From: Frederic Weisbecker <frederic@kernel.org>
To: LKML <linux-kernel@vger.kernel.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>,
Sebastian Andrzej Siewior <bigeasy@linutronix.de>,
Peter Zijlstra <peterz@infradead.org>,
"David S . Miller" <davem@davemloft.net>,
Linus Torvalds <torvalds@linux-foundation.org>,
Thomas Gleixner <tglx@linutronix.de>,
Frederic Weisbecker <frederic@kernel.org>,
"Paul E . McKenney" <paulmck@linux.vnet.ibm.com>,
Ingo Molnar <mingo@kernel.org>,
Mauro Carvalho Chehab <mchehab@s-opensource.com>
Subject: [RFC PATCH 24/30] softirq: Introduce Local_bh_enter/exit()
Date: Thu, 11 Oct 2018 01:12:11 +0200 [thread overview]
Message-ID: <1539213137-13953-25-git-send-email-frederic@kernel.org> (raw)
In-Reply-To: <1539213137-13953-1-git-send-email-frederic@kernel.org>
From: Frederic Weisbecker <fweisbec@gmail.com>
So far, the state of handling the disablement of softirqs and processing
their callbacks have been handled the same: increment the softirq offset,
trace softirqs off, preempt off, etc...
The only difference remains in the way the preempt count is incremented:
by 1 for softirq processing (can't nest as softirqs processing aren't
re-entrant) and by 2 for softirq disablement (can nest).
Now their behaviour is going to drift entirely. Softirq processing will
need to be reentrant and accept stacking SOFTIRQ_OFFSET increments.
OTOH softirq disablement will be driven by the vector enabled mask and
toggled only once any vector get disabled.
Maintaining both behaviours under the same handler is going to be messy,
so move the preempt count related code on softirq processing to its own
handlers.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: David S. Miller <davem@davemloft.net>
Cc: Mauro Carvalho Chehab <mchehab@s-opensource.com>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
---
kernel/softirq.c | 74 ++++++++++++++++++++++++++++++++++++++++++++------------
1 file changed, 58 insertions(+), 16 deletions(-)
diff --git a/kernel/softirq.c b/kernel/softirq.c
index ae9e29f..22cc0a7 100644
--- a/kernel/softirq.c
+++ b/kernel/softirq.c
@@ -139,19 +139,6 @@ void __local_bh_disable_ip(unsigned long ip, unsigned int cnt)
EXPORT_SYMBOL(__local_bh_disable_ip);
#endif /* CONFIG_TRACE_IRQFLAGS */
-static void __local_bh_enable(unsigned int cnt)
-{
- lockdep_assert_irqs_disabled();
-
- if (preempt_count() == cnt)
- trace_preempt_on(CALLER_ADDR0, get_lock_parent_ip());
-
- if (softirq_count() == (cnt & SOFTIRQ_MASK))
- trace_softirqs_on(_RET_IP_);
-
- __preempt_count_sub(cnt);
-}
-
/*
* Special-case - softirqs can safely be enabled by __do_softirq(),
* without processing still-pending softirqs:
@@ -159,7 +146,16 @@ static void __local_bh_enable(unsigned int cnt)
void local_bh_enable_no_softirq(void)
{
WARN_ON_ONCE(in_irq());
- __local_bh_enable(SOFTIRQ_DISABLE_OFFSET);
+ lockdep_assert_irqs_disabled();
+
+ if (preempt_count() == SOFTIRQ_DISABLE_OFFSET)
+ trace_preempt_on(CALLER_ADDR0, get_lock_parent_ip());
+
+ if (softirq_count() == SOFTIRQ_DISABLE_OFFSET)
+ trace_softirqs_on(_RET_IP_);
+
+ __preempt_count_sub(SOFTIRQ_DISABLE_OFFSET);
+
}
EXPORT_SYMBOL(local_bh_enable_no_softirq);
@@ -207,6 +203,52 @@ void local_bh_enable_all(void)
local_bh_enable(SOFTIRQ_ALL_MASK);
}
+static void local_bh_enter(unsigned long ip)
+{
+ unsigned long flags;
+
+ WARN_ON_ONCE(in_irq());
+
+ raw_local_irq_save(flags);
+ /*
+ * The preempt tracer hooks into preempt_count_add and will break
+ * lockdep because it calls back into lockdep after SOFTIRQ_OFFSET
+ * is set and before current->softirq_enabled is cleared.
