[PATCH] statfs.2 does not describe f_flag field

[PATCH] statfs.2 does not describe f_flag field

[Jan Chaloupka]

Based on kernel commit [1], new flag f_flag is added to statfs struct.

Before kernel 2.6.36, glibc's statvfs.f_flags was computed be iterating
through all mounted points. This was inefficient and could cause a delay.
Since kernel 2.6.36, statfs struct has new field f_flag, which already
extracts mount options without iterating through mount point.
statfs.f_flag flags of glibc and kernel are the same (plus kernel adds
additional flag ST_VALID used only internally). So no matter what version
of kernel is used, flags does not change.

Flags used in statfs.f_flags can be enumerated from upstreams's
glibc/internal_statvfs.c, lines 155-173. Nine flags in sum:

ST_RDONLY = 1
ST_NOSUID = 2
ST_NODEV = 4
ST_NOEXEC = 8
ST_SYNCHRONOUS = 16
ST_MANDLOCK = 64
ST_NOATIME = 1024
ST_NODIRATIME = 2048
ST_RELATIME = 4096

This patch describes missing flag.

[1] https://github.com/torvalds/linux/commit/365b18189789bfa1acd9939e6312b8a4b4577b28

Signed-off-by: Jan Chaloupka <jchaloup-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
---
 man2/statfs.2 |   28 +++++++++++++++++++++++++++-
 1 file changed, 27 insertions(+), 1 deletion(-)

diff --git a/man2/statfs.2 b/man2/statfs.2
index 714bc7c..46cfa5f 100644
--- a/man2/statfs.2
+++ b/man2/statfs.2
@@ -65,7 +65,9 @@ struct statfs {
     fsid_t       f_fsid;    /* filesystem id */
     __SWORD_TYPE f_namelen; /* maximum length of filenames */
     __SWORD_TYPE f_frsize;  /* fragment size (since Linux 2.6) */
-    __SWORD_TYPE f_spare[5];
+    __SWORD_TYPE f_flags;   /* mount flags of filesystem
+                               (since Linux 2.6.36) */
+    __SWORD_TYPE f_spare[4];
 };
 
 Filesystem types:
@@ -146,6 +148,30 @@ Most of these MAGIC constants are defined in
 .IR /usr/include/linux/magic.h ,
 and some are hardcoded in kernel sources.
 .PP
+Mount flags of
+.I f_flags
+field are extracted from an option string of a mount point defined in
+.BR mount (8).
+The field can be combination of the following flags
+(corresponding mount options are in the right column):
+.in
+.nf
+
+   ST_RDONLY          ro
+   ST_NOSUID          nosuid
+   ST_NOEXEC          noexec
+   ST_NODEV           nodev
+   ST_SYNCHRONOUS     sync
+   ST_MANDLOCK        mand
+   ST_NOATIME         noatime
+   ST_NODIRATIME      nodiratime
+   ST_RELATIME        relatime
+
+.fi
+.in
+For kernel less than 2.6.36 option flags are extracted by iterating
+through individual mount points which can be inefficient and cause delay.
+.PP
 Nobody knows what
 .I f_fsid
 is supposed to contain (but see below).

--
To unsubscribe from this list: send the line "unsubscribe linux-man" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] statfs.2 does not describe f_flag field

[Michael Kerrisk (man-pages)]

Hi Jan,

On 4 February 2015 at 09:48, Jan Chaloupka <jchaloup-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org> wrote:
> Based on kernel commit [1], new flag f_flag is added to statfs struct.
>
> Before kernel 2.6.36, glibc's statvfs.f_flags was computed be iterating
> through all mounted points. This was inefficient and could cause a delay.
> Since kernel 2.6.36, statfs struct has new field f_flag, which already
> extracts mount options without iterating through mount point.
> statfs.f_flag flags of glibc and kernel are the same (plus kernel adds
> additional flag ST_VALID used only internally). So no matter what version
> of kernel is used, flags does not change.
>
> Flags used in statfs.f_flags can be enumerated from upstreams's
> glibc/internal_statvfs.c, lines 155-173. Nine flags in sum:
>
> ST_RDONLY = 1
> ST_NOSUID = 2
> ST_NODEV = 4
> ST_NOEXEC = 8
> ST_SYNCHRONOUS = 16
> ST_MANDLOCK = 64
> ST_NOATIME = 1024
> ST_NODIRATIME = 2048
> ST_RELATIME = 4096
>
> This patch describes missing flag.
>
> [1] https://github.com/torvalds/linux/commit/365b18189789bfa1acd9939e6312b8a4b4577b28

Thanks for the heads up. But this patch is a bit muddled, since it
mixes details about statfs(2) and statvfs(3). I've instead made some
more detailed changes for the statfs(2) and statvfs(2) pages.

