_sched_node *node,
- struct i915_sched_node *signal)
+ struct i915_sched_node *signal,
+ unsigned long flags)
{
struct i915_dependency *dep;
@@ -473,8 +474,7 @@ int i915_sched_node_add_dependency(struct i915_sched_node *node,
local_bh_disable();
if (!__i915_sched_node_add_dependency(node, signal, dep,
- I915_DEPENDENCY_EXTERNAL |
- I915_DEPENDENCY_ALLOC))
+ flags | I915_DEPENDENCY_ALLOC))
i915_dependency_free(dep);
local_bh_enable(); /* kick submission tasklet */
diff --git a/drivers/gpu/drm/i915/i915_scheduler.h b/drivers/gpu/drm/i915/i915_scheduler.h
index d1dc4efef77b..6f0bf00fc569 100644
--- a/drivers/gpu/drm/i915/i915_scheduler.h
+++ b/drivers/gpu/drm/i915/i915_scheduler.h
@@ -34,7 +34,8 @@ bool __i915_sched_node_add_dependency(struct i915_sched_node *node,
unsigned long flags);
int i915_sched_node_add_dependency(struct i915_sched_node *node,
- struct i915_sched_node *signal);
+ struct i915_sched_node *signal,
+ unsigned long flags);
void i915_sched_node_fini(struct i915_sched_node *node);
diff --git a/drivers/gpu/drm/i915/i915_scheduler_types.h b/drivers/gpu/drm/i915/i915_scheduler_types.h
index d18e70550054..7186875088a0 100644
--- a/drivers/gpu/drm/i915/i915_scheduler_types.h
+++ b/drivers/gpu/drm/i915/i915_scheduler_types.h
@@ -78,6 +78,7 @@ struct i915_dependency {
unsigned long flags;
#define I915_DEPENDENCY_ALLOC BIT(0)
#define I915_DEPENDENCY_EXTERNAL BIT(1)
+#define I915_DEPENDENCY_WEAK BIT(2)
};
#endif /* _I915_SCHEDULER_TYPES_H_ */
--
2.20.1
From: Sean Paul <seanpaul(a)chromium.org>
On HDCP disable, clear the repeater bit. This ensures if we connect a
non-repeater sink after a repeater, the bit is in the state we expect.
Fixes: ee5e5e7a5e0f (drm/i915: Add HDCP framework + base implementation)
Cc: Chris Wilson <chris(a)chris-wilson.co.uk>
Cc: Ramalingam C <ramalingam.c(a)intel.com>
Cc: Daniel Vetter <daniel.vetter(a)ffwll.ch>
Cc: Sean Paul <seanpaul(a)chromium.org>
Cc: Jani Nikula <jani.nikula(a)linux.intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen(a)linux.intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi(a)intel.com>
Cc: intel-gfx(a)lists.freedesktop.org
Cc: <stable(a)vger.kernel.org> # v4.17+
Reviewed-by: Ramalingam C <ramalingam.c(a)intel.com>
Signed-off-by: Sean Paul <seanpaul(a)chromium.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20191212190230.188505-3-sean@… #v2
Link: https://patchwork.freedesktop.org/patch/msgid/20200117193103.156821-3-sean@… #v3
Link: https://patchwork.freedesktop.org/patch/msgid/20200218220242.107265-3-sean@… #v4
Link: https://patchwork.freedesktop.org/patch/msgid/20200305201236.152307-3-sean@… #v5
Changes in v2:
-Added to the set
Changes in v3:
-None
I had previously agreed that clearing the rep_ctl bits on enable would
also be a good idea. However when I committed that idea to code, it
didn't look right. So let's rely on enables and disables being paired
and everything outside of that will be considered a bug
Changes in v4:
-s/I915_(READ|WRITE)/intel_de_(read|write)/
