From: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
To: linux-rt-users@vger.kernel.org
Cc: linux-kernel@vger.kernel.org, tglx@linutronix.de,
Steven Rostedt <rostedt@goodmis.org>,
Catalin Marinas <catalin.marinas@arm.com>,
Will Deacon <will.deacon@arm.com>,
linux-arm-kernel@lists.infradead.org
Subject: [PATCH RT] arm64: fpsimd: use a local_lock() in addition to local_bh_disable()
Date: Thu, 17 May 2018 14:40:06 +0200 [thread overview]
Message-ID: <20180517124006.ohygrrpg7z2moqqt@linutronix.de> (raw)
In v4.16-RT I noticed a number of warnings from task_fpsimd_load(). The
code disables BH and expects that it is not preemptible. On -RT the
task remains preemptible but remains the same CPU. This may corrupt the
content of the SIMD registers if the task is preempted during
saving/restoring those registers.
Add a locallock around next to the local_bh_disable(). This fulfill the
requirement that the code is not invoked again in different context on
the same CPU while it remains preemptible.
The preempt_disable() + local_bh_enable() combo in kernel_neon_begin()
is not working on -RT. We don't use NEON in kernel mode on RT right now
but this still should be addressed.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
---
arch/arm64/kernel/fpsimd.c | 20 ++++++++++++++++++--
1 file changed, 18 insertions(+), 2 deletions(-)
diff --git a/arch/arm64/kernel/fpsimd.c b/arch/arm64/kernel/fpsimd.c
index e7226c4c7493..3a5cd1908874 100644
--- a/arch/arm64/kernel/fpsimd.c
+++ b/arch/arm64/kernel/fpsimd.c
@@ -38,6 +38,7 @@
#include <linux/signal.h>
#include <linux/slab.h>
#include <linux/sysctl.h>
+#include <linux/locallock.h>
#include <asm/fpsimd.h>
#include <asm/cputype.h>
@@ -235,7 +236,7 @@ static void sve_user_enable(void)
* whether TIF_SVE is clear or set, since these are not vector length
* dependent.
*/
-
+static DEFINE_LOCAL_IRQ_LOCK(fpsimd_lock);
/*
* Update current's FPSIMD/SVE registers from thread_struct.
*
@@ -594,6 +595,7 @@ int sve_set_vector_length(struct task_struct *task,
* non-SVE thread.
*/
if (task == current) {
+ local_lock(fpsimd_lock);
local_bh_disable();
task_fpsimd_save();
@@ -604,8 +606,10 @@ int sve_set_vector_length(struct task_struct *task,
if (test_and_clear_tsk_thread_flag(task, TIF_SVE))
sve_to_fpsimd(task);
- if (task == current)
+ if (task == current) {
+ local_unlock(fpsimd_lock);
local_bh_enable();
+ }
/*
* Force reallocation of task SVE state to the correct size
@@ -838,6 +842,7 @@ asmlinkage void do_sve_acc(unsigned int esr, struct pt_regs *regs)
sve_alloc(current);
local_bh_disable();
+ local_lock(fpsimd_lock);
task_fpsimd_save();
fpsimd_to_sve(current);
@@ -849,6 +854,7 @@ asmlinkage void do_sve_acc(unsigned int esr, struct pt_regs *regs)
if (test_and_set_thread_flag(TIF_SVE))
WARN_ON(1); /* SVE access shouldn't have trapped */
+ local_unlock(fpsimd_lock);
local_bh_enable();
}
@@ -926,6 +932,7 @@ void fpsimd_flush_thread(void)
return;
local_bh_disable();
+ local_lock(fpsimd_lock);
memset(¤t->thread.fpsimd_state, 0, sizeof(struct fpsimd_state));
fpsimd_flush_task_state(current);
@@ -967,6 +974,7 @@ void fpsimd_flush_thread(void)
set_thread_flag(TIF_FOREIGN_FPSTATE);
+ local_unlock(fpsimd_lock);
local_bh_enable();
}
@@ -980,7 +988,9 @@ void fpsimd_preserve_current_state(void)
return;
local_bh_disable();
+ local_lock(fpsimd_lock);
task_fpsimd_save();
