* bug#68504: [PATCH] Add copy-on-write support to scm_copy_file.
2024-01-16 12:48 bug#68504: [PATCH] Add copy-on-write support to scm_copy_file Tomas Volf
@ 2024-01-24 10:26 ` Ludovic Courtès
2024-01-24 19:19 ` Tomas Volf
2024-01-24 19:10 ` bug#68504: [PATCH v2] " Tomas Volf
2024-01-24 19:14 ` bug#68504: [PATCH v3] " Tomas Volf
2 siblings, 1 reply; 8+ messages in thread
From: Ludovic Courtès @ 2024-01-24 10:26 UTC (permalink / raw)
To: Tomas Volf; +Cc: 68504
Hi,
Tomas Volf <~@wolfsden.cz> skribis:
> On modern file-systems (BTRFS, ZFS) it is possible to copy a file using
> copy-on-write method. For large files it has the advantage of being
> much faster and saving disk space (since identical extents are not
> duplicated). This feature is stable and for example coreutils' `cp'
> does use it automatically (see --reflink).
>
> This commit adds support for this feature into our
> copy-file (scm_copy_file) procedure. Same as `cp', it defaults to
> 'auto, meaning the copy-on-write is attempted, and in case of failure
> the regular copy is performed.
>
> No tests are provided, because the behavior depends on the system,
> underlying file-system and its configuration. That makes it challenging
> to write a test for it. Manual testing was performed instead:
>
> $ btrfs filesystem du /tmp/cow*
> Total Exclusive Set shared Filename
> 36.00KiB 36.00KiB 0.00B /tmp/cow
>
> $ cat cow-test.scm
> (copy-file "/tmp/cow" "/tmp/cow-unspecified")
> (copy-file "/tmp/cow" "/tmp/cow-always" #:copy-on-write 'always)
> (copy-file "/tmp/cow" "/tmp/cow-auto" #:copy-on-write 'auto)
> (copy-file "/tmp/cow" "/tmp/cow-never" #:copy-on-write 'never)
> (copy-file "/tmp/cow" "/dev/shm/cow-unspecified")
> (copy-file "/tmp/cow" "/dev/shm/cow-auto" #:copy-on-write 'auto)
> (copy-file "/tmp/cow" "/dev/shm/cow-never" #:copy-on-write 'never)
> $ ./meta/guile -s cow-test.scm
>
> $ btrfs filesystem du /tmp/cow*
> Total Exclusive Set shared Filename
> 36.00KiB 0.00B 36.00KiB /tmp/cow
> 36.00KiB 0.00B 36.00KiB /tmp/cow-always
> 36.00KiB 0.00B 36.00KiB /tmp/cow-auto
> 36.00KiB 36.00KiB 0.00B /tmp/cow-never
> 36.00KiB 0.00B 36.00KiB /tmp/cow-unspecified
>
> $ sha1sum /tmp/cow* /dev/shm/cow*
> 4c665f87b5dc2e7d26279c4b48968d085e1ace32 /tmp/cow
> 4c665f87b5dc2e7d26279c4b48968d085e1ace32 /tmp/cow-always
> 4c665f87b5dc2e7d26279c4b48968d085e1ace32 /tmp/cow-auto
> 4c665f87b5dc2e7d26279c4b48968d085e1ace32 /tmp/cow-never
> 4c665f87b5dc2e7d26279c4b48968d085e1ace32 /tmp/cow-unspecified
> 4c665f87b5dc2e7d26279c4b48968d085e1ace32 /dev/shm/cow-auto
> 4c665f87b5dc2e7d26279c4b48968d085e1ace32 /dev/shm/cow-never
> 4c665f87b5dc2e7d26279c4b48968d085e1ace32 /dev/shm/cow-unspecified
>
> This commit also adds to new failure modes for (copy-file).
>
> Failure to copy-on-write when 'always was passed in:
>
> scheme@(guile-user)> (copy-file "/tmp/cow" "/dev/shm/cow" #:copy-on-write 'always)
> ice-9/boot-9.scm:1676:22: In procedure raise-exception:
> In procedure copy-file: copy-on-write failed: Invalid cross-device link
>
> Passing in invalid value for the #:copy-on-write keyword argument:
>
> scheme@(guile-user)> (copy-file "/tmp/cow" "/dev/shm/cow" #:copy-on-write 'nevr)
> ice-9/boot-9.scm:1676:22: In procedure raise-exception:
> In procedure copy-file: invalid value for #:copy-on-write: nevr
>
> * NEWS: Add note for copy-file supporting copy-on-write.
> * configure.ac: Check for linux/fs.h.
> * doc/ref/posix.texi (File System)[copy-file]: Document the new
> signature.
> * libguile/filesys.c (clone_file): New function cloning a file using
> FICLONE, if supported.
> (k_copy_on_write): New keyword.
> (sym_always, sym_auto, sym_never): New symbols.
> (scm_copy_file): New #:copy-on-write keyword argument. Attempt
> copy-on-write copy by default.
> * libguile/filesys.h: Update signature for scm_copy_file.
The patch looks great (and very useful) to me, modulo one issue:
> -SCM_API SCM scm_copy_file (SCM oldfile, SCM newfile);
> +SCM_API SCM scm_copy_file (SCM oldfile, SCM newfile, SCM rest);
Since this is a public interface, we cannot change this function’s
signature during the 3.0 stable series.
Thus, I would suggest keeping the public ‘scm_copy_file’ unchanged and
internally having a three-argument variant. The Scheme-level
‘copy-file’ would map to that three-argument variant. (See how
‘scm_pipe’ and ‘scm accept’ as examples.)