+ * We must manually increment preempt_count here and manually
+ * call the trace_preempt_off later.
+ */
+ __preempt_count_add(SOFTIRQ_OFFSET);
+ /*
+ * Were softirqs turned off above:
+ */
+ if (softirq_count() == SOFTIRQ_OFFSET)
+ trace_softirqs_off(ip);
+ raw_local_irq_restore(flags);
+
+ if (preempt_count() == SOFTIRQ_OFFSET) {
+#ifdef CONFIG_DEBUG_PREEMPT
+ current->preempt_disable_ip = get_lock_parent_ip();
+#endif
+ trace_preempt_off(CALLER_ADDR0, get_lock_parent_ip());
+ }
+}
+
+static void local_bh_exit(void)
+{
+ lockdep_assert_irqs_disabled();
+
+ if (preempt_count() == SOFTIRQ_OFFSET)
+ trace_preempt_on(CALLER_ADDR0, get_lock_parent_ip());
+
+ if (softirq_count() == SOFTIRQ_OFFSET)
+ trace_softirqs_on(_RET_IP_);
+
+ __preempt_count_sub(SOFTIRQ_OFFSET);
+}
+
+
+
+
/*
* We restart softirq processing for at most MAX_SOFTIRQ_RESTART times,
* but break the loop if need_resched() is set or after 2 ms.
@@ -276,7 +318,7 @@ asmlinkage __visible void __softirq_entry __do_softirq(void)
pending = local_softirq_pending() & local_softirq_enabled();
account_irq_enter_time(current);
- __local_bh_disable_ip(_RET_IP_, SOFTIRQ_OFFSET);
+ local_bh_enter(_RET_IP_);
in_hardirq = lockdep_softirq_start();
restart:
@@ -325,7 +367,7 @@ asmlinkage __visible void __softirq_entry __do_softirq(void)
lockdep_softirq_end(in_hardirq);
account_irq_exit_time(current);
- __local_bh_enable(SOFTIRQ_OFFSET);
+ local_bh_exit();
WARN_ON_ONCE(in_interrupt());
current_restore_flags(old_flags, PF_MEMALLOC);
}
--
2.7.4
next prev parent reply other threads:[~2018-10-10 23:13 UTC|newest]
Thread overview: 40+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-10-10 23:11 [RFC PATCH 00/30] softirq: Make softirqs soft-interruptible (+ per vector disablement) Frederic Weisbecker
2018-10-10 23:11 ` [RFC PATCH 01/30] x86: Revert "x86/irq: Demote irq_cpustat_t::__softirq_pending to u16" Frederic Weisbecker
2018-10-10 23:11 ` [RFC PATCH 02/30] arch/softirq: Rename softirq_pending fields to softirq_data Frederic Weisbecker
2018-10-10 23:11 ` [RFC PATCH 03/30] softirq: Implement local_softirq_pending() below softirq vector definition Frederic Weisbecker
2018-10-10 23:11 ` [RFC PATCH 04/30] softirq: Normalize softirq_pending naming scheme Frederic Weisbecker
2018-10-10 23:11 ` [RFC PATCH 05/30] softirq: Convert softirq_pending_set() to softirq_pending_nand() Frederic Weisbecker
2018-10-10 23:11 ` [RFC PATCH 06/30] softirq: Introduce disabled softirq vectors bits Frederic Weisbecker
2018-10-10 23:11 ` [RFC PATCH 07/30] softirq: Rename _local_bh_enable() to local_bh_enable_no_softirq() Frederic Weisbecker
2018-10-10 23:11 ` [RFC PATCH 08/30] softirq: Move vectors bits to bottom_half.h Frederic Weisbecker
2018-10-10 23:11 ` [RFC PATCH 09/30] x86: Init softirq enabled field Frederic Weisbecker
2018-10-10 23:11 ` [RFC PATCH 10/30] softirq: Check enabled bits on the softirq loop Frederic Weisbecker
