Skip to contentSkip to conversation

block: improve error reporting for unsupported O_DIRECT

PATCH Superseded Daniel P. Berrangé 4 participants 8 messages Jul 1, 2020 – Jul 2, 2020 Block layer

v1 of 6 Superseded Latest is v6 →

lore
Share & tools
lore
#6 lore raw

PATCH 3/3 block: switch to use qemu_open_err for improved errors

In reply to Daniel P. Berrangé

Currently at startup if using cache=none on a filesystem lacking O_DIRECT such as tmpfs, at startup QEMU prints

qemu-system-x86_64: -drive file=/tmp/foo.img,cache=none: file system may not support O_DIRECT
qemu-system-x86_64: -drive file=/tmp/foo.img,cache=none: Could not open '/tmp/foo.img': Invalid argument

while at QMP level the hint is missing, so QEMU reports just

  "error": {
      "class": "GenericError",
      "desc": "Could not open '/tmp/foo.img': Invalid argument"
  }

which is close to useless for the end user trying to figure out what they did wrong

With this change at startup QEMU prints

qemu-system-x86_64: -drive file=/tmp/foo.img,cache=none: Unable to open '/tmp/foo.img' flags 0x4000: filesystem does not support O_DIRECT

while at the QMP level QEMU reports a massively more informative

  "error": {
     "class": "GenericError",
     "desc": "Unable to open '/tmp/foo.img' flags 0x4002: filesystem does not support O_DIRECT"
  }
---
 block/file-posix.c | 10 ++++------
 1 file changed, 4 insertions(+), 6 deletions(-)
block/file-posix.c +4 -6
index 3ab8f5a0fa..2865b789fb 100644
@@ -574,11 +574,10 @@ static int raw_open_common(BlockDriverState *bs, QDict *options,
     raw_parse_flags(bdrv_flags, &s->open_flags, false);
 
     s->fd = -1;
-    fd = qemu_open(filename, s->open_flags, 0644);
+    fd = qemu_open_err(filename, s->open_flags, errp, 0644);
     ret = fd < 0 ? -errno : 0;
 
     if (ret < 0) {
-        error_setg_file_open(errp, -ret, filename);
         if (ret == -EROFS) {
             ret = -EACCES;
         }
@@ -970,9 +969,8 @@ static int raw_reconfigure_getfd(BlockDriverState *bs, int flags,
         ret = raw_normalize_devicepath(&normalized_filename, errp);
         if (ret >= 0) {
             assert(!(*open_flags & O_CREAT));
-            fd = qemu_open(normalized_filename, *open_flags);
+            fd = qemu_open_err(normalized_filename, *open_flags, errp);
             if (fd == -1) {
-                error_setg_errno(errp, errno, "Could not reopen file");
                 return -1;
             }
         }
@@ -2324,10 +2322,10 @@ raw_co_create(BlockdevCreateOptions *options, Error **errp)
     }
 
     /* Create file */
-    fd = qemu_open(file_opts->filename, O_RDWR | O_CREAT | O_BINARY, 0644);
+    fd = qemu_open_err(file_opts->filename, O_RDWR | O_CREAT | O_BINARY,
+                       errp, 0644);
     if (fd < 0) {
         result = -errno;
-        error_setg_errno(errp, -result, "Could not create file");
         goto out;
     }
 
Signature · 1 line
2.26.2
#1 lore raw

PATCH 1/3 util: validate whether O_DIRECT is supported after failure

In reply to Daniel P. Berrangé

Currently we suggest that a filesystem may not support O_DIRECT after seeing an EINVAL. Other things can cause EINVAL though, so it is better to do an explicit check, and then report a definitive error message.

---
 util/osdep.c | 13 +++++++++++--
 1 file changed, 11 insertions(+), 2 deletions(-)
util/osdep.c +11 -2
index 4829c07ff6..4bdbe81cec 100644
@@ -342,8 +342,17 @@ int qemu_open(const char *name, int flags, ...)
 
 #ifdef O_DIRECT
     if (ret == -1 && errno == EINVAL && (flags & O_DIRECT)) {
-        error_report("file system may not support O_DIRECT");
-        errno = EINVAL; /* in case it was clobbered */
+        int newflags = flags & ~O_DIRECT;
+# ifdef O_CLOEXEC
+        ret = open(name, newflags | O_CLOEXEC, mode);
+# else
+        ret = open(name, newflags, mode);
+# endif
+        if (ret != -1) {
+            close(ret);
+            error_report("file system does not support O_DIRECT");
+            errno = EINVAL;
+        }
     }
 #endif /* O_DIRECT */
 
Signature · 1 line
2.26.2
#0 lore raw

cover block: improve error reporting for unsupported O_DIRECT

To repeat the commit message from patch 3...