Could you send an updated patch?
BTW, copyright assignment to the FSF is now optional but encouraged.
Please see
<https://lists.gnu.org/archive/html/guile-devel/2022-10/msg00008.html>.
Thanks,
Ludo’.
^ permalink raw reply [flat|nested] 8+ messages in thread
* bug#68504: [PATCH] Add copy-on-write support to scm_copy_file.
2024-01-24 10:26 ` Ludovic Courtès
@ 2024-01-24 19:19 ` Tomas Volf
0 siblings, 0 replies; 8+ messages in thread
From: Tomas Volf @ 2024-01-24 19:19 UTC (permalink / raw)
To: Ludovic Courtès; +Cc: 68504
[-- Attachment #1: Type: text/plain, Size: 1684 bytes --]
On 2024-01-24 11:26:56 +0100, Ludovic Courtès wrote:
>
> The patch looks great (and very useful) to me, modulo one issue:
>
> > -SCM_API SCM scm_copy_file (SCM oldfile, SCM newfile);
> > +SCM_API SCM scm_copy_file (SCM oldfile, SCM newfile, SCM rest);
>
> Since this is a public interface, we cannot change this function’s
> signature during the 3.0 stable series.
>
> Thus, I would suggest keeping the public ‘scm_copy_file’ unchanged and
> internally having a three-argument variant. The Scheme-level
> ‘copy-file’ would map to that three-argument variant. (See how
> ‘scm_pipe’ and ‘scm accept’ as examples.)
That is a very good point, which I did not realize at all. Thanks to the
examples you provided, it was not that hard to do (well, assuming I did it
right).
> Could you send an updated patch?
Done. However now that I read it after myself, I overlooked this occurrence of
scm_copy_file in the commit message:
This commit adds support for this feature into our
copy-file (scm_copy_file) procedure. Same as `cp', it defaults to
So I just sent v3 right after v2, sorry for the noise, should have been more
careful.
>
> BTW, copyright assignment to the FSF is now optional but encouraged.
> Please see
> <https://lists.gnu.org/archive/html/guile-devel/2022-10/msg00008.html>.
Since it is optional, I will currently opt into not doing the assignment, I do
not like the concept that much. I will try to find time to actually form an
opinion based on facts.
Have a nice day,
Tomas
--
There are only two hard things in Computer Science:
cache invalidation, naming things and off-by-one errors.
[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]
^ permalink raw reply [flat|nested] 8+ messages in thread
* bug#68504: [PATCH v2] Add copy-on-write support to scm_copy_file.
2024-01-16 12:48 bug#68504: [PATCH] Add copy-on-write support to scm_copy_file Tomas Volf
2024-01-24 10:26 ` Ludovic Courtès
@ 2024-01-24 19:10 ` Tomas Volf
2024-01-24 19:14 ` bug#68504: [PATCH v3] " Tomas Volf
2 siblings, 0 replies; 8+ messages in thread
From: Tomas Volf @ 2024-01-24 19:10 UTC (permalink / raw)
To: 68504; +Cc: Tomas Volf
On modern file-systems (BTRFS, ZFS) it is possible to copy a file using
copy-on-write method. For large files it has the advantage of being
much faster and saving disk space (since identical extents are not
duplicated). This feature is stable and for example coreutils' `cp'
does use it automatically (see --reflink).
This commit adds support for this feature into our
copy-file (scm_copy_file) procedure. Same as `cp', it defaults to
'auto, meaning the copy-on-write is attempted, and in case of failure
the regular copy is performed.
No tests are provided, because the behavior depends on the system,
underlying file-system and its configuration. That makes it challenging
to write a test for it. Manual testing was performed instead:
$ btrfs filesystem du /tmp/cow*
Total Exclusive Set shared Filename
36.00KiB 36.00KiB 0.00B /tmp/cow
$ cat cow-test.scm
(copy-file "/tmp/cow" "/tmp/cow-unspecified")
(copy-file "/tmp/cow" "/tmp/cow-always" #:copy-on-write 'always)
(copy-file "/tmp/cow" "/tmp/cow-auto" #:copy-on-write 'auto)
(copy-file "/tmp/cow" "/tmp/cow-never" #:copy-on-write 'never)
(copy-file "/tmp/cow" "/dev/shm/cow-unspecified")
(copy-file "/tmp/cow" "/dev/shm/cow-auto" #:copy-on-write 'auto)
(copy-file "/tmp/cow" "/dev/shm/cow-never" #:copy-on-write 'never)
$ ./meta/guile -s cow-test.scm
$ btrfs filesystem du /tmp/cow*
Total Exclusive Set shared Filename
36.00KiB 0.00B 36.00KiB /tmp/cow
36.00KiB 0.00B 36.00KiB /tmp/cow-always
36.00KiB 0.00B 36.00KiB /tmp/cow-auto
36.00KiB 36.00KiB 0.00B /tmp/cow-never
36.00KiB 0.00B 36.00KiB /tmp/cow-unspecified
$ sha1sum /tmp/cow* /dev/shm/cow*
4c665f87b5dc2e7d26279c4b48968d085e1ace32 /tmp/cow
4c665f87b5dc2e7d26279c4b48968d085e1ace32 /tmp/cow-always
4c665f87b5dc2e7d26279c4b48968d085e1ace32 /tmp/cow-auto
4c665f87b5dc2e7d26279c4b48968d085e1ace32 /tmp/cow-never
4c665f87b5dc2e7d26279c4b48968d085e1ace32 /tmp/cow-unspecified
4c665f87b5dc2e7d26279c4b48968d085e1ace32 /dev/shm/cow-auto
4c665f87b5dc2e7d26279c4b48968d085e1ace32 /dev/shm/cow-never
4c665f87b5dc2e7d26279c4b48968d085e1ace32 /dev/shm/cow-unspecified
This commit also adds to new failure modes for (copy-file).
