Skip to contentSkip to conversation

add futimesat syscall

PATCH In review Riku Voipio 5 participants 11 messages Sep 17, 2008 – Sep 18, 2008

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 add futimesat syscall

Add futimes patch, originally from scratchbox qemu devkit.

Signature · 1 line
"rm -rf" only sounds scary if you don't have backups
linux-user/syscall.c +24 -0
Index: trunk/linux-user/syscall.c
===================================================================
@@ -156,6 +156,7 @@
 #define __NR_sys_faccessat __NR_faccessat
 #define __NR_sys_fchmodat __NR_fchmodat
 #define __NR_sys_fchownat __NR_fchownat
+#define __NR_sys_futimesat __NR_futimesat
 #define __NR_sys_getcwd1 __NR_getcwd
 #define __NR_sys_getdents __NR_getdents
 #define __NR_sys_getdents64 __NR_getdents64
@@ -200,6 +201,10 @@
 _syscall5(int,sys_fchownat,int,dirfd,const char *,pathname,
           uid_t,owner,gid_t,group,int,flags)
 #endif
+#if defined(TARGET_NR_futimesat) && defined(__NR_futimesat)
+_syscall3(int,sys_futimesat,int,dirfd,const char *,pathname,
+          const struct timeval *,times)
+#endif
 _syscall2(int,sys_getcwd1,char *,buf,size_t,size)
 #if TARGET_ABI_BITS == 32
 _syscall3(int, sys_getdents, uint, fd, struct dirent *, dirp, uint, count);
@@ -5743,6 +5748,25 @@
 	break;
 #endif
 
+#if defined(TARGET_NR_futimesat) && defined(__NR_futimesat)
+    case TARGET_NR_futimesat:
+        {
+            struct timeval tv[2];
+            if (copy_from_user_timeval(tv, arg3)
+                || copy_from_user_timeval(tv+1, arg3+sizeof(struct target_timeval))) {
+                goto efault;
+            }
+            if (!arg2)
+                ret = get_errno(sys_futimesat(arg1, NULL, tv));
+            else {
+                p = lock_user_string(arg2);
+                ret = get_errno(sys_futimesat(arg1, path(p), tv));
+                unlock_user(p, arg2, 0);
+            }
+        }
+        break;
+#endif
+
 #ifdef TARGET_NR_set_robust_list
     case TARGET_NR_set_robust_list:
 	goto unimplemented_nowarn;
#2 lore raw
Quoted 9 lines · andrzej zaborowski

2008/9/17 Riku Voipio <[email protected]>:

Quoted 1 line

Add futimes patch, originally from scratchbox qemu devkit.

Kirill Shutemov submitted an imlementation of futimensat() at
http://lists.gnu.org/archive/html/qemu-devel/2008-09/msg00439.html , it looks cleaner to me.

I would like to commit Kirill's series at some point if nobody finds any issues with them.

Please, don't commit binfmt_misc related patch. I'll fix it soon.

Signature · 3 lines
Regards,  Kirill A. Shutemov
 + Belarus, Minsk
 + ALT Linux Team, http://www.altlinux.com/
#4 lore raw
Quoted 9 lines · Riku Voipio
Quoted 6 lines · andrzej zaborowski

2008/9/17 Riku Voipio <[email protected]>:

Quoted 1 line

Add futimes patch, originally from scratchbox qemu devkit.

Kirill Shutemov submitted an imlementation of futimensat() at
http://lists.gnu.org/archive/html/qemu-devel/2008-09/msg00439.html , it looks cleaner to me.

Kirill's patch looks good to me.

There's a functional difference between Kirill's patch and yours:

Kirill + ret = get_errno(futimesat(arg1, p, tvp));
you + ret = get_errno(sys_futimesat(arg1, path(p), tv));

Laurent

#5 lore raw
Quoted 6 lines · Laurent Desnogues
Quoted 4 lines · Riku Voipio
Quoted 2 lines · andrzej zaborowski

http://lists.gnu.org/archive/html/qemu-devel/2008-09/msg00439.html , it looks cleaner to me.

Quoted 6 lines
Quoted 1 line

Kirill's patch looks good to me.

There's a functional difference between Kirill's patch and yours:

Kirill + ret = get_errno(futimesat(arg1, p, tvp));
you + ret = get_errno(sys_futimesat(arg1, path(p), tv));

The sbox patch defines sys_futimesat earlier in the file, presumably to deal with supporting a stone-age version of libc/kernel headers. path() is related to interp_prefix. The old utimes handlers don't use path() either, and it's unlikely that any library-loading routine uses futimes(). Thus, I believe Kirill's version is more correct.

That said, path() should probably renamed to something more obvious (interp_path() ?), or path mangling could be made more generic (allowing things like chroot() emulation).