Currently at startup if using cache=none on a filesystem lacking O_DIRECT such as tmpfs, at startup QEMU prints

qemu-system-x86_64: -drive file=/tmp/foo.img,cache=none: file system may not support O_DIRECT
qemu-system-x86_64: -drive file=/tmp/foo.img,cache=none: Could not open '/tmp/foo.img': Invalid argument

while at QMP level the hint is missing, so QEMU reports just

  "error": {
      "class": "GenericError",
      "desc": "Could not open '/tmp/foo.img': Invalid argument"
  }

which is close to useless for the end user trying to figure out what they did wrong

With this change at startup QEMU prints

qemu-system-x86_64: -drive file=/tmp/foo.img,cache=none: Unable to open '/tmp/foo.img' flags 0x4000: filesystem does not support O_DIRECT

while at the QMP level QEMU reports a massively more informative

  "error": {
     "class": "GenericError",
     "desc": "Unable to open '/tmp/foo.img' flags 0x4002: filesystem does not support O_DIRECT"
  }

qemu_open is used in many more places besides block layer, but converting those to qemu_open_err is left as an exercise for other maintainers.

Daniel P. Berrangé (3):
util: validate whether O_DIRECT is supported after failure
util: support detailed error reporting for qemu_open
block: switch to use qemu_open_err for improved errors

 block/file-posix.c   | 10 +++----
 include/qemu/osdep.h |  1 +
 util/osdep.c         | 70 ++++++++++++++++++++++++++++++++++++++------
 3 files changed, 66 insertions(+), 15 deletions(-)
Signature · 1 line
2.26.2
#3 lore raw

PATCH 2/3 util: support detailed error reporting for qemu_open

Create a "qemu_open_err" method which does the same as "qemu_open", but with a "Error **errp" for error reporting. There should be no behavioural difference for existing callers at this stage.

---
 include/qemu/osdep.h |  1 +
 util/osdep.c         | 71 +++++++++++++++++++++++++++++++++++---------
 2 files changed, 58 insertions(+), 14 deletions(-)
include/qemu/osdep.h +1 -0
index 0d26a1b9bd..e41701a308 100644
@@ -483,6 +483,7 @@ int qemu_madvise(void *addr, size_t len, int advice);
 int qemu_mprotect_rwx(void *addr, size_t size);
 int qemu_mprotect_none(void *addr, size_t size);
 
+int qemu_open_err(const char *name, int flags, Error **errp, ...);
 int qemu_open(const char *name, int flags, ...);
 int qemu_close(int fd);
 int qemu_unlink(const char *name);
util/osdep.c +57 -14

130 lines not on this page · raw message

Signature · 1 line
2.26.2
#7 lore raw
In reply to Daniel P. Berrangé
Patchew URL: https://patchew.org/QEMU/[email protected]/

Patchew URL: https://patchew.org/QEMU/[email protected]/

Hi,

This series failed the docker-quick@centos7 build test. Please find the testing commands and their output below. If you have Docker installed, you can probably reproduce it locally.

=== TEST SCRIPT BEGIN ===
#!/bin/bash
make docker-image-centos7 V=1 NETWORK=1
time make docker-test-quick@centos7 SHOW_ENV=1 J=14 NETWORK=1 === TEST SCRIPT END ===

+++ /tmp/qemu-test/build/tests/qemu-iotests/226.out.bad 2020-07-01 16:58:11.321029717 +0000 @@ -6,7 +6,7 @@
qemu-io: can't open: A regular file was expected by the 'file' driver, but something else was given
qemu-io: warning: Opening a character device as a file using the 'file' driver is deprecated == Testing RW ==
-qemu-io: can't open: Could not open 'TEST_DIR/t.IMGFMT': Is a directory
+qemu-io: can't open: Unable to open 'TEST_DIR/t.IMGFMT' flags 0x2: Is a directory
qemu-io: warning: Opening a character device as a file using the 'file' driver is deprecated

=== Testing with driver:host_device ===
@@ -14,13 +14,13 @@
---
Not run: 259
Failures: 061 069 111 226 244
Failed 5 of 119 iotests
make: *** [check-tests/check-block.sh] Error 1 Traceback (most recent call last):
File "./tests/docker/docker.py", line 669, in <module>

    sys.exit(main())

