Fix MS_LAZYTIME not defined on uclibc

[giuliobenetti]

No description provided.

[Nikratio]

Where does this value come from? (please document in the comment)

[giuliobenetti]

Hi @Nikratio,

it comes from Linux header mount.h:
https://elixir.bootlin.com/linux/v6.2.8/source/include/uapi/linux/mount.h#L39
and it never changed value throug various versions since it's been added with Linux 4.0:
https://elixir.bootlin.com/linux/v4.0.9/source/include/uapi/linux/fs.h#L93

Unforunately uclibc doesn't add it to <sys/mount.h>. The other chance to fix the bug is to include Linux's:
include/uapi/linux/mount.h
but before Linux 5.0 that define is contained in:
https://elixir.bootlin.com/linux/v4.20.17/source/include/uapi/linux/fs.h#L133
So I don't know if that's a good idea. Or if we do that I can check against Linux version and include the 2 different headers.

[Nikratio]

There's nothing wrong with hardcoding the value, I just want you to add the source to the (C) comment in the file :-)

[giuliobenetti]

You're right, I don't know why I've left a commit so bad written :-/

[bsbernd]

This patch would miss util/fusermount.c.

Can you try out this?

diff --git a/lib/mount.c b/lib/mount.c
index 1f1ee860..b7965d6c 100644
--- a/lib/mount.c
+++ b/lib/mount.c
@@ -27,6 +27,9 @@
 #include <sys/wait.h>
 #include <sys/mount.h>
 
+#include <linux/fs.h>
+#include <linux/mount.h>
+
 #ifdef __NetBSD__
 #include <perfuse.h>
 
@@ -47,9 +50,11 @@
 #define fork() vfork()
 #endif
 
+#if 0
 #ifndef MS_DIRSYNC
 #define MS_DIRSYNC 128
 #endif
+#endif
 
 enum {
        KEY_KERN_FLAG,
diff --git a/util/fusermount.c b/util/fusermount.c
index 6e72f0d3..41327e58 100644
--- a/util/fusermount.c
+++ b/util/fusermount.c
@@ -32,13 +32,18 @@
 #include <stdbool.h>
 #include <sys/vfs.h>
 
+#include <linux/fs.h>
+#include <linux/mount.h>
+
 #define FUSE_COMMFD_ENV                "_FUSE_COMMFD"
 
 #define FUSE_DEV "/dev/fuse"
 
+#if 0
 #ifndef MS_DIRSYNC
 #define MS_DIRSYNC 128
 #endif
+
 #ifndef MS_REC
 #define MS_REC 16384
 #endif
@@ -46,6 +51,8 @@
 #define MS_PRIVATE (1<<18)
 #endif
 
+#endif
+
 #ifndef UMOUNT_DETACH
 #define UMOUNT_DETACH  0x00000002      /* Just detach from the tree */
 #endif

I don't see a problem to include multiple header files - nothing bad should happen, except there would be a bug in the headers.

And I have the feeling we need a test for ulibc compilation.

[giuliobenetti]

This patch would miss util/fusermount.c.

Can you try out this?

diff --git a/lib/mount.c b/lib/mount.c
index 1f1ee860..b7965d6c 100644
--- a/lib/mount.c
+++ b/lib/mount.c
@@ -27,6 +27,9 @@
 #include <sys/wait.h>
 #include <sys/mount.h>
 
+#include <linux/fs.h>
+#include <linux/mount.h>
+
 #ifdef __NetBSD__
 #include <perfuse.h>
 
@@ -47,9 +50,11 @@
 #define fork() vfork()
 #endif
 
+#if 0
 #ifndef MS_DIRSYNC
 #define MS_DIRSYNC 128
 #endif
+#endif
 
 enum {
        KEY_KERN_FLAG,
diff --git a/util/fusermount.c b/util/fusermount.c
index 6e72f0d3..41327e58 100644
--- a/util/fusermount.c
+++ b/util/fusermount.c
@@ -32,13 +32,18 @@
 #include <stdbool.h>
 #include <sys/vfs.h>
 
+#include <linux/fs.h>
+#include <linux/mount.h>
+
 #define FUSE_COMMFD_ENV                "_FUSE_COMMFD"
 
 #define FUSE_DEV "/dev/fuse"
 
+#if 0
 #ifndef MS_DIRSYNC
 #define MS_DIRSYNC 128
 #endif
+
 #ifndef MS_REC
 #define MS_REC 16384
 #endif
@@ -46,6 +51,8 @@
 #define MS_PRIVATE (1<<18)
 #endif
 
+#endif
+
 #ifndef UMOUNT_DETACH
 #define UMOUNT_DETACH  0x00000002      /* Just detach from the tree */
 #endif

I don't see a problem to include multiple header files - nothing bad should happen, except there would be a bug in the headers.

Unfortunately <linux/mount.h> is not always present because we don't always have Linux Headers installed:

/home/giuliobenetti/git/upstream/test-libfuse3/bootlin-armv5-uclibc/host/bin/arm-linux-gcc -Iutil/fusermount3.p -Iutil -I../util -Iinclude -I../include -Ilib -I../lib -I. -I.. -fdiagnostics-color=always -Wall -Winvalid-pch -Wextra -O3 -D_REENTRANT -DHAVE_LIBFUSE_PRIVATE_CONFIG_H -Wno-sign-compare -Wstrict-prototypes -Wmissing-declarations -Wwrite-strings -fno-strict-aliasing -Wno-unused-result -DHAVE_SYMVER_ATTRIBUTE -D_LARGEFILE_SOURCE -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -Os -g0 '-DFUSE_CONF="/etc/fuse.conf"' -MD -MQ util/fusermount3.p/fusermount.c.o -MF util/fusermount3.p/fusermount.c.o.d -o util/fusermount3.p/fusermount.c.o -c ../util/fusermount.c
../util/fusermount.c:36:10: fatal error: linux/mount.h: No such file or directory
   36 | #include <linux/mount.h>
      |          ^~~~~~~~~~~~~~~
compilation terminated.

I understand your goal to use includes instead of local redefinition of macros, but while dealing with kernel includes
can change between versions the same way as defines, so I don't what's better :-/
IMHO I'd leave it as it is.
Patch(without the comment) for libfuse3.14.1 is already pending on Buildroot's patchwork:
https://patchwork.ozlabs.org/project/buildroot/patch/20230327192628.566995-1-giulio.benetti@benettiengineering.com/

And I have the feeling we need a test for ulibc compilation.

It would be great!

[bsbernd]

Ok, yeah, if linux/mount.h is not always available it makes it complicated.

[bsbernd]

Also needs to be defined in util/fusermount.c. Maybe we should add a mount_compat.h?

[giuliobenetti]

Also needs to be defined in util/fusermount.c. Maybe we should add a mount_compat.h?

Good idea, indeed I was looking exactly for something like that to touch only one file instead of 2.
I didn't know how to call that file and yes, mount_compat.h is the name :-)

[Nikratio]

Tagging this as "draft" per the comments above. Please let me know when this is ready for review.

[giuliobenetti]

Tagging this as "draft" per the comments above. Please let me know when this is ready for review.

I've updated the patch. Hope it looks good to you. I start the building with Buildroot, so please wait because I still have to try to build!

[giuliobenetti]

@Nikratio it built successfully all 44 combinations in Buildroot.
Best regards