um: Accept /dev/fd/* uml block devices

Useful for
* limiting privileges
* opening block devices O_EXCL

Use dup to work around the fact /proc/self/fd can't be opened after dropping privileges.
This proc behaviour doesn't match TLPI and might be a bug.

Qemu has a slightly more complex fdset approach that provides fds with different access permissions.

Signed-off-by Gabriel de Perthuis <[email protected]>

---
 arch/um/os-Linux/file.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)

arch/um/os-Linux/file.c +7 -3

index c17bd6f..cee65ba 100644
@@ -169,11 +169,11 @@ int os_file_mode(const char *file, struct openflags *mode_out)
 	return err;
 }
 
 int os_open_file(const char *file, struct openflags flags, int mode)
 {
-	int fd, err, f = 0;
+	int fd, fd0, err, f = 0;
 
 	if (flags.r && flags.w)
 		f = O_RDWR;
 	else if (flags.r)
 		f = O_RDONLY;

@@ -190,11 +190,15 @@ int os_open_file(const char *file, struct openflags flags, int mode)
 	if (flags.e)
 		f |= O_EXCL;
 	if (flags.a)
 		f |= O_APPEND;
 
-	fd = open64(file, f, mode);
+	if (!strncmp(file, "/dev/fd/", 8)
+	    && sscanf(file, "/dev/fd/%d", &fd0) == 1)
+		fd = dup(fd0);
+	else
+		fd = open64(file, f, mode);
 	if (fd < 0)
 		return -errno;
 
 	if (flags.cl && fcntl(fd, F_SETFD, 1)) {
 		err = -errno;

@@ -280,11 +284,11 @@ int os_file_size(const char *file, unsigned long long *size_out)
 
 	if (S_ISBLK(buf.ust_mode)) {
 		int fd;
 		long blocks;
 
-		fd = open(file, O_RDONLY, 0);
+		fd = os_open_file(file, of_read(OPENFLAGS()), 0);
 		if (fd < 0) {
 			err = -errno;
 			printk(UM_KERN_ERR "Couldn't open \"%s\", "
 			       "errno = %d\n", file, errno);
 			return err;

Am 27.07.2013 17:23, schrieb Gabriel de Perthuis:

Quoted 3 lines

Useful for
* limiting privileges
* opening block devices O_EXCL

So, the goal of this patch is to allow passing a file descriptor number as block device instead of a file?

I assume you have already a wrapper around UML which exec()'s it such that it can reuse a fd?

Quoted 6 lines

Use dup to work around the fact /proc/self/fd can't be opened after dropping privileges.
This proc behaviour doesn't match TLPI and might be a bug.

Qemu has a slightly more complex fdset approach that provides fds with different access permissions.

I really don't like that you patch os_open_file(), this is a generic function.

What about this one?
Allow ubda= (and all other UML block device kernel parameters) to accept arguments like file:/foo/bar and fd:N.
Where N is a number and file: is default such that we do not break old kernels.

Quoted 7 lines · Richard Weinberger

Am 27.07.2013 17:23, schrieb Gabriel de Perthuis:

Quoted 3 lines

Useful for
* limiting privileges
* opening block devices O_EXCL

So, the goal of this patch is to allow passing a file descriptor number as block device instead of a file?

Yes. It turns out it already works, but not after dropping privileges.

Quoted 2 lines

I assume you have already a wrapper around UML which exec()'s it such that it can reuse a fd?

Yes, vido: https://github.com/g2p/vido

Here's the relevant commit:
https://github.com/g2p/vido/commit/42d4b86eab13d90ee63138b73146485dc4e47ec6

Quoted 9 lines

Quoted 6 lines

Use dup to work around the fact /proc/self/fd can't be opened after dropping privileges.
This proc behaviour doesn't match TLPI and might be a bug.

Qemu has a slightly more complex fdset approach that provides fds with different access permissions.

I really don't like that you patch os_open_file(), this is a generic function.

The justification was that it unbreaks open("/dev/fd") to be more like standards suggest, but I can see how that makes it a special case.

Quoted 5 lines

What about this one?
Allow ubda= (and all other UML block device kernel parameters) to accept arguments like file:/foo/bar and fd:N.
Where N is a number and file: is default such that we do not break old kernels.

Okay, I'll add a prefix. Maybe file:// + /abs/path | rel/path since that's already standard.