[musl] Feature request: expose Linux-specific renameat2 function

[musl] Feature request: expose Linux-specific renameat2 function

[vi0oss]

[-- Attachment #1: Type: text/plain, Size: 0 bytes --]



[-- Attachment #2: 0001-expose-renameat2-function-if-such-syscall-is-availab.patch --]
[-- Type: text/x-diff, Size: 1071 bytes --]

From 70d1747dd0fb31df780d2e0255a8e0417899b6ca Mon Sep 17 00:00:00 2001
From: Vitaly _Vi Shukela <vi0oss@gmail.com>
Date: Tue, 21 Feb 2023 00:23:45 +0100
Subject: [PATCH] expose renameat2 function if such syscall is available

Does not include RENAME_EXCHANGE, RENAME_NOREPLACE or
RENAME_WHITEOUT constants that are likely to be the
reason to use renameat2 instead of renameat, but this
should still simplify linking issues arising from the
use of renameat2 and enable simpler workarounds.
---
 src/unistd/renameat.c | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/src/unistd/renameat.c b/src/unistd/renameat.c
index c3b40a25..fe1adaf4 100644
--- a/src/unistd/renameat.c
+++ b/src/unistd/renameat.c
@@ -9,3 +9,10 @@ int renameat(int oldfd, const char *old, int newfd, const char *new)
 	return syscall(SYS_renameat2, oldfd, old, newfd, new, 0);
 #endif
 }
+
+#ifdef SYS_renameat2
+int renameat2(int oldfd, const char *old, int newfd, const char *new, unsigned int flags)
+{
+	return syscall(SYS_renameat2, oldfd, old, newfd, new, flags);
+}
+#endif
-- 
2.39.1

Re: [musl] Feature request: expose Linux-specific renameat2 function

[Markus Wichmann]

On Tue, Feb 21, 2023 at 12:28:41AM +0100, vi0oss wrote:
> Does not include RENAME_EXCHANGE, RENAME_NOREPLACE or
> RENAME_WHITEOUT constants that are likely to be the
> reason to use renameat2 instead of renameat, but this
> should still simplify linking issues arising from the
> use of renameat2 and enable simpler workarounds.

I'm wondering why this system call cannot be handled like every other
new system interface: By making the users copy the necessary definitions
into their projects and wrapping the syscall themselves while the libc
does not contain them.

I don't think the patch could be applied as-is. There is little obvious
reason (and no reason given) why this needs to be added to the
renameat.c file instead of its own file. Nor is it obvious why the
abovementioned constants cannot be added under _GNU_SOURCE. If they are
kernel interface, they cannot change anymore.

Ciao,
Markus