Skip to contentSkip to conversation

nfs client, kernel 2.4.31: readlink result overflow

PATCH In review Assar 4 participants 21 messages Sep 12, 2005 – Sep 14, 2005 NFS client

lore
Share & tools
lore

Assar

author
#0 lore raw

PATCH 1/1 nfs client, kernel 2.4.31: readlink result overflow

In 2.4.31, the v2/3 nfs readlink accepts too long symlinks. I have tested this by having a server return long symlinks. 2.6.13 does not to my reading have this problem.

fs/nfs/nfs2xdr.c +2 -2
diff -u linux-2.4.31.orig/fs/nfs/nfs2xdr.c linux-2.4.31/fs/nfs/nfs2xdr.c
@@ -571,8 +571,8 @@
 	strlen = (u32*)kmap(rcvbuf->pages[0]);
 	/* Convert length of symlink */
 	len = ntohl(*strlen);
-	if (len > rcvbuf->page_len)
-		len = rcvbuf->page_len;
+	if (len > rcvbuf->page_len - 1 - 4)
+		len = rcvbuf->page_len - 1 - 4;
 	*strlen = len;
 	/* NULL terminate the string we got */
 	string = (char *)(strlen + 1);
fs/nfs/nfs3xdr.c +2 -2
diff -u linux-2.4.31.orig/fs/nfs/nfs3xdr.c linux-2.4.31/fs/nfs/nfs3xdr.c
@@ -759,8 +759,8 @@
 	strlen = (u32*)kmap(rcvbuf->pages[0]);
 	/* Convert length of symlink */
 	len = ntohl(*strlen);
-	if (len > rcvbuf->page_len)
-		len = rcvbuf->page_len;
+	if (len > rcvbuf->page_len - 1 - 4)
+		len = rcvbuf->page_len - 1 - 4;
 	*strlen = len;
 	/* NULL terminate the string we got */
 	string = (char *)(strlen + 1);
#1 lore raw
Quoted 3 lines · Assar

In 2.4.31, the v2/3 nfs readlink accepts too long symlinks. I have tested this by having a server return long symlinks. 2.6.13 does not to my reading have this problem.

Odd, as it has similar code - 2.6.13-mm1 nfs2xdr.c has:

        /* Convert length of symlink */
        len = ntohl(*p++);
        if (len >= rcvbuf->page_len || len <= 0) {

Is there some *other* code in 2.6 that prevents this from being a problem, if it's a problem on 2.4?

fs/nfs/nfs2xdr.c line 575

Quoted 11 lines

diff -u linux-2.4.31.orig/fs/nfs/nfs2xdr.c linux-2.4.31/fs/nfs/nfs2xdr.c

fs/nfs/nfs2xdr.c +2 -2
@@ -571,8 +571,8 @@
 	strlen = (u32*)kmap(rcvbuf->pages[0]);
 	/* Convert length of symlink */
 	len = ntohl(*strlen);
-	if (len > rcvbuf->page_len)
-		len = rcvbuf->page_len;
+	if (len > rcvbuf->page_len - 1 - 4)
+		len = rcvbuf->page_len - 1 - 4;

Am I the only one who finds an uncommented "-1 -4" construct scary beyond belief?

author
#2 lore raw
Quoted 4 lines · [email protected] writes:

Odd, as it has similar code - 2.6.13-mm1 nfs2xdr.c has:

        /* Convert length of symlink */
        len = ntohl(*p++);
        if (len >= rcvbuf->page_len || len <= 0) {

To my reading, the 2.6.13 code does not copy the 4 bytes of length to rcvbuf.