---
raise CalledProcessError(retcode, cmd)
subprocess.CalledProcessError: Command '['sudo', '-n', 'docker', 'run', '--label', 'com.qemu.instance.uuid=67228a70e7844d70b07a864fef5c4569', '-u', '1003', '--security-opt', 'seccomp=unconfined', '--rm', '-e', 'TARGET_LIST=', '-e', 'EXTRA_CONFIGURE_OPTS=', '-e', 'V=', '-e', 'J=14', '-e', 'DEBUG=', '-e', 'SHOW_ENV=1', '-e', 'CCACHE_DIR=/var/tmp/ccache', '-v', '/home/patchew2/.cache/qemu-docker-ccache:/var/tmp/ccache:z', '-v', '/var/tmp/patchew-tester-tmp-y2quzqfb/src/docker-src.2020-07-01-12.41.12.23538:/var/tmp/qemu:z,ro', 'qemu:centos7', '/var/tmp/qemu/run', 'test-quick']' returned non-zero exit status 2. filter=--filter=label=com.qemu.instance.uuid=67228a70e7844d70b07a864fef5c4569
make[1]: *** [docker-run] Error 1
make[1]: Leaving directory `/var/tmp/patchew-tester-tmp-y2quzqfb/src'
make: *** [docker-run-test-quick@centos7] Error 2

real 17m41.892s
user 0m5.790s

The full log is available at
http://patchew.org/logs/[email protected]/testing.docker-quick@centos7/?type=message.
---
Email generated automatically by Patchew [https://patchew.org/].
Please send your feedback to [email protected]

Reviewed-by Eric Blake on v1 2/3
#4 lore raw

Re: [PATCH 2/3] util: support detailed error reporting for qemu_open

In reply to Daniel P. Berrangé
Quoted 108 lines · Daniel P. Berrangé

Create a "qemu_open_err" method which does the same as "qemu_open", but with a "Error **errp" for error reporting. There should be no behavioural difference for existing callers at this stage.

---
 include/qemu/osdep.h |  1 +
 util/osdep.c         | 71 +++++++++++++++++++++++++++++++++++---------
 2 files changed, 58 insertions(+), 14 deletions(-)
include/qemu/osdep.h +1 -0
index 0d26a1b9bd..e41701a308 100644
@@ -483,6 +483,7 @@ int qemu_madvise(void *addr, size_t len, int advice);
 int qemu_mprotect_rwx(void *addr, size_t size);
 int qemu_mprotect_none(void *addr, size_t size);
 
+int qemu_open_err(const char *name, int flags, Error **errp, ...);
 int qemu_open(const char *name, int flags, ...);
 int qemu_close(int fd);
 int qemu_unlink(const char *name);
util/osdep.c +18 -9
index 4bdbe81cec..450b3a5da3 100644
@@ -22,6 +22,7 @@
  * THE SOFTWARE.
  */
 #include "qemu/osdep.h"
+#include "qapi/error.h"
 
 /* Needed early for CONFIG_BSD etc. */
 
@@ -282,7 +283,7 @@ int qemu_lock_fd_test(int fd, int64_t start, int64_t len, bool exclusive)
 /*
  * Opens a file with FD_CLOEXEC set
  */
-int qemu_open(const char *name, int flags, ...)
+static int qemu_openv(const char *name, int flags, Error **errp, va_list ap)
 {
     int ret;
     int mode = 0;
@@ -297,24 +298,31 @@ int qemu_open(const char *name, int flags, ...)
 
         fdset_id = qemu_parse_fdset(fdset_id_str);
         if (fdset_id == -1) {
+            error_setg(errp, "Unable to parse fdset %s", name);
             errno = EINVAL;
             return -1;
         }
 
         fd = monitor_fdset_get_fd(fdset_id, flags);
         if (fd < 0) {
+            error_setg_errno(errp, -fd, "Unable acquire FD for %s flags %x",
+                             name, flags);
             errno = -fd;
             return -1;
         }
 
         dupfd = qemu_dup_flags(fd, flags);
         if (dupfd == -1) {
+            error_setg_errno(errp, errno, "Unable dup FD for %s flags %x",
+                             name, flags);
             return -1;
         }
 
         ret = monitor_fdset_dup_fd_add(fdset_id, dupfd);
         if (ret == -1) {
             close(dupfd);
+            error_setg(errp, "Unable save FD for %s flags %x",
+                       name, flags);
             errno = EINVAL;
             return -1;
         }
@@ -324,11 +332,7 @@ int qemu_open(const char *name, int flags, ...)
 #endif
 
     if (flags & O_CREAT) {
-        va_list ap;
-
-        va_start(ap, flags);
         mode = va_arg(ap, int);
-        va_end(ap);
     }
 
