Skip to contentSkip to conversation

syscalls/preadv203: Add basic RWF_NOWAIT test

PATCH Superseded Cyril Hrubis 2 participants 5 messages Apr 4, 2019 – May 29, 2019

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 syscalls/preadv203: Add basic RWF_NOWAIT test

We are attempting to trigger the EAGAIN path for the RWF_NOWAIT flag.

In order to do so the test runs three threads:

* nowait_reader: reads from a random offset from a random file with RWF_NOWAIT flag and expects to get EAGAIN and short read sooner or later

* writer_thread: rewrites random file in order to keep the underlying device bussy so that pages evicted from cache cannot be faulted immediatelly

* cache_dropper: attempts to evict pages from a cache in order for reader to hit evicted page sooner or later

---

This is just the same as previous patch + small (typo) fixes and fallback definition for RWF_NOWAIT.

I've also looked into the BLKFLSBUF ioctl() however that one only seem to drop cached pages but no pages schedulled for writeback, hence it does not generate I/O pressure at all.

 include/lapi/preadv2.h                        |   4 +
 runtest/syscalls                              |   2 +
 testcases/kernel/syscalls/preadv2/.gitignore  |   2 +
 testcases/kernel/syscalls/preadv2/Makefile    |   4 +
 testcases/kernel/syscalls/preadv2/preadv203.c | 267 ++++++++++++++++++
 5 files changed, 279 insertions(+)
 create mode 100644 testcases/kernel/syscalls/preadv2/preadv203.c
include/lapi/preadv2.h +4 -0
index aacf96fbf..538ed7218 100644
@@ -10,6 +10,10 @@
 #include "config.h"
 #include "lapi/syscalls.h"
 
+#ifndef RWF_NOWAIT
+# define RWF_NOWAIT 0x00000008
+#endif
+
 #if !defined(HAVE_PREADV2)
 
 /* LO_HI_LONG taken from glibc */
runtest/syscalls +2 -0
index cf8189ebd..81906082c 100644
@@ -873,6 +873,8 @@ preadv201 preadv201
 preadv201_64 preadv201_64
 preadv202 preadv202
 preadv202_64 preadv202_64
+preadv203 preadv203
+preadv203_64 preadv203_64
 
 profil01 profil01
 
testcases/kernel/syscalls/preadv2/.gitignore +2 -0
index 759d9ef5b..98b81abea 100644
@@ -2,3 +2,5 @@
 /preadv201_64
 /preadv202
 /preadv202_64
+/preadv203
+/preadv203_64
testcases/kernel/syscalls/preadv2/Makefile +4 -0
index fc1fbf3c7..fbedd0287 100644
@@ -11,4 +11,8 @@ include $(abs_srcdir)/../utils/newer_64.mk
 
 %_64: CPPFLAGS += -D_FILE_OFFSET_BITS=64
 
+preadv203: CFLAGS += -pthread
+preadv203_64: CFLAGS += -pthread
+preadv203_64: LDFLAGS += -pthread
+
 include $(top_srcdir)/include/mk/generic_leaf_target.mk
testcases/kernel/syscalls/preadv2/preadv203.c new file +267 -0

267 lines not on this page · raw message

Signature · 1 line
2.19.2
#1 lore raw

PATCH v2 1/1 syscalls/preadv203: Add basic RWF_NOWAIT test

Hi!
Ping, anybody has any comments or can be this be this merged?

Signature · 2 lines
Cyril Hrubis
[email protected]
#2 lore raw

PATCH v2 1/1 syscalls/preadv203: Add basic RWF_NOWAIT test

Quoted 4 lines · Cyril Hrubis

Hi!
Ping, anybody has any comments or can be this be this merged?

I have no objection to this test, but I have a question. Your claim is:
"We are attempting to trigger the EAGAIN path for the RWF_NOWAIT flag."

Seems like you have gone to a great length of effort to achieve this. Wouldn't it be sufficient to do a single preadv2 after fsync+drop_caches in order to observe EAGAIN?

There are several other ways to get EAGAIN, like DIO read/write when page cache *is* populated.

Question is, how much is the test trying to cover.

Thanks,
Amir.

#3 lore raw

PATCH v2 1/1 syscalls/preadv203: Add basic RWF_NOWAIT test

Hi!

Quoted 9 lines
Quoted 2 lines

Ping, anybody has any comments or can be this be this merged?

I have no objection to this test, but I have a question. Your claim is:
"We are attempting to trigger the EAGAIN path for the RWF_NOWAIT flag."

Seems like you have gone to a great length of effort to achieve this. Wouldn't it be sufficient to do a single preadv2 after fsync+drop_caches in order to observe EAGAIN?

As far as I can tell the I/O to the device has to be saturated in order to get EAGAIN at least for a subset of filesystems. If I just drop the cache once then attempt to read with the NOWAIT flag it strangely works only for ext2 for me.

So I guess that the manual page is a bit misleading and the NOWAIT flag actually returns EAGAIN only if the calling thread would end up in a queue or waiting for a lock, but I haven't looked into the kernel code to make sure.

Quoted 2 lines

There are several other ways to get EAGAIN, like DIO read/write when page cache *is* populated.

That is a good hint, I will have a look at that case as well.

Quoted 1 line

Question is, how much is the test trying to cover.

Well ideally we should excercise as much kernel code paths as possible... :-)

Signature · 2 lines
Cyril Hrubis
[email protected]