Failure to copy-on-write when 'always was passed in:
scheme@(guile-user)> (copy-file "/tmp/cow" "/dev/shm/cow" #:copy-on-write 'always)
ice-9/boot-9.scm:1676:22: In procedure raise-exception:
In procedure copy-file: copy-on-write failed: Invalid cross-device link
Passing in invalid value for the #:copy-on-write keyword argument:
scheme@(guile-user)> (copy-file "/tmp/cow" "/dev/shm/cow" #:copy-on-write 'nevr)
ice-9/boot-9.scm:1676:22: In procedure raise-exception:
In procedure copy-file: invalid value for #:copy-on-write: nevr
* NEWS: Add note for copy-file supporting copy-on-write.
* configure.ac: Check for linux/fs.h.
* doc/ref/posix.texi (File System)[copy-file]: Document the new
signature.
* libguile/filesys.c (clone_file): New function cloning a file using
FICLONE, if supported.
(k_copy_on_write): New keyword.
(sym_always, sym_auto, sym_never): New symbols.
(scm_copy_file2): Renamed from scm_copy_file. New #:copy-on-write
keyword argument. Attempt copy-on-write copy by default.
(scm_copy_file): Call scm_copy_file2.
* libguile/filesys.h: Add scm_copy_file2 as SCM_INTERNAL.
---
v2: Introduce scm_copy_file2 in order to preserve backwards compatibility.
NEWS | 9 +++++
configure.ac | 1 +
doc/ref/posix.texi | 9 ++++-
libguile/filesys.c | 82 +++++++++++++++++++++++++++++++++++++++-------
libguile/filesys.h | 1 +
5 files changed, 89 insertions(+), 13 deletions(-)
diff --git a/NEWS b/NEWS
index b319404d7..9147098c9 100644
--- a/NEWS
+++ b/NEWS
@@ -21,6 +21,15 @@ definitely unused---this is notably the case for modules that are only
used at macro-expansion time, such as (srfi srfi-26). In those cases,
the compiler reports it as "possibly unused".
+** copy-file now supports copy-on-write
+
+The copy-file procedure now takes an additional keyword argument,
+#:copy-on-write, specifying whether copy-on-write should be done, if the
+underlying file-system supports it. Possible values are 'always, 'auto
+and 'never, with 'auto being the default.
+
+This speeds up copying large files a lot while saving the disk space.
+
* Bug fixes
** (ice-9 suspendable-ports) incorrect UTF-8 decoding
diff --git a/configure.ac b/configure.ac
index d0a2dc79b..c46586e9b 100644
--- a/configure.ac
+++ b/configure.ac
@@ -418,6 +418,7 @@ AC_SUBST([SCM_I_GSC_HAVE_STRUCT_DIRENT64])
# sys/sendfile.h - non-POSIX, found in glibc
#
AC_CHECK_HEADERS([complex.h fenv.h io.h memory.h process.h \
+linux/fs.h \
sys/dir.h sys/ioctl.h sys/select.h \
sys/time.h sys/timeb.h sys/times.h sys/stdtypes.h sys/types.h \
sys/utime.h unistd.h utime.h pwd.h grp.h sys/utsname.h \
diff --git a/doc/ref/posix.texi b/doc/ref/posix.texi
index fec42d061..d26808d91 100644
--- a/doc/ref/posix.texi
+++ b/doc/ref/posix.texi
@@ -896,10 +896,17 @@ of @code{delete-file}. Why doesn't POSIX have a @code{rmdirat} function
for this instead? No idea!
@end deffn
-@deffn {Scheme Procedure} copy-file oldfile newfile
+@deffn {Scheme Procedure} copy-file @var{oldfile} @var{newfile} @
+ [#:copy-on-write='auto]
@deffnx {C Function} scm_copy_file (oldfile, newfile)
Copy the file specified by @var{oldfile} to @var{newfile}.
The return value is unspecified.
+
+@code{#:copy-on-write} keyword argument determines whether copy-on-write
+copy should be attempted and the behavior in case of failure. Possible
+values are @code{'always} (attempt the copy-on-write, return error if it
+fails), @code{'auto} (attempt the copy-on-write, fallback to regular
+copy if it fails) and @code{'never} (perform the regular copy).