Signature · 1 line
"rm -rf" only sounds scary if you don't have backups
#6 lore raw
Quoted 9 lines · Riku Voipio
Quoted 7 lines · Laurent Desnogues
Quoted 6 lines · Riku Voipio
Quoted 2 lines · andrzej zaborowski

http://lists.gnu.org/archive/html/qemu-devel/2008-09/msg00439.html , it looks cleaner to me.

Kirill's patch looks good to me.

Quoted 7 lines
Quoted 4 lines

There's a functional difference between Kirill's patch and yours:

Kirill + ret = get_errno(futimesat(arg1, p, tvp));
you + ret = get_errno(sys_futimesat(arg1, path(p), tv));

The sbox patch defines sys_futimesat earlier in the file, presumably to deal with supporting a stone-age version of libc/kernel headers.

I don't think that it's great idea. If libc/kernel headers doesn't provide the syscall we shouldn't implement them. The exception is syscalls without libc's wrapper. Like gettid(2).

Quoted 3 lines

path() is related to interp_prefix. The old utimes handlers don't use path() either, and it's unlikely that any library-loading routine uses futimes(). Thus, I believe Kirill's version is more correct.

I have written futimeat based on utimes(). If utimes() emulation work properly with interp_prefix != "/", futimesat() also will. But I have never test it with interp_prefix != "/"

Signature · 3 lines
Regards,  Kirill A. Shutemov
 + Belarus, Minsk
 + ALT Linux Team, http://www.altlinux.com/
#7 lore raw
Quoted 4 lines · Kirill A. Shutemov

On Thu, Sep 18, 2008 at 12:08:21PM +0300, Riku Voipio wrote:
I don't think that it's great idea. If libc/kernel headers doesn't provide the syscall we shouldn't implement them. The exception is syscalls without libc's wrapper. Like gettid(2).

Well, defining syscalls appears to be a common practice in qemu for the *at family of syscalls (openat, linkat, ...), so it doesn't seem that far off for futimesat().

Signature · 1 line
"rm -rf" only sounds scary if you don't have backups
#8 lore raw
Quoted 9 lines · Riku Voipio
Quoted 4 lines · Kirill A. Shutemov

On Thu, Sep 18, 2008 at 12:08:21PM +0300, Riku Voipio wrote:
I don't think that it's great idea. If libc/kernel headers doesn't provide the syscall we shouldn't implement them. The exception is syscalls without libc's wrapper. Like gettid(2).

Well, defining syscalls appears to be a common practice in qemu for the *at family of syscalls (openat, linkat, ...), so it doesn't seem that far off for futimesat().

*at syscalls ware implemented in linux 2.6.16. glibc's wrappers to this syscalls were implemented in glibc 2.4. Both of them relesed more than two and half years ago.

I think we can drop this crap now.

Signature · 3 lines
Regards,  Kirill A. Shutemov
 + Belarus, Minsk
 + ALT Linux Team, http://www.altlinux.com/
#9 lore raw
Quoted 11 lines · Kirill A. Shutemov
Quoted 9 lines · Riku Voipio
Quoted 4 lines · Kirill A. Shutemov

On Thu, Sep 18, 2008 at 12:08:21PM +0300, Riku Voipio wrote:
I don't think that it's great idea. If libc/kernel headers doesn't provide the syscall we shouldn't implement them. The exception is syscalls without libc's wrapper. Like gettid(2).

Well, defining syscalls appears to be a common practice in qemu for the *at family of syscalls (openat, linkat, ...), so it doesn't seem that far off for futimesat().

Quoted 5 lines

*at syscalls ware implemented in linux 2.6.16. glibc's wrappers to this syscalls were implemented in glibc 2.4. Both of them relesed more than two and half years ago.

I think we can drop this crap now.

Current Debian stable uses glibc 2.3.6.

Thiemo

#10 lore raw
Quoted 12 lines · Thiemo Seufer
Quoted 11 lines · Kirill A. Shutemov
Quoted 9 lines · Riku Voipio
Quoted 4 lines · Kirill A. Shutemov

On Thu, Sep 18, 2008 at 12:08:21PM +0300, Riku Voipio wrote:
I don't think that it's great idea. If libc/kernel headers doesn't provide the syscall we shouldn't implement them. The exception is syscalls without libc's wrapper. Like gettid(2).

Well, defining syscalls appears to be a common practice in qemu for the *at family of syscalls (openat, linkat, ...), so it doesn't seem that far off for futimesat().

Quoted 7 lines
Quoted 5 lines

*at syscalls ware implemented in linux 2.6.16. glibc's wrappers to this syscalls were implemented in glibc 2.4. Both of them relesed more than two and half years ago.

I think we can drop this crap now.

Current Debian stable uses glibc 2.3.6.

Hm... Ok, I'll fix my patches.

Signature · 3 lines
Regards,  Kirill A. Shutemov
 + Belarus, Minsk
 + ALT Linux Team, http://www.altlinux.com/