Changes in v5:
-None
Changes in v6:
-None
---
drivers/gpu/drm/i915/display/intel_hdcp.c | 6 ++++++
1 file changed, 6 insertions(+)
diff --git a/drivers/gpu/drm/i915/display/intel_hdcp.c b/drivers/gpu/drm/i915/display/intel_hdcp.c
index 525658fd201f..20175a53643d 100644
--- a/drivers/gpu/drm/i915/display/intel_hdcp.c
+++ b/drivers/gpu/drm/i915/display/intel_hdcp.c
@@ -795,6 +795,7 @@ static int _intel_hdcp_disable(struct intel_connector *connector)
struct intel_hdcp *hdcp = &connector->hdcp;
enum port port = intel_dig_port->base.port;
enum transcoder cpu_transcoder = hdcp->cpu_transcoder;
+ u32 repeater_ctl;
int ret;
drm_dbg_kms(&dev_priv->drm, "[%s:%d] HDCP is being disabled...\n",
@@ -810,6 +811,11 @@ static int _intel_hdcp_disable(struct intel_connector *connector)
return -ETIMEDOUT;
}
+ repeater_ctl = intel_hdcp_get_repeater_ctl(dev_priv, cpu_transcoder,
+ port);
+ intel_de_write(dev_priv, HDCP_REP_CTL,
+ intel_de_read(dev_priv, HDCP_REP_CTL) & ~repeater_ctl);
+
ret = hdcp->shim->toggle_signalling(intel_dig_port, false);
if (ret) {
drm_err(&dev_priv->drm, "Failed to disable HDCP signalling\n");
--
Sean Paul, Software Engineer, Google / Chromium OS
From: Sean Paul <seanpaul(a)chromium.org>
This patch fixes a few bugs:
1- We weren't taking into account sha_leftovers when adding multiple
ksvs to sha_text. As such, we were or'ing the end of ksv[j - 1] with
the beginning of ksv[j]
2- In the sha_leftovers == 2 and sha_leftovers == 3 case, bstatus was
being placed on the wrong half of sha_text, overlapping the leftover
ksv value
3- In the sha_leftovers == 2 case, we need to manually terminate the
byte stream with 0x80 since the hardware doesn't have enough room to
add it after writing M0
The upside is that all of the HDCP supported HDMI repeaters I could
find on Amazon just strip HDCP anyways, so it turns out to be _really_
hard to hit any of these cases without an MST hub, which is not (yet)
supported. Oh, and the sha_leftovers == 1 case works perfectly!
Fixes: ee5e5e7a5e0f (drm/i915: Add HDCP framework + base implementation)
Cc: Chris Wilson <chris(a)chris-wilson.co.uk>
Cc: Ramalingam C <ramalingam.c(a)intel.com>
Cc: Daniel Vetter <daniel.vetter(a)ffwll.ch>
Cc: Sean Paul <seanpaul(a)chromium.org>
Cc: Jani Nikula <jani.nikula(a)linux.intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen(a)linux.intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi(a)intel.com>
Cc: intel-gfx(a)lists.freedesktop.org
Cc: <stable(a)vger.kernel.org> # v4.17+
Reviewed-by: Ramalingam C <ramalingam.c(a)intel.com>
Signed-off-by: Sean Paul <seanpaul(a)chromium.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20191203173638.94919-2-sean@p… #v1
Link: https://patchwork.freedesktop.org/patch/msgid/20191212190230.188505-2-sean@… #v2
Link: https://patchwork.freedesktop.org/patch/msgid/20200117193103.156821-2-sean@… #v3
Link: https://patchwork.freedesktop.org/patch/msgid/20200218220242.107265-2-sean@… #v4
Link: https://patchwork.freedesktop.org/patch/msgid/20200305201236.152307-2-sean@… #v5