@end deffn
@deffn {Scheme Procedure} sendfile out in count [offset]
diff --git a/libguile/filesys.c b/libguile/filesys.c
index 1f0bba556..5be42b825 100644
--- a/libguile/filesys.c
+++ b/libguile/filesys.c
@@ -67,6 +67,11 @@
# include <sys/sendfile.h>
#endif
+#if defined(HAVE_SYS_IOCTL_H) && defined(HAVE_LINUX_FS_H)
+# include <linux/fs.h>
+# include <sys/ioctl.h>
+#endif
+
#include "async.h"
#include "boolean.h"
#include "dynwind.h"
@@ -75,6 +80,7 @@
#include "fports.h"
#include "gsubr.h"
#include "iselect.h"
+#include "keywords.h"
#include "list.h"
#include "load.h" /* for scm_i_mirror_backslashes */
#include "modules.h"
@@ -1255,20 +1261,49 @@ SCM_DEFINE (scm_readlink, "readlink", 1, 0, 0,
}
#undef FUNC_NAME
-SCM_DEFINE (scm_copy_file, "copy-file", 2, 0, 0,
- (SCM oldfile, SCM newfile),
+static int
+clone_file (int oldfd, int newfd)
+{
+#ifdef FICLONE
+ return ioctl (newfd, FICLONE, oldfd);
+#else
+ (void)oldfd;
+ (void)newfd;
+ errno = EOPNOTSUPP;
+ return -1;
+#endif
+}
+
+SCM_KEYWORD (k_copy_on_write, "copy-on-write");
+SCM_SYMBOL (sym_always, "always");
+SCM_SYMBOL (sym_auto, "auto");
+SCM_SYMBOL (sym_never, "never");
+
+SCM_DEFINE (scm_copy_file2, "copy-file", 2, 0, 1,
+ (SCM oldfile, SCM newfile, SCM rest),
"Copy the file specified by @var{oldfile} to @var{newfile}.\n"
- "The return value is unspecified.")
-#define FUNC_NAME s_scm_copy_file
+ "The return value is unspecified.\n"
+ "\n"
+ "@code{#:copy-on-write} keyword argument determines whether "
+ "copy-on-write copy should be attempted and the "
+ "behavior in case of failure. Possible values are "
+ "@code{'always} (attempt the copy-on-write, return error if "
+ "it fails), @code{'auto} (attempt the copy-on-write, "
+ "fallback to regular copy if it fails) and @code{'never} "
+ "(perform the regular copy)."
+ )
+#define FUNC_NAME s_scm_copy_file2
{
char *c_oldfile, *c_newfile;
int oldfd, newfd;
int n, rv;
+ SCM cow = sym_auto;
+ int clone_res;
char buf[BUFSIZ];
struct stat_or_stat64 oldstat;
scm_dynwind_begin (0);
-
+
c_oldfile = scm_to_locale_string (oldfile);
scm_dynwind_free (c_oldfile);
c_newfile = scm_to_locale_string (newfile);
@@ -1292,13 +1327,30 @@ SCM_DEFINE (scm_copy_file, "copy-file", 2, 0, 0,
SCM_SYSERROR;
}
- while ((n = read (oldfd, buf, sizeof buf)) > 0)
- if (write (newfd, buf, n) != n)
- {
- close (oldfd);
- close (newfd);
- SCM_SYSERROR;
- }
+ scm_c_bind_keyword_arguments ("copy-file", rest, 0,
+ k_copy_on_write, &cow,
+ SCM_UNDEFINED);
+
+ if (scm_is_eq (cow, sym_always) || scm_is_eq (cow, sym_auto))
+ clone_res = clone_file(oldfd, newfd);
+ else if (scm_is_eq (cow, sym_never))
+ clone_res = -1;
+ else
+ scm_misc_error ("copy-file",
+ "invalid value for #:copy-on-write: ~S",
+ scm_list_1 (cow));
+
+ if (scm_is_eq (cow, sym_always) && clone_res)
+ scm_syserror ("copy-file: copy-on-write failed");
+
+ if (clone_res)
+ while ((n = read (oldfd, buf, sizeof buf)) > 0)
+ if (write (newfd, buf, n) != n)
+ {
+ close (oldfd);
+ close (newfd);
+ SCM_SYSERROR;
+ }
close (oldfd);
if (close (newfd) == -1)
SCM_SYSERROR;
@@ -1308,6 +1360,12 @@ SCM_DEFINE (scm_copy_file, "copy-file", 2, 0, 0,
}
#undef FUNC_NAME
+SCM
+scm_copy_file (SCM oldfile, SCM newfile)
+{
+ return scm_copy_file2 (oldfile, newfile, SCM_UNSPECIFIED);
+}
+
SCM_DEFINE (scm_sendfile, "sendfile", 3, 1, 0,
(SCM out, SCM in, SCM count, SCM offset),
"Send @var{count} bytes from @var{in} to @var{out}, both of which "
diff --git a/libguile/filesys.h b/libguile/filesys.h
index 1ce50d30e..8e849fe7a 100644
--- a/libguile/filesys.h
+++ b/libguile/filesys.h
@@ -74,6 +74,7 @@ SCM_API SCM scm_symlinkat (SCM dir, SCM oldpath, SCM newpath);
SCM_API SCM scm_readlink (SCM path);
SCM_API SCM scm_lstat (SCM str);
SCM_API SCM scm_copy_file (SCM oldfile, SCM newfile);
+SCM_INTERNAL SCM scm_copy_file2 (SCM oldfile, SCM newfile, SCM rest);
SCM_API SCM scm_mkstemp (SCM tmpl);
SCM_API SCM scm_mkdtemp (SCM tmpl);
SCM_API SCM scm_dirname (SCM filename);
--
2.41.0
^ permalink raw reply related [flat|nested] 8+ messages in thread* bug#68504: [PATCH v3] Add copy-on-write support to scm_copy_file.
2024-01-16 12:48 bug#68504: [PATCH] Add copy-on-write support to scm_copy_file Tomas Volf
2024-01-24 10:26 ` Ludovic Courtès
2024-01-24 19:10 ` bug#68504: [PATCH v2] " Tomas Volf
@ 2024-01-24 19:14 ` Tomas Volf
2024-03-12 13:06 ` Ludovic Courtès
2 siblings, 1 reply; 8+ messages in thread
From: Tomas Volf @ 2024-01-24 19:14 UTC (permalink / raw)
To: 68504; +Cc: Tomas Volf
On modern file-systems (BTRFS, ZFS) it is possible to copy a file using
copy-on-write method. For large files it has the advantage of being
much faster and saving disk space (since identical extents are not
duplicated). This feature is stable and for example coreutils' `cp'
does use it automatically (see --reflink).