 #ifdef O_CLOEXEC
@@ -340,25 +344,64 @@ int qemu_open(const char *name, int flags, ...)
     }
 #endif
 
+    if (ret == -1) {
+        const char *action = "open";
+        if (flags & O_CREAT) {
+            action = "create";
+        }
 #ifdef O_DIRECT
-    if (ret == -1 && errno == EINVAL && (flags & O_DIRECT)) {
-        int newflags = flags & ~O_DIRECT;
+        if (errno == EINVAL && (flags & O_DIRECT)) {
+            int newflags = flags & ~O_DIRECT;
 # ifdef O_CLOEXEC
-        ret = open(name, newflags | O_CLOEXEC, mode);
+            ret = open(name, newflags | O_CLOEXEC, mode);
 # else
-        ret = open(name, newflags, mode);
+            ret = open(name, newflags, mode);
 # endif

util/osdep.c int qemu_open(const char *name, int flags, ...) · line 365

Quoted 10 lines
util/osdep.c +6 -4
-        if (ret != -1) {
-            close(ret);
-            error_report("file system does not support O_DIRECT");
-            errno = EINVAL;
+            if (ret != -1) {
+                close(ret);
+                error_setg(errp, "Unable to %s '%s' flags 0x%x: "
+                           "filesystem does not support O_DIRECT",
+                           action, name, flags);
+                if (!errp) {

If the caller ignores errors, ...

util/osdep.c int qemu_open(const char *name, int flags, ...) · line 366

Quoted 1 line
util/osdep.c +1 -0
+                    error_report("file system does not support O_DIRECT");

... we report this error to stderr (but not any of the other ones). This is weird. I figure you do it here to reproduce the weirdness of qemu_open() before the patch. Goes back to

commit a5813077aac7865f69b7ee46a594f3705429f7cd
Author: Stefan Hajnoczi <[email protected]>
Date: Thu Aug 22 11:29:03 2013 +0200

osdep: warn if open(O_DIRECT) on fails with EINVAL

Print a warning when opening a file O_DIRECT fails with EINVAL. This saves users a lot of time trying to figure out the EINVAL error, which is typical when attempting to open a file O_DIRECT on Linux tmpfs.

The message isn't phrased as a warning, though. Should it use warn_report()? Stefan?

Regardless, since use of error_report() in a function that returns errors through its Error ** parameter is an anti-pattern, this use needs a comment. I'd make it a TODO comment: convert all callers where the warning is wanted to qemu_open_err(), then drop it.

Quoted 28 lines
util/osdep.c +21 -1
+                }
+                errno = EINVAL;
+                return -1;
+            }
         }
-    }
 #endif /* O_DIRECT */
+        error_setg_errno(errp, errno, "Unable to %s '%s' flags 0x%x",
+                         action, name, flags);
+    }
+
 
     return ret;
 }
 
+int qemu_open_err(const char *name, int flags, Error **errp, ...)
+{
+    va_list ap;
+    int rv;
+
+    va_start(ap, errp);
+    rv = qemu_openv(name, flags, errp, ap);
+    va_end(ap);
+
+    return rv;
+}
+
+int qemu_open(const char *name, int flags, ...)

util/osdep.c int qemu_open(const char *name, int flags, ...) · line 403

Quoted 10 lines
util/osdep.c +10 -0
+{
+    va_list ap;
+    int rv;
+
+    va_start(ap, flags);
+    rv = qemu_openv(name, flags, NULL, ap);
+    va_end(ap);
+
+    return rv;
+}

I'd rename this to qemu_open_with_bad_error_messages().

For better ones, callers can use

    if (qemu_open_err(name, flags, &err) < 0) {
        error_report_err(err);
        ...
    }

or, where the error is fatal

    qemu_open_err(name, flags, &error_fatal);

If you prefer not to rename it now, please add a comment why it should not be used in new code, and existing uses should be converted.

If you rename, call the new one qemu_open().