Thanks,

Michael


> Signed-off-by: Jan Chaloupka <jchaloup-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
> ---
>  man2/statfs.2 |   28 +++++++++++++++++++++++++++-
>  1 file changed, 27 insertions(+), 1 deletion(-)
>
> diff --git a/man2/statfs.2 b/man2/statfs.2
> index 714bc7c..46cfa5f 100644
> --- a/man2/statfs.2
> +++ b/man2/statfs.2
> @@ -65,7 +65,9 @@ struct statfs {
>      fsid_t       f_fsid;    /* filesystem id */
>      __SWORD_TYPE f_namelen; /* maximum length of filenames */
>      __SWORD_TYPE f_frsize;  /* fragment size (since Linux 2.6) */
> -    __SWORD_TYPE f_spare[5];
> +    __SWORD_TYPE f_flags;   /* mount flags of filesystem
> +                               (since Linux 2.6.36) */
> +    __SWORD_TYPE f_spare[4];
>  };
>
>  Filesystem types:
> @@ -146,6 +148,30 @@ Most of these MAGIC constants are defined in
>  .IR /usr/include/linux/magic.h ,
>  and some are hardcoded in kernel sources.
>  .PP
> +Mount flags of
> +.I f_flags
> +field are extracted from an option string of a mount point defined in
> +.BR mount (8).
> +The field can be combination of the following flags
> +(corresponding mount options are in the right column):
> +.in
> +.nf
> +
> +   ST_RDONLY          ro
> +   ST_NOSUID          nosuid
> +   ST_NOEXEC          noexec
> +   ST_NODEV           nodev
> +   ST_SYNCHRONOUS     sync
> +   ST_MANDLOCK        mand
> +   ST_NOATIME         noatime
> +   ST_NODIRATIME      nodiratime
> +   ST_RELATIME        relatime
> +
> +.fi
> +.in
> +For kernel less than 2.6.36 option flags are extracted by iterating
> +through individual mount points which can be inefficient and cause delay.
> +.PP
>  Nobody knows what
>  .I f_fsid
>  is supposed to contain (but see below).
>



-- 
Michael Kerrisk
Linux man-pages maintainer; http://www.kernel.org/doc/man-pages/
Linux/UNIX System Programming Training: http://man7.org/training/
--
To unsubscribe from this list: send the line "unsubscribe linux-man" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] statfs.2 does not describe f_flag field

[Jan Chaloupka]

On 02/04/2015 12:51 PM, Michael Kerrisk (man-pages) wrote:
> Hi Jan,
>
> On 4 February 2015 at 09:48, Jan Chaloupka <jchaloup-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org> wrote:
>> Based on kernel commit [1], new flag f_flag is added to statfs struct.
>>
>> Before kernel 2.6.36, glibc's statvfs.f_flags was computed be iterating
>> through all mounted points. This was inefficient and could cause a delay.
>> Since kernel 2.6.36, statfs struct has new field f_flag, which already
>> extracts mount options without iterating through mount point.
>> statfs.f_flag flags of glibc and kernel are the same (plus kernel adds
>> additional flag ST_VALID used only internally). So no matter what version
>> of kernel is used, flags does not change.
>>
>> Flags used in statfs.f_flags can be enumerated from upstreams's
>> glibc/internal_statvfs.c, lines 155-173. Nine flags in sum:
>>
>> ST_RDONLY = 1
>> ST_NOSUID = 2
>> ST_NODEV = 4
>> ST_NOEXEC = 8
>> ST_SYNCHRONOUS = 16
>> ST_MANDLOCK = 64
>> ST_NOATIME = 1024
>> ST_NODIRATIME = 2048
>> ST_RELATIME = 4096
>>
>> This patch describes missing flag.
>>
>> [1] https://github.com/torvalds/linux/commit/365b18189789bfa1acd9939e6312b8a4b4577b28
> Thanks for the heads up. But this patch is a bit muddled, since it
> mixes details about statfs(2) and statvfs(3).

Yes. As kernel copied all the flags, it is now duplicated. But you are 
right it more suites for statvfs(2).

>   I've instead made some more detailed changes for the statfs(2) and statvfs(2) pages.
>

Thanks.

>
> Thanks,
>
> Michael
>
>
>> Signed-off-by: Jan Chaloupka <jchaloup-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
>> ---
>>   man2/statfs.2 |   28 +++++++++++++++++++++++++++-
>>   1 file changed, 27 insertions(+), 1 deletion(-)
>>
>> diff --git a/man2/statfs.2 b/man2/statfs.2
>> index 714bc7c..46cfa5f 100644
>> --- a/man2/statfs.2
>> +++ b/man2/statfs.2
>> @@ -65,7 +65,9 @@ struct statfs {
>>       fsid_t       f_fsid;    /* filesystem id */
>>       __SWORD_TYPE f_namelen; /* maximum length of filenames */
>>       __SWORD_TYPE f_frsize;  /* fragment size (since Linux 2.6) */
>> -    __SWORD_TYPE f_spare[5];
>> +    __SWORD_TYPE f_flags;   /* mount flags of filesystem
>> +                               (since Linux 2.6.36) */
>> +    __SWORD_TYPE f_spare[4];
>>   };
>>
>>   Filesystem types:
>> @@ -146,6 +148,30 @@ Most of these MAGIC constants are defined in
>>   .IR /usr/include/linux/magic.h ,
>>   and some are hardcoded in kernel sources.
>>   .PP
>> +Mount flags of
>> +.I f_flags
>> +field are extracted from an option string of a mount point defined in
>> +.BR mount (8).
>> +The field can be combination of the following flags
>> +(corresponding mount options are in the right column):
>> +.in
>> +.nf
>> +
>> +   ST_RDONLY          ro
>> +   ST_NOSUID          nosuid
>> +   ST_NOEXEC          noexec
>> +   ST_NODEV           nodev
>> +   ST_SYNCHRONOUS     sync
>> +   ST_MANDLOCK        mand
>> +   ST_NOATIME         noatime
>> +   ST_NODIRATIME      nodiratime
>> +   ST_RELATIME        relatime
>> +
>> +.fi
>> +.in
>> +For kernel less than 2.6.36 option flags are extracted by iterating
>> +through individual mount points which can be inefficient and cause delay.
>> +.PP
>>   Nobody knows what
>>   .I f_fsid
>>   is supposed to contain (but see below).
>>
>
>
--
To unsubscribe from this list: send the line "unsubscribe linux-man" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html