2018-10-10 23:11 ` [RFC PATCH 11/30] net: Prepare netif_tx_lock_bh/netif_tx_unlock_bh() for handling softirq mask Frederic Weisbecker
2018-10-10 23:11 ` [RFC PATCH 12/30] rcu: Prepare rcu_read_[un]lock_bh() " Frederic Weisbecker
2018-10-16 5:28 ` Joel Fernandes
2018-10-17 0:44 ` Frederic Weisbecker
2018-10-17 0:55 ` Joel Fernandes
2018-10-10 23:12 ` [RFC PATCH 13/30] net: Prepare tcp_get_md5sig_pool() " Frederic Weisbecker
2018-10-10 23:12 ` [RFC PATCH 14/30] softirq: Introduce local_bh_disable_all() Frederic Weisbecker
2018-10-10 23:12 ` [RFC PATCH 15/30] net: Prepare [un]lock_sock_fast() for handling softirq mask Frederic Weisbecker
2018-10-10 23:12 ` [RFC PATCH 16/30] net: Prepare nf_log_buf_open() " Frederic Weisbecker
2018-10-10 23:12 ` [RFC PATCH 17/30] isdn: Prepare isdn_net_get_locked_lp() " Frederic Weisbecker
2018-10-10 23:12 ` [RFC PATCH 18/30] softirq: Prepare local_bh_disable() " Frederic Weisbecker
2018-10-10 23:12 ` [RFC PATCH 19/30] diva: Prepare diva_os_enter_spin_lock() " Frederic Weisbecker
2018-10-10 23:12 ` [RFC PATCH 20/30] tg3: Prepare tg3_full_[un]lock() " Frederic Weisbecker
2018-10-10 23:12 ` [RFC PATCH 21/30] locking: Prepare spin_lock_bh() " Frederic Weisbecker
2018-10-10 23:12 ` [RFC PATCH 22/30] seqlock: Prepare write_seq[un]lock_bh() " Frederic Weisbecker
2018-10-10 23:12 ` [RFC PATCH 23/30] rwlock: Prepare write_[un]lock_bh() " Frederic Weisbecker
2018-10-10 23:12 ` Frederic Weisbecker [this message]
2018-10-10 23:12 ` [RFC PATCH 25/30] softirq: Push down softirq mask to __local_bh_disable_ip() Frederic Weisbecker
2018-10-10 23:12 ` [RFC PATCH 26/30] softirq: Increment the softirq offset on top of enabled bits Frederic Weisbecker
2018-10-10 23:12 ` [RFC PATCH 27/30] softirq: Swap softirq serving VS disable on preempt mask layout Frederic Weisbecker
2018-10-10 23:12 ` [RFC PATCH 28/30] softirq: Disable vector on execution Frederic Weisbecker
2018-10-10 23:12 ` [RFC PATCH 29/30] softirq: Make softirq processing softinterruptible Frederic Weisbecker
2018-10-16 4:15 ` Pavan Kondeti
2018-10-17 0:26 ` Frederic Weisbecker
2018-10-22 8:12 ` Pavan Kondeti
2018-10-10 23:12 ` [RFC PATCH 30/30] softirq: Tasklet/net-rx fixup Frederic Weisbecker
2018-10-16 22:03 ` [RFC PATCH 00/30] softirq: Make softirqs soft-interruptible (+ per vector disablement) Jonathan Corbet
2018-10-16 23:37 ` Richard Cochran
2018-10-17 1:20 ` Frederic Weisbecker
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=1539213137-13953-25-git-send-email-frederic@kernel.org \
--to=frederic@kernel.org \
--cc=bigeasy@linutronix.de \
--cc=davem@davemloft.net \
--cc=fweisbec@gmail.com \
--cc=linux-kernel@vger.kernel.org \
--cc=mchehab@s-opensource.com \
--cc=mingo@kernel.org \
--cc=paulmck@linux.vnet.ibm.com \
--cc=peterz@infradead.org \
--cc=tglx@linutronix.de \
--cc=torvalds@linux-foundation.org \
/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