Quoted 16 lines

Is there some *other* code in 2.6 that prevents this from being a problem, if it's a problem on 2.4?

fs/nfs/nfs2xdr.c line 575

Quoted 11 lines

diff -u linux-2.4.31.orig/fs/nfs/nfs2xdr.c linux-2.4.31/fs/nfs/nfs2xdr.c

fs/nfs/nfs2xdr.c +2 -2
@@ -571,8 +571,8 @@
 	strlen = (u32*)kmap(rcvbuf->pages[0]);
 	/* Convert length of symlink */
 	len = ntohl(*strlen);
-	if (len > rcvbuf->page_len)
-		len = rcvbuf->page_len;
+	if (len > rcvbuf->page_len - 1 - 4)
+		len = rcvbuf->page_len - 1 - 4;

Am I the only one who finds an uncommented "-1 -4" construct scary beyond belief?

Probably not. What do you think looks better? sizeof(u32) ?

#3 lore raw
Quoted 8 lines · Assar
Quoted 4 lines · [email protected] writes:

Odd, as it has similar code - 2.6.13-mm1 nfs2xdr.c has:

        /* Convert length of symlink */
        len = ntohl(*p++);
        if (len >= rcvbuf->page_len || len <= 0) {

To my reading, the 2.6.13 code does not copy the 4 bytes of length to rcvbuf.

Hmm... it still does this:

	kaddr[len+rcvbuf->page_base] = '\0';

which still has a possible off-by-one? (Was that why you have -1 -4?)

Quoted 3 lines
Quoted 1 line

Am I the only one who finds an uncommented "-1 -4" construct scary beyond belief?

Probably not. What do you think looks better? sizeof(u32) ?

sizeof(actual_var) is even better, as that way it's clear what you're allowing space for.

author
#4 lore raw
Quoted 6 lines · [email protected] writes:
Quoted 2 lines

To my reading, the 2.6.13 code does not copy the 4 bytes of length to rcvbuf.

Hmm... it still does this:

	kaddr[len+rcvbuf->page_base] = '\0';

which still has a possible off-by-one? (Was that why you have -1 -4?)

The check is different. 2.6.13 is using ">=" instead of ">", so hence I think that's fine.

Quoted 2 lines

sizeof(actual_var) is even better, as that way it's clear what you're allowing space for.

fs/nfs/nfs2xdr.c +2 -2
diff -u linux-2.4.31.orig/fs/nfs/nfs2xdr.c linux-2.4.31/fs/nfs/nfs2xdr.c
@@ -571,8 +571,8 @@
 	strlen = (u32*)kmap(rcvbuf->pages[0]);
 	/* Convert length of symlink */
 	len = ntohl(*strlen);
-	if (len > rcvbuf->page_len)
-		len = rcvbuf->page_len;
+	if (len > rcvbuf->page_len - sizeof(*strlen) - 1)
+		len = rcvbuf->page_len - sizeof(*strlen) - 1;
 	*strlen = len;
 	/* NULL terminate the string we got */
 	string = (char *)(strlen + 1);
fs/nfs/nfs3xdr.c +2 -2
diff -u linux-2.4.31.orig/fs/nfs/nfs3xdr.c linux-2.4.31/fs/nfs/nfs3xdr.c
@@ -759,8 +759,8 @@
 	strlen = (u32*)kmap(rcvbuf->pages[0]);
 	/* Convert length of symlink */
 	len = ntohl(*strlen);
-	if (len > rcvbuf->page_len)
-		len = rcvbuf->page_len;
+	if (len > rcvbuf->page_len - sizeof(*strlen) - 1)
+		len = rcvbuf->page_len - sizeof(*strlen) - 1;
 	*strlen = len;
 	/* NULL terminate the string we got */
 	string = (char *)(strlen + 1);
#5 lore raw
Quoted 10 lines · Assar
Quoted 6 lines · [email protected] writes:
Quoted 2 lines

To my reading, the 2.6.13 code does not copy the 4 bytes of length to rcvbuf.

Hmm... it still does this:

	kaddr[len+rcvbuf->page_base] = '\0';

which still has a possible off-by-one? (Was that why you have -1 -4?)