+ local_unlock(fpsimd_lock);
local_bh_enable();
}
@@ -1022,12 +1032,14 @@ void fpsimd_restore_current_state(void)
return;
local_bh_disable();
+ local_lock(fpsimd_lock);
if (test_and_clear_thread_flag(TIF_FOREIGN_FPSTATE)) {
task_fpsimd_load();
fpsimd_bind_to_cpu();
}
+ local_unlock(fpsimd_lock);
local_bh_enable();
}
@@ -1042,6 +1054,7 @@ void fpsimd_update_current_state(struct user_fpsimd_state const *state)
return;
local_bh_disable();
+ local_lock(fpsimd_lock);
current->thread.fpsimd_state.user_fpsimd = *state;
if (system_supports_sve() && test_thread_flag(TIF_SVE))
@@ -1052,6 +1065,7 @@ void fpsimd_update_current_state(struct user_fpsimd_state const *state)
if (test_and_clear_thread_flag(TIF_FOREIGN_FPSTATE))
fpsimd_bind_to_cpu();
+ local_unlock(fpsimd_lock);
local_bh_enable();
}
@@ -1116,6 +1130,7 @@ void kernel_neon_begin(void)
BUG_ON(!may_use_simd());
local_bh_disable();
+ local_lock(fpsimd_lock);
__this_cpu_write(kernel_neon_busy, true);
@@ -1128,6 +1143,7 @@ void kernel_neon_begin(void)
/* Invalidate any task state remaining in the fpsimd regs: */
fpsimd_flush_cpu_state();
+ local_unlock(fpsimd_lock);
preempt_disable();
local_bh_enable();
--
2.17.0
next reply other threads:[~2018-05-17 12:40 UTC|newest]
Thread overview: 34+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-05-17 12:40 Sebastian Andrzej Siewior [this message]
2018-05-17 18:19 ` [PATCH RT] arm64: fpsimd: use a local_lock() in addition to local_bh_disable() Dave Martin
2018-05-18 12:46 ` Dave Martin
2018-05-23 14:34 ` Sebastian Andrzej Siewior
2018-05-23 14:31 ` Sebastian Andrzej Siewior
2018-05-23 14:55 ` Dave Martin
2018-05-22 17:10 ` Steven Rostedt
2018-05-22 17:21 ` Sebastian Andrzej Siewior
2018-05-22 17:24 ` Steven Rostedt
2018-05-22 17:33 ` Sebastian Andrzej Siewior
2018-07-11 13:25 ` Steven Rostedt
2018-07-11 13:31 ` Sebastian Andrzej Siewior
2018-07-11 13:33 ` Steven Rostedt
2018-07-13 17:49 ` [PATCH RT v2] " Sebastian Andrzej Siewior
2018-07-13 17:50 ` [PATCH RT] locallock: add local_lock_bh() Sebastian Andrzej Siewior
2018-07-13 22:03 ` [PATCH RT v2] arm64: fpsimd: use a local_lock() in addition to local_bh_disable() Mike Galbraith
2018-07-15 7:22 ` Mike Galbraith
2018-07-18 10:30 ` Mike Galbraith
2018-07-18 9:27 ` Sebastian Andrzej Siewior
2018-07-18 10:28 ` Mike Galbraith
2018-07-18 10:36 ` Sebastian Andrzej Siewior
2018-07-16 15:17 ` Dave Martin
2018-07-18 9:12 ` Sebastian Andrzej Siewior
2018-07-18 9:24 ` Sebastian Andrzej Siewior
2018-07-24 14:45 ` Dave Martin
2018-07-24 15:15 ` Ard Biesheuvel
2018-07-24 13:46 ` Steven Rostedt
2018-07-24 13:57 ` Sebastian Andrzej Siewior
2018-07-26 15:06 ` [PATCH RT v3] arm64: fpsimd: use preemp_disable " Sebastian Andrzej Siewior
2018-07-27 3:17 ` Mike Galbraith
2018-07-27 7:56 ` Sebastian Andrzej Siewior
2018-07-27 15:35 ` Dave Martin
2018-07-27 16:26 ` Sebastian Andrzej Siewior
2018-07-11 17:07 ` [PATCH RT] arm64: fpsimd: use a local_lock() " Mike Galbraith
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20180517124006.ohygrrpg7z2moqqt@linutronix.de \
--to=bigeasy@linutronix.de \
--cc=catalin.marinas@arm.com \
--cc=linux-arm-kernel@lists.infradead.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-rt-users@vger.kernel.org \
--cc=rostedt@goodmis.org \
--cc=tglx@linutronix.de \
--cc=will.deacon@arm.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).