This commit adds support for this feature into our copy-file procedure.
Same as `cp', it defaults to 'auto, meaning the copy-on-write is
attempted, and in case of failure the regular copy is performed.
No tests are provided, because the behavior depends on the system,
underlying file-system and its configuration. That makes it challenging
to write a test for it. Manual testing was performed instead:
$ btrfs filesystem du /tmp/cow*
Total Exclusive Set shared Filename
36.00KiB 36.00KiB 0.00B /tmp/cow
$ cat cow-test.scm
(copy-file "/tmp/cow" "/tmp/cow-unspecified")
(copy-file "/tmp/cow" "/tmp/cow-always" #:copy-on-write 'always)
(copy-file "/tmp/cow" "/tmp/cow-auto" #:copy-on-write 'auto)
(copy-file "/tmp/cow" "/tmp/cow-never" #:copy-on-write 'never)
(copy-file "/tmp/cow" "/dev/shm/cow-unspecified")
(copy-file "/tmp/cow" "/dev/shm/cow-auto" #:copy-on-write 'auto)
(copy-file "/tmp/cow" "/dev/shm/cow-never" #:copy-on-write 'never)
$ ./meta/guile -s cow-test.scm
$ btrfs filesystem du /tmp/cow*
Total Exclusive Set shared Filename
36.00KiB 0.00B 36.00KiB /tmp/cow
36.00KiB 0.00B 36.00KiB /tmp/cow-always
36.00KiB 0.00B 36.00KiB /tmp/cow-auto
36.00KiB 36.00KiB 0.00B /tmp/cow-never
36.00KiB 0.00B 36.00KiB /tmp/cow-unspecified
$ sha1sum /tmp/cow* /dev/shm/cow*
4c665f87b5dc2e7d26279c4b48968d085e1ace32 /tmp/cow
4c665f87b5dc2e7d26279c4b48968d085e1ace32 /tmp/cow-always
4c665f87b5dc2e7d26279c4b48968d085e1ace32 /tmp/cow-auto
4c665f87b5dc2e7d26279c4b48968d085e1ace32 /tmp/cow-never
4c665f87b5dc2e7d26279c4b48968d085e1ace32 /tmp/cow-unspecified
4c665f87b5dc2e7d26279c4b48968d085e1ace32 /dev/shm/cow-auto
4c665f87b5dc2e7d26279c4b48968d085e1ace32 /dev/shm/cow-never
4c665f87b5dc2e7d26279c4b48968d085e1ace32 /dev/shm/cow-unspecified
This commit also adds to new failure modes for (copy-file).
Failure to copy-on-write when 'always was passed in:
scheme@(guile-user)> (copy-file "/tmp/cow" "/dev/shm/cow" #:copy-on-write 'always)
ice-9/boot-9.scm:1676:22: In procedure raise-exception:
In procedure copy-file: copy-on-write failed: Invalid cross-device link
Passing in invalid value for the #:copy-on-write keyword argument:
scheme@(guile-user)> (copy-file "/tmp/cow" "/dev/shm/cow" #:copy-on-write 'nevr)
ice-9/boot-9.scm:1676:22: In procedure raise-exception:
In procedure copy-file: invalid value for #:copy-on-write: nevr
* NEWS: Add note for copy-file supporting copy-on-write.
* configure.ac: Check for linux/fs.h.
* doc/ref/posix.texi (File System)[copy-file]: Document the new
signature.
* libguile/filesys.c (clone_file): New function cloning a file using
FICLONE, if supported.
(k_copy_on_write): New keyword.
(sym_always, sym_auto, sym_never): New symbols.
(scm_copy_file2): Renamed from scm_copy_file. New #:copy-on-write
keyword argument. Attempt copy-on-write copy by default.
(scm_copy_file): Call scm_copy_file2.
* libguile/filesys.h: Add scm_copy_file2 as SCM_INTERNAL.
---
v2: Introduce scm_copy_file2 in order to preserve backwards compatibility.
v3: Remove mention of scm_copy_file from the commit message.
NEWS | 9 +++++
configure.ac | 1 +
doc/ref/posix.texi | 9 ++++-
libguile/filesys.c | 82 +++++++++++++++++++++++++++++++++++++++-------
libguile/filesys.h | 1 +
5 files changed, 89 insertions(+), 13 deletions(-)
diff --git a/NEWS b/NEWS
index b319404d7..9147098c9 100644
--- a/NEWS
+++ b/NEWS
@@ -21,6 +21,15 @@ definitely unused---this is notably the case for modules that are only
used at macro-expansion time, such as (srfi srfi-26). In those cases,
the compiler reports it as "possibly unused".
+** copy-file now supports copy-on-write
+
+The copy-file procedure now takes an additional keyword argument,
+#:copy-on-write, specifying whether copy-on-write should be done, if the
+underlying file-system supports it. Possible values are 'always, 'auto
+and 'never, with 'auto being the default.
+
+This speeds up copying large files a lot while saving the disk space.