The check is different. 2.6.13 is using ">=" instead of ">", so hence I think that's fine.

Argh. Where's my reading glasses? ;) Yes, a >= check works there....

Review

Quoted 2 lines
(quoted diff) +2 -0
+	if (len > rcvbuf->page_len - sizeof(*strlen) - 1)
+		len = rcvbuf->page_len - sizeof(*strlen) - 1;

OK, looks saner to me, but Trond and Marcelo probably get to make the final decision ;)

#6 lore raw
In reply to Assar

Hi Assar,

Quoted 18 lines · Assar
Quoted 6 lines · [email protected] writes:
Quoted 2 lines

To my reading, the 2.6.13 code does not copy the 4 bytes of length to rcvbuf.

Hmm... it still does this:

	kaddr[len+rcvbuf->page_base] = '\0';

which still has a possible off-by-one? (Was that why you have -1 -4?)

The check is different. 2.6.13 is using ">=" instead of ">", so hence I think that's fine.

Quoted 2 lines

sizeof(actual_var) is even better, as that way it's clear what you're allowing space for.

diff -u linux-2.4.31.orig/fs/nfs/nfs2xdr.c linux-2.4.31/fs/nfs/nfs2xdr.c

fs/nfs/nfs2xdr.c line 575

Quoted 7 lines
fs/nfs/nfs2xdr.c +2 -2
@@ -571,8 +571,8 @@
 	strlen = (u32*)kmap(rcvbuf->pages[0]);
 	/* Convert length of symlink */
 	len = ntohl(*strlen);
-	if (len > rcvbuf->page_len)
-		len = rcvbuf->page_len;
+	if (len > rcvbuf->page_len - sizeof(*strlen) - 1)
+		len = rcvbuf->page_len - sizeof(*strlen) - 1;

So the problem is that the "len" variable encapsulated in (u32 *)rcvbuf->pages[0] does not account for its own length (4 bytes)?

If thats the reason, you don't need the "-1" there?

Someone with better understanding to ACK this would be nice. Trond?

Quoted 17 lines
fs/nfs/nfs2xdr.c +0 -0
 	*strlen = len;
 	/* NULL terminate the string we got */
 	string = (char *)(strlen + 1);

diff -u linux-2.4.31.orig/fs/nfs/nfs3xdr.c linux-2.4.31/fs/nfs/nfs3xdr.c

fs/nfs/nfs3xdr.c +2 -2
@@ -759,8 +759,8 @@
 	strlen = (u32*)kmap(rcvbuf->pages[0]);
 	/* Convert length of symlink */
 	len = ntohl(*strlen);
-	if (len > rcvbuf->page_len)
-		len = rcvbuf->page_len;
+	if (len > rcvbuf->page_len - sizeof(*strlen) - 1)
+		len = rcvbuf->page_len - sizeof(*strlen) - 1;
 	*strlen = len;
 	/* NULL terminate the string we got */
 	string = (char *)(strlen + 1);
author
#7 lore raw

Hi, Marcelo.

Quoted 6 lines · Marcelo Tosatti
Quoted 6 lines

diff -u linux-2.4.31.orig/fs/nfs/nfs2xdr.c linux-2.4.31/fs/nfs/nfs2xdr.c

fs/nfs/nfs2xdr.c +0 -0
@@ -571,8 +571,8 @@
 	strlen = (u32*)kmap(rcvbuf->pages[0]);
 	/* Convert length of symlink */
Quoted 8 lines

fs/nfs/nfs2xdr.c line 575

Quoted 5 lines
fs/nfs/nfs2xdr.c +2 -2
 	len = ntohl(*strlen);
-	if (len > rcvbuf->page_len)
-		len = rcvbuf->page_len;
+	if (len > rcvbuf->page_len - sizeof(*strlen) - 1)
+		len = rcvbuf->page_len - sizeof(*strlen) - 1;

So the problem is that the "len" variable encapsulated in (u32 *)rcvbuf->pages[0] does not account for its own length (4 bytes)?