Quoted 4 lines
util/osdep.c +1 -0
+
 int qemu_close(int fd)
 {
     int64_t fdset_id;
#5 lore raw

Re: [PATCH 2/3] util: support detailed error reporting for qemu_open

Quoted 139 lines · Markus Armbruster
Quoted 108 lines · Daniel P. Berrangé

Create a "qemu_open_err" method which does the same as "qemu_open", but with a "Error **errp" for error reporting. There should be no behavioural difference for existing callers at this stage.

---
 include/qemu/osdep.h |  1 +
 util/osdep.c         | 71 +++++++++++++++++++++++++++++++++++---------
 2 files changed, 58 insertions(+), 14 deletions(-)
include/qemu/osdep.h +1 -0
index 0d26a1b9bd..e41701a308 100644
@@ -483,6 +483,7 @@ int qemu_madvise(void *addr, size_t len, int advice);
 int qemu_mprotect_rwx(void *addr, size_t size);
 int qemu_mprotect_none(void *addr, size_t size);
 
+int qemu_open_err(const char *name, int flags, Error **errp, ...);
 int qemu_open(const char *name, int flags, ...);
 int qemu_close(int fd);
 int qemu_unlink(const char *name);
util/osdep.c +18 -9
index 4bdbe81cec..450b3a5da3 100644
@@ -22,6 +22,7 @@
  * THE SOFTWARE.
  */
 #include "qemu/osdep.h"
+#include "qapi/error.h"
 
 /* Needed early for CONFIG_BSD etc. */
 
@@ -282,7 +283,7 @@ int qemu_lock_fd_test(int fd, int64_t start, int64_t len, bool exclusive)
 /*
  * Opens a file with FD_CLOEXEC set
  */
-int qemu_open(const char *name, int flags, ...)
+static int qemu_openv(const char *name, int flags, Error **errp, va_list ap)
 {
     int ret;
     int mode = 0;
@@ -297,24 +298,31 @@ int qemu_open(const char *name, int flags, ...)
 
         fdset_id = qemu_parse_fdset(fdset_id_str);
         if (fdset_id == -1) {
+            error_setg(errp, "Unable to parse fdset %s", name);
             errno = EINVAL;
             return -1;
         }
 
         fd = monitor_fdset_get_fd(fdset_id, flags);
         if (fd < 0) {
+            error_setg_errno(errp, -fd, "Unable acquire FD for %s flags %x",
+                             name, flags);
             errno = -fd;
             return -1;
         }
 
         dupfd = qemu_dup_flags(fd, flags);
         if (dupfd == -1) {
+            error_setg_errno(errp, errno, "Unable dup FD for %s flags %x",
+                             name, flags);
             return -1;
         }
 
         ret = monitor_fdset_dup_fd_add(fdset_id, dupfd);
         if (ret == -1) {
             close(dupfd);
+            error_setg(errp, "Unable save FD for %s flags %x",
+                       name, flags);
             errno = EINVAL;
             return -1;
         }
@@ -324,11 +332,7 @@ int qemu_open(const char *name, int flags, ...)
 #endif
 
     if (flags & O_CREAT) {
-        va_list ap;
-
-        va_start(ap, flags);
         mode = va_arg(ap, int);
-        va_end(ap);
     }
 
 #ifdef O_CLOEXEC
@@ -340,25 +344,64 @@ int qemu_open(const char *name, int flags, ...)
     }
 #endif
 
+    if (ret == -1) {
+        const char *action = "open";
+        if (flags & O_CREAT) {
+            action = "create";
+        }
 #ifdef O_DIRECT
-    if (ret == -1 && errno == EINVAL && (flags & O_DIRECT)) {
-        int newflags = flags & ~O_DIRECT;
+        if (errno == EINVAL && (flags & O_DIRECT)) {
+            int newflags = flags & ~O_DIRECT;
 # ifdef O_CLOEXEC
-        ret = open(name, newflags | O_CLOEXEC, mode);
+            ret = open(name, newflags | O_CLOEXEC, mode);
 # else
-        ret = open(name, newflags, mode);
+            ret = open(name, newflags, mode);
 # endif

util/osdep.c int qemu_open(const char *name, int flags, ...) · line 365

Quoted 10 lines
util/osdep.c +6 -4
-        if (ret != -1) {
-            close(ret);
-            error_report("file system does not support O_DIRECT");
-            errno = EINVAL;
+            if (ret != -1) {
+                close(ret);
+                error_setg(errp, "Unable to %s '%s' flags 0x%x: "
+                           "filesystem does not support O_DIRECT",
+                           action, name, flags);
+                if (!errp) {