+
* Bug fixes
** (ice-9 suspendable-ports) incorrect UTF-8 decoding
diff --git a/configure.ac b/configure.ac
index d0a2dc79b..c46586e9b 100644
--- a/configure.ac
+++ b/configure.ac
@@ -418,6 +418,7 @@ AC_SUBST([SCM_I_GSC_HAVE_STRUCT_DIRENT64])
# sys/sendfile.h - non-POSIX, found in glibc
#
AC_CHECK_HEADERS([complex.h fenv.h io.h memory.h process.h \
+linux/fs.h \
sys/dir.h sys/ioctl.h sys/select.h \
sys/time.h sys/timeb.h sys/times.h sys/stdtypes.h sys/types.h \
sys/utime.h unistd.h utime.h pwd.h grp.h sys/utsname.h \
diff --git a/doc/ref/posix.texi b/doc/ref/posix.texi
index fec42d061..d26808d91 100644
--- a/doc/ref/posix.texi
+++ b/doc/ref/posix.texi
@@ -896,10 +896,17 @@ of @code{delete-file}. Why doesn't POSIX have a @code{rmdirat} function
for this instead? No idea!
@end deffn
-@deffn {Scheme Procedure} copy-file oldfile newfile
+@deffn {Scheme Procedure} copy-file @var{oldfile} @var{newfile} @
+ [#:copy-on-write='auto]
@deffnx {C Function} scm_copy_file (oldfile, newfile)
Copy the file specified by @var{oldfile} to @var{newfile}.
The return value is unspecified.
+
+@code{#:copy-on-write} keyword argument determines whether copy-on-write
+copy should be attempted and the behavior in case of failure. Possible
+values are @code{'always} (attempt the copy-on-write, return error if it
+fails), @code{'auto} (attempt the copy-on-write, fallback to regular
+copy if it fails) and @code{'never} (perform the regular copy).
@end deffn
@deffn {Scheme Procedure} sendfile out in count [offset]
diff --git a/libguile/filesys.c b/libguile/filesys.c
index 1f0bba556..5be42b825 100644
--- a/libguile/filesys.c
+++ b/libguile/filesys.c
@@ -67,6 +67,11 @@
# include <sys/sendfile.h>
#endif
+#if defined(HAVE_SYS_IOCTL_H) && defined(HAVE_LINUX_FS_H)
+# include <linux/fs.h>
+# include <sys/ioctl.h>
+#endif
+
#include "async.h"
#include "boolean.h"
#include "dynwind.h"
@@ -75,6 +80,7 @@
#include "fports.h"
#include "gsubr.h"
#include "iselect.h"
+#include "keywords.h"
#include "list.h"
#include "load.h" /* for scm_i_mirror_backslashes */
#include "modules.h"
@@ -1255,20 +1261,49 @@ SCM_DEFINE (scm_readlink, "readlink", 1, 0, 0,
}
#undef FUNC_NAME
-SCM_DEFINE (scm_copy_file, "copy-file", 2, 0, 0,
- (SCM oldfile, SCM newfile),
+static int
+clone_file (int oldfd, int newfd)
+{
+#ifdef FICLONE
+ return ioctl (newfd, FICLONE, oldfd);
+#else
+ (void)oldfd;
+ (void)newfd;
+ errno = EOPNOTSUPP;
+ return -1;
+#endif
+}
+
+SCM_KEYWORD (k_copy_on_write, "copy-on-write");
+SCM_SYMBOL (sym_always, "always");
+SCM_SYMBOL (sym_auto, "auto");
+SCM_SYMBOL (sym_never, "never");
+
+SCM_DEFINE (scm_copy_file2, "copy-file", 2, 0, 1,
+ (SCM oldfile, SCM newfile, SCM rest),
"Copy the file specified by @var{oldfile} to @var{newfile}.\n"
- "The return value is unspecified.")
-#define FUNC_NAME s_scm_copy_file
+ "The return value is unspecified.\n"
+ "\n"
+ "@code{#:copy-on-write} keyword argument determines whether "
+ "copy-on-write copy should be attempted and the "
+ "behavior in case of failure. Possible values are "
+ "@code{'always} (attempt the copy-on-write, return error if "
+ "it fails), @code{'auto} (attempt the copy-on-write, "
+ "fallback to regular copy if it fails) and @code{'never} "
+ "(perform the regular copy)."