Changes in v2:
-None
Changes in v3:
-None
Changes in v4:
-Rebased on intel_de_write changes
Changes in v5:
-None
Changes in v6:
-None
---
drivers/gpu/drm/i915/display/intel_hdcp.c | 26 +++++++++++++++++------
include/drm/drm_hdcp.h | 3 +++
2 files changed, 23 insertions(+), 6 deletions(-)
diff --git a/drivers/gpu/drm/i915/display/intel_hdcp.c b/drivers/gpu/drm/i915/display/intel_hdcp.c
index 2cbc4619b4ce..525658fd201f 100644
--- a/drivers/gpu/drm/i915/display/intel_hdcp.c
+++ b/drivers/gpu/drm/i915/display/intel_hdcp.c
@@ -336,8 +336,10 @@ int intel_hdcp_validate_v_prime(struct intel_connector *connector,
/* Fill up the empty slots in sha_text and write it out */
sha_empty = sizeof(sha_text) - sha_leftovers;
- for (j = 0; j < sha_empty; j++)
- sha_text |= ksv[j] << ((sizeof(sha_text) - j - 1) * 8);
+ for (j = 0; j < sha_empty; j++) {
+ u8 off = ((sizeof(sha_text) - j - 1 - sha_leftovers) * 8);
+ sha_text |= ksv[j] << off;
+ }
ret = intel_write_sha_text(dev_priv, sha_text);
if (ret < 0)
@@ -435,7 +437,7 @@ int intel_hdcp_validate_v_prime(struct intel_connector *connector,
/* Write 32 bits of text */
intel_de_write(dev_priv, HDCP_REP_CTL,
rep_ctl | HDCP_SHA1_TEXT_32);
- sha_text |= bstatus[0] << 24 | bstatus[1] << 16;
+ sha_text |= bstatus[0] << 8 | bstatus[1];
ret = intel_write_sha_text(dev_priv, sha_text);
if (ret < 0)
return ret;
@@ -450,17 +452,29 @@ int intel_hdcp_validate_v_prime(struct intel_connector *connector,
return ret;
sha_idx += sizeof(sha_text);
}
+
+ /*
+ * Terminate the SHA-1 stream by hand. For the other leftover
+ * cases this is appended by the hardware.
+ */
+ intel_de_write(dev_priv, HDCP_REP_CTL,
+ rep_ctl | HDCP_SHA1_TEXT_32);
+ sha_text = DRM_HDCP_SHA1_TERMINATOR << 24;
+ ret = intel_write_sha_text(dev_priv, sha_text);
+ if (ret < 0)
+ return ret;
+ sha_idx += sizeof(sha_text);
} else if (sha_leftovers == 3) {
- /* Write 32 bits of text */
+ /* Write 32 bits of text (filled from LSB) */
intel_de_write(dev_priv, HDCP_REP_CTL,
rep_ctl | HDCP_SHA1_TEXT_32);
- sha_text |= bstatus[0] << 24;
+ sha_text |= bstatus[0];
ret = intel_write_sha_text(dev_priv, sha_text);
if (ret < 0)
return ret;
sha_idx += sizeof(sha_text);
- /* Write 8 bits of text, 24 bits of M0 */
+ /* Write 8 bits of text (filled from LSB), 24 bits of M0 */
intel_de_write(dev_priv, HDCP_REP_CTL,
rep_ctl | HDCP_SHA1_TEXT_8);
ret = intel_write_sha_text(dev_priv, bstatus[1]);
diff --git a/include/drm/drm_hdcp.h b/include/drm/drm_hdcp.h
index c6bab4986a65..fe58dbb46962 100644
--- a/include/drm/drm_hdcp.h
+++ b/include/drm/drm_hdcp.h
@@ -29,6 +29,9 @@
/* Slave address for the HDCP registers in the receiver */
#define DRM_HDCP_DDC_ADDR 0x3A
+/* Value to use at the end of the SHA-1 bytestream used for repeaters */
+#define DRM_HDCP_SHA1_TERMINATOR 0x80
+
/* HDCP register offsets for HDMI/DVI devices */
#define DRM_HDCP_DDC_BKSV 0x00
#define DRM_HDCP_DDC_RI_PRIME 0x08
--
Sean Paul, Software Engineer, Google / Chromium OS