If the caller ignores errors, ...

util/osdep.c int qemu_open(const char *name, int flags, ...) · line 366

Quoted 1 line
util/osdep.c +1 -0
+                    error_report("file system does not support O_DIRECT");

... we report this error to stderr (but not any of the other ones). This is weird. I figure you do it here to reproduce the weirdness of qemu_open() before the patch. Goes back to

commit a5813077aac7865f69b7ee46a594f3705429f7cd
Author: Stefan Hajnoczi <[email protected]>
Date: Thu Aug 22 11:29:03 2013 +0200

osdep: warn if open(O_DIRECT) on fails with EINVAL

Print a warning when opening a file O_DIRECT fails with EINVAL. This saves users a lot of time trying to figure out the EINVAL error, which is typical when attempting to open a file O_DIRECT on Linux tmpfs.

Quoted 6 lines

The message isn't phrased as a warning, though. Should it use warn_report()? Stefan?

I was really in two minds as to whether to keep this error_report or not. It depends depends on whether any other callers of qemu_open still need it. In fact if nothing outside the block layer uses O_DIRECT then I think we can remove the error_report line.

Quoted 17 lines
Quoted 17 lines
util/osdep.c +17 -0
+int qemu_open_err(const char *name, int flags, Error **errp, ...)
+{
+    va_list ap;
+    int rv;
+
+    va_start(ap, errp);
+    rv = qemu_openv(name, flags, errp, ap);
+    va_end(ap);
+
+    return rv;
+}
+
+int qemu_open(const char *name, int flags, ...)
+{
+    va_list ap;
+    int rv;
+
Quoted 8 lines

util/osdep.c int qemu_open(const char *name, int flags, ...) · line 403

Quoted 6 lines
util/osdep.c +6 -0
+    va_start(ap, flags);
+    rv = qemu_openv(name, flags, NULL, ap);
+    va_end(ap);
+
+    return rv;
+}

I'd rename this to qemu_open_with_bad_error_messages().

My goal was to avoid a big bang conversion of all callers.

Quoted 8 lines

For better ones, callers can use

    if (qemu_open_err(name, flags, &err) < 0) {
        error_report_err(err);
        ...
    }
Quoted 8 lines

or, where the error is fatal

    qemu_open_err(name, flags, &error_fatal);

If you prefer not to rename it now, please add a comment why it should not be used in new code, and existing uses should be converted.

If you rename, call the new one qemu_open().

Maybe it is better to rename upfront though, instead of waiting till everything uses the qemu_open_err and then renaming back to qemu_open.

Regards,
Daniel

Signature · 3 lines
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|
Reviewed-by on v1 1/3
#2 lore raw

Re: [PATCH 1/3] util: validate whether O_DIRECT is supported after failure

In reply to Daniel P. Berrangé
Quoted 15 lines · Daniel P. Berrangé

Currently we suggest that a filesystem may not support O_DIRECT after seeing an EINVAL. Other things can cause EINVAL though, so it is better to do an explicit check, and then report a definitive error message.

---
 util/osdep.c | 13 +++++++++++--
 1 file changed, 11 insertions(+), 2 deletions(-)
util/osdep.c +0 -0
index 4829c07ff6..4bdbe81cec 100644
@@ -342,8 +342,17 @@ int qemu_open(const char *name, int flags, ...)
 

util/osdep.c int qemu_open(const char *name, int flags, ...) · line 350

Quoted 10 lines
util/osdep.c +6 -2
 #ifdef O_DIRECT
     if (ret == -1 && errno == EINVAL && (flags & O_DIRECT)) {
-        error_report("file system may not support O_DIRECT");
-        errno = EINVAL; /* in case it was clobbered */
+        int newflags = flags & ~O_DIRECT;
+# ifdef O_CLOEXEC
+        ret = open(name, newflags | O_CLOEXEC, mode);
+# else
+        ret = open(name, newflags, mode);
+# endif

Can we use this simpler form (easier to set a breakpoint)?

   #ifdef O_CLOEXEC
           newflags |= O_CLOEXEC;
   #endif
           ret = open(name, newflags, mode);

Whichever you prefer:

Quoted 9 lines
util/osdep.c +5 -0
+        if (ret != -1) {
+            close(ret);
+            error_report("file system does not support O_DIRECT");
+            errno = EINVAL;
+        }
     }
 #endif /* O_DIRECT */