+ )
+#define FUNC_NAME s_scm_copy_file2
{
char *c_oldfile, *c_newfile;
int oldfd, newfd;
int n, rv;
+ SCM cow = sym_auto;
+ int clone_res;
char buf[BUFSIZ];
struct stat_or_stat64 oldstat;
scm_dynwind_begin (0);
-
+
c_oldfile = scm_to_locale_string (oldfile);
scm_dynwind_free (c_oldfile);
c_newfile = scm_to_locale_string (newfile);
@@ -1292,13 +1327,30 @@ SCM_DEFINE (scm_copy_file, "copy-file", 2, 0, 0,
SCM_SYSERROR;
}
- while ((n = read (oldfd, buf, sizeof buf)) > 0)
- if (write (newfd, buf, n) != n)
- {
- close (oldfd);
- close (newfd);
- SCM_SYSERROR;
- }
+ scm_c_bind_keyword_arguments ("copy-file", rest, 0,
+ k_copy_on_write, &cow,
+ SCM_UNDEFINED);
+
+ if (scm_is_eq (cow, sym_always) || scm_is_eq (cow, sym_auto))
+ clone_res = clone_file(oldfd, newfd);
+ else if (scm_is_eq (cow, sym_never))
+ clone_res = -1;
+ else
+ scm_misc_error ("copy-file",
+ "invalid value for #:copy-on-write: ~S",
+ scm_list_1 (cow));
+
+ if (scm_is_eq (cow, sym_always) && clone_res)
+ scm_syserror ("copy-file: copy-on-write failed");
+
+ if (clone_res)
+ while ((n = read (oldfd, buf, sizeof buf)) > 0)
+ if (write (newfd, buf, n) != n)
+ {
+ close (oldfd);
+ close (newfd);
+ SCM_SYSERROR;
+ }
close (oldfd);
if (close (newfd) == -1)
SCM_SYSERROR;
@@ -1308,6 +1360,12 @@ SCM_DEFINE (scm_copy_file, "copy-file", 2, 0, 0,
}
#undef FUNC_NAME
+SCM
+scm_copy_file (SCM oldfile, SCM newfile)
+{
+ return scm_copy_file2 (oldfile, newfile, SCM_UNSPECIFIED);
+}
+
SCM_DEFINE (scm_sendfile, "sendfile", 3, 1, 0,
(SCM out, SCM in, SCM count, SCM offset),
"Send @var{count} bytes from @var{in} to @var{out}, both of which "
diff --git a/libguile/filesys.h b/libguile/filesys.h
index 1ce50d30e..8e849fe7a 100644
--- a/libguile/filesys.h
+++ b/libguile/filesys.h
@@ -74,6 +74,7 @@ SCM_API SCM scm_symlinkat (SCM dir, SCM oldpath, SCM newpath);
SCM_API SCM scm_readlink (SCM path);
SCM_API SCM scm_lstat (SCM str);
SCM_API SCM scm_copy_file (SCM oldfile, SCM newfile);
+SCM_INTERNAL SCM scm_copy_file2 (SCM oldfile, SCM newfile, SCM rest);
SCM_API SCM scm_mkstemp (SCM tmpl);
SCM_API SCM scm_mkdtemp (SCM tmpl);
SCM_API SCM scm_dirname (SCM filename);
--
2.41.0
^ permalink raw reply related [flat|nested] 8+ messages in thread* bug#68504: [PATCH v3] Add copy-on-write support to scm_copy_file.
2024-01-24 19:14 ` bug#68504: [PATCH v3] " Tomas Volf
@ 2024-03-12 13:06 ` Ludovic Courtès
2024-03-12 23:19 ` Tomas Volf
0 siblings, 1 reply; 8+ messages in thread
From: Ludovic Courtès @ 2024-03-12 13:06 UTC (permalink / raw)
To: Tomas Volf; +Cc: 68504-done
Hi Tomas,
Tomas Volf <~@wolfsden.cz> skribis:
> On modern file-systems (BTRFS, ZFS) it is possible to copy a file using
> copy-on-write method. For large files it has the advantage of being
> much faster and saving disk space (since identical extents are not
> duplicated). This feature is stable and for example coreutils' `cp'
> does use it automatically (see --reflink).
>
> This commit adds support for this feature into our copy-file procedure.
> Same as `cp', it defaults to 'auto, meaning the copy-on-write is
> attempted, and in case of failure the regular copy is performed.
>
> No tests are provided, because the behavior depends on the system,
> underlying file-system and its configuration. That makes it challenging
> to write a test for it. Manual testing was performed instead:
>
> $ btrfs filesystem du /tmp/cow*
> Total Exclusive Set shared Filename
> 36.00KiB 36.00KiB 0.00B /tmp/cow
>
> $ cat cow-test.scm
> (copy-file "/tmp/cow" "/tmp/cow-unspecified")
> (copy-file "/tmp/cow" "/tmp/cow-always" #:copy-on-write 'always)
> (copy-file "/tmp/cow" "/tmp/cow-auto" #:copy-on-write 'auto)
> (copy-file "/tmp/cow" "/tmp/cow-never" #:copy-on-write 'never)
> (copy-file "/tmp/cow" "/dev/shm/cow-unspecified")
> (copy-file "/tmp/cow" "/dev/shm/cow-auto" #:copy-on-write 'auto)
> (copy-file "/tmp/cow" "/dev/shm/cow-never" #:copy-on-write 'never)
> $ ./meta/guile -s cow-test.scm
>
> $ btrfs filesystem du /tmp/cow*
> Total Exclusive Set shared Filename
> 36.00KiB 0.00B 36.00KiB /tmp/cow
> 36.00KiB 0.00B 36.00KiB /tmp/cow-always
> 36.00KiB 0.00B 36.00KiB /tmp/cow-auto
> 36.00KiB 36.00KiB 0.00B /tmp/cow-never
> 36.00KiB 0.00B 36.00KiB /tmp/cow-unspecified
>
> $ sha1sum /tmp/cow* /dev/shm/cow*
> 4c665f87b5dc2e7d26279c4b48968d085e1ace32 /tmp/cow
> 4c665f87b5dc2e7d26279c4b48968d085e1ace32 /tmp/cow-always
> 4c665f87b5dc2e7d26279c4b48968d085e1ace32 /tmp/cow-auto
> 4c665f87b5dc2e7d26279c4b48968d085e1ace32 /tmp/cow-never
> 4c665f87b5dc2e7d26279c4b48968d085e1ace32 /tmp/cow-unspecified
> 4c665f87b5dc2e7d26279c4b48968d085e1ace32 /dev/shm/cow-auto
> 4c665f87b5dc2e7d26279c4b48968d085e1ace32 /dev/shm/cow-never
> 4c665f87b5dc2e7d26279c4b48968d085e1ace32 /dev/shm/cow-unspecified
>
> This commit also adds to new failure modes for (copy-file).