That's one problem.

Quoted 1 line

If thats the reason, you don't need the "-1" there?

It also writes a 0 byte. I think it looks like this:

---- ------------ -
len string... 0

#8 lore raw
Quoted 18 lines · Assar

Hi, Marcelo.

Quoted 6 lines · Marcelo Tosatti
Quoted 6 lines

diff -u linux-2.4.31.orig/fs/nfs/nfs2xdr.c linux-2.4.31/fs/nfs/nfs2xdr.c

fs/nfs/nfs2xdr.c +0 -0
@@ -571,8 +571,8 @@
 	strlen = (u32*)kmap(rcvbuf->pages[0]);
 	/* Convert length of symlink */
Quoted 8 lines

fs/nfs/nfs2xdr.c line 575

Quoted 5 lines
fs/nfs/nfs2xdr.c +2 -2
 	len = ntohl(*strlen);
-	if (len > rcvbuf->page_len)
-		len = rcvbuf->page_len;
+	if (len > rcvbuf->page_len - sizeof(*strlen) - 1)
+		len = rcvbuf->page_len - sizeof(*strlen) - 1;

So the problem is that the "len" variable encapsulated in (u32 *)rcvbuf->pages[0] does not account for its own length (4 bytes)?

Quoted 8 lines

That's one problem.

Quoted 1 line

If thats the reason, you don't need the "-1" there?

It also writes a 0 byte. I think it looks like this:

---- ------------ -
len string... 0

If an overflow happens (len > rcvbuf->page_len) the last character will get truncated anyway, so there is no need for the "-1" AFAICS.

author
#9 lore raw
Quoted 11 lines · Marcelo Tosatti
Quoted 8 lines

That's one problem.

Quoted 1 line

If thats the reason, you don't need the "-1" there?

It also writes a 0 byte. I think it looks like this:

---- ------------ -
len string... 0

If an overflow happens (len > rcvbuf->page_len) the last character will get truncated anyway, so there is no need for the "-1" AFAICS.

I'm not sure I follow.

The code writes a 0 at rcvbuf->pages[0][sizeof(u32) + len], right? Doesn't that make the maximum allowed value of len should be 'rcvbuf->page_len - sizeof(u32) - 1' ?

#10 lore raw
Quoted 6 lines · Assar
Quoted 4 lines · Peter Staubach
Quoted 4 lines

Yes, but fs/nfs/nfs2xdr.c:nfs_xdr_readlinkres on 2.4.31 writes a 0 at the end of string after having received it, which is what started this thread. Look at the end of nfs_xdr_readlinkres.

Quoted 7 lines
Quoted 1 line

Yes, I know that. For C purposes, the string must be null terminated.

Then I'm sorry but I don't understand what your point was. Do you believe there's a bug in nfs_xdr_readlinkres and if so, how do you think it should work?

Yes, I think that there is a bug in the boundary checking. I think that:

       if (len > rcvbuf->page_len)

should be

       if (len >= rcvbuf->page_len - sizeof(u32) || len > 1024)

because the code puts the length in the first 4 bytes and then the contents of the symbolic link is stored in the rest of the page. The ">=" accounts for the null byte will be appended to the length. The additional check for 1024 is due to the NFS Version 2 protocol limiting the size of the contents of a symbolic link which can be returned to 1024 bytes.

Also, the NFS Version 3, nfs3_xdr_readlinkres, is broken in the same way and will need to be changed in the same fashion, except that the NFS Version 3 protocol does not place an arbitrary limit on the size of the contents of the symbolic which can be returned. The comparison against 1024 won't be needed here.

--

The 2.6 kernel code is also broken, but in a different, but once again, similar fashions.

ps

author
#11 lore raw
Quoted 7 lines · Peter Staubach

Yes, I think that