[PATCH 1/1 linux-next RESEND] task IO accounting: add conditional read/write counters

[PATCH 1/1 linux-next RESEND] task IO accounting: add conditional read/write counters

[Fabian Frederick]

This patch adds syscre and syscwe respectively effective read/write
syscall counters. We already had syscr/syscw which were
unconditionally incremented. Doing the difference between those
variables could give interesting statistics.

Signed-off-by: Fabian Frederick <fabf@skynet.be>
---
 Documentation/filesystems/proc.txt |   16 ++++++++++++++++
 fs/proc/base.c                     |    4 ++++
 include/linux/sched.h              |    2 ++
 include/linux/task_io_accounting.h |    4 ++++
 4 files changed, 26 insertions(+)

diff --git a/Documentation/filesystems/proc.txt b/Documentation/filesystems/proc.txt
index 6d37062..34fb7ee 100644
--- a/Documentation/filesystems/proc.txt
+++ b/Documentation/filesystems/proc.txt
@@ -1541,6 +1541,22 @@ I/O counter: write syscalls
 Attempt to count the number of write I/O operations, i.e. syscalls like
 write() and pwrite().
 
+syscre
+------
+
+I/O counter: Effective read syscalls
+Attempt to count the number of effective read I/O operations,
+i.e. syscalls like read() and pread().
+
+
+syscwe
+------
+
+I/O counter: Effective write syscalls
+Attempt to count the number of effective write I/O operations,
+i.e. syscalls like read() and pread().
+
+
 
 read_bytes
 ----------
diff --git a/fs/proc/base.c b/fs/proc/base.c
index 7a3b82f..54a7c93 100644
--- a/fs/proc/base.c
+++ b/fs/proc/base.c
@@ -2367,6 +2367,8 @@ static int do_io_accounting(struct task_struct *task, struct seq_file *m, int wh
 		   "wchar: %llu\n"
 		   "syscr: %llu\n"
 		   "syscw: %llu\n"
+		   "syscre: %llu\n"
+		   "syscwe: %llu\n"
 		   "read_bytes: %llu\n"
 		   "write_bytes: %llu\n"
 		   "cancelled_write_bytes: %llu\n",
@@ -2374,6 +2376,8 @@ static int do_io_accounting(struct task_struct *task, struct seq_file *m, int wh
 		   (unsigned long long)acct.wchar,
 		   (unsigned long long)acct.syscr,
 		   (unsigned long long)acct.syscw,
+		   (unsigned long long)acct.syscre,
+		   (unsigned long long)acct.syscwe,
 		   (unsigned long long)acct.read_bytes,
 		   (unsigned long long)acct.write_bytes,
 		   (unsigned long long)acct.cancelled_write_bytes);
diff --git a/include/linux/sched.h b/include/linux/sched.h
index 6d77432..ec316ee 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -3020,11 +3020,13 @@ extern int task_can_switch_user(struct user_struct *up,
 #ifdef CONFIG_TASK_XACCT
 static inline void add_rchar(struct task_struct *tsk, ssize_t amt)
 {
+	tsk->ioac.syscre++;
 	tsk->ioac.rchar += amt;
 }
 
 static inline void add_wchar(struct task_struct *tsk, ssize_t amt)
 {
+	tsk->ioac.syscwe++;
 	tsk->ioac.wchar += amt;
 }
 
diff --git a/include/linux/task_io_accounting.h b/include/linux/task_io_accounting.h
index bdf855c..bc94291 100644
--- a/include/linux/task_io_accounting.h
+++ b/include/linux/task_io_accounting.h
@@ -18,6 +18,10 @@ struct task_io_accounting {
 	u64 syscr;
 	/* # of write syscalls */
 	u64 syscw;
+	/* # of effective read syscalls */
+	u64 syscre;
+	/* # of effective write syscalls */
+	u64 syscwe;
 #endif /* CONFIG_TASK_XACCT */
 
 #ifdef CONFIG_TASK_IO_ACCOUNTING
-- 
1.7.9.5

Re: [PATCH 1/1 linux-next RESEND] task IO accounting: add conditional read/write counters

[Andrew Morton]

On Tue,  3 Mar 2015 21:16:35 +0100 Fabian Frederick <fabf@skynet.be> wrote:

> This patch adds syscre and syscwe respectively effective read/write
> syscall counters. We already had syscr/syscw which were
> unconditionally incremented. Doing the difference between those
> variables could give interesting statistics.

I don't think that "could give interesting statistics" is sufficient
justification for adding overhead and extending the Linux userspace
interface.

More specificity would help, including real-worldish use cases which
demonstrate the value.