>
> Failure to copy-on-write when 'always was passed in:
>
> scheme@(guile-user)> (copy-file "/tmp/cow" "/dev/shm/cow" #:copy-on-write 'always)
> ice-9/boot-9.scm:1676:22: In procedure raise-exception:
> In procedure copy-file: copy-on-write failed: Invalid cross-device link
>
> Passing in invalid value for the #:copy-on-write keyword argument:
>
> scheme@(guile-user)> (copy-file "/tmp/cow" "/dev/shm/cow" #:copy-on-write 'nevr)
> ice-9/boot-9.scm:1676:22: In procedure raise-exception:
> In procedure copy-file: invalid value for #:copy-on-write: nevr
>
> * NEWS: Add note for copy-file supporting copy-on-write.
> * configure.ac: Check for linux/fs.h.
> * doc/ref/posix.texi (File System)[copy-file]: Document the new
> signature.
> * libguile/filesys.c (clone_file): New function cloning a file using
> FICLONE, if supported.
> (k_copy_on_write): New keyword.
> (sym_always, sym_auto, sym_never): New symbols.
> (scm_copy_file2): Renamed from scm_copy_file. New #:copy-on-write
> keyword argument. Attempt copy-on-write copy by default.
> (scm_copy_file): Call scm_copy_file2.
> * libguile/filesys.h: Add scm_copy_file2 as SCM_INTERNAL.
> ---
> v2: Introduce scm_copy_file2 in order to preserve backwards compatibility.
>
> v3: Remove mention of scm_copy_file from the commit message.
Finally pushed as e1690f3fd251d69b3687ec12c6f4b41034047f0f. Note that I
added copyright lines for you, let me know if I got it wrong.
As a followup, we should add support for ‘copy_file_range’ when FICLONE
cannot be used; glibc supports it on all platforms but it returns ENOSYS
on GNU/Hurd currently.
WDYT?
Thank you!
Ludo’.
^ permalink raw reply [flat|nested] 8+ messages in thread
* bug#68504: [PATCH v3] Add copy-on-write support to scm_copy_file.
2024-03-12 13:06 ` Ludovic Courtès
@ 2024-03-12 23:19 ` Tomas Volf
2024-03-21 14:23 ` Ludovic Courtès
0 siblings, 1 reply; 8+ messages in thread
From: Tomas Volf @ 2024-03-12 23:19 UTC (permalink / raw)
To: Ludovic Courtès; +Cc: 68504
[-- Attachment #1: Type: text/plain, Size: 1413 bytes --]
On 2024-03-12 14:06:34 +0100, Ludovic Courtès wrote:
>
> Finally pushed as e1690f3fd251d69b3687ec12c6f4b41034047f0f. Note that I
> added copyright lines for you, let me know if I got it wrong.
Thank you for merging it, and thanks for the copyright, looks correct :)
> As a followup, we should add support for ‘copy_file_range’ when FICLONE
> cannot be used; glibc supports it on all platforms but it returns ENOSYS
> on GNU/Hurd currently.
>
> WDYT?
Sure, I am willing to do my part. I managed to find this blog post[0], so after
some minor troubles I did manage to get a VM with GNU/Hurd running. Next I will
read up on copy_file_range and try to put together a patch.
Just to make sure, your idea here is exactly what? Always try to use
copy_file_range before the regular copy? So the flow would be
For 'always case:
CoW ---fail--> FAIL
For 'auto case:
CoW ---fail--> copy_file_range ---fail--> current copy ---fail--> FAIL
For 'never case:
copy_file_range ---fail--> current copy ---fail--> FAIL
Is that an accurate summary? Or did you mean only as a fallback for the CoW, so
only for 'auto, but not for the 'never?
Tomas
0: https://guix.gnu.org/en/blog/2020/a-hello-world-virtual-machine-running-the-hurd/
--
There are only two hard things in Computer Science:
cache invalidation, naming things and off-by-one errors.
[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]
^ permalink raw reply [flat|nested] 8+ messages in thread* bug#68504: [PATCH v3] Add copy-on-write support to scm_copy_file.
2024-03-12 23:19 ` Tomas Volf
@ 2024-03-21 14:23 ` Ludovic Courtès
0 siblings, 0 replies; 8+ messages in thread
From: Ludovic Courtès @ 2024-03-21 14:23 UTC (permalink / raw)
To: Tomas Volf; +Cc: 68504
Hi,
Tomas Volf <~@wolfsden.cz> skribis:
> Sure, I am willing to do my part. I managed to find this blog post[0], so after
> some minor troubles I did manage to get a VM with GNU/Hurd running. Next I will
> read up on copy_file_range and try to put together a patch.
It’s really just (service hurd-vm-service-type) on Guix System:
https://guix.gnu.org/manual/devel/en/html_node/Virtualization-Services.html#The-Hurd-in-a-Virtual-Machine
> Just to make sure, your idea here is exactly what? Always try to use
> copy_file_range before the regular copy? So the flow would be
>
> For 'always case:
>
> CoW ---fail--> FAIL
>
> For 'auto case:
>
> CoW ---fail--> copy_file_range ---fail--> current copy ---fail--> FAIL
>
> For 'never case:
>
> copy_file_range ---fail--> current copy ---fail--> FAIL
>
> Is that an accurate summary?
Yes, that’s exactly what I had in mind.
Actually it might be better to use sendfile(2), which is slightly less
generic but otherwise equivalent AFAICS, and which happens to have a
Hurd implementation in glibc.
Thanks for your help!
Ludo’.
^ permalink raw reply [flat|nested] 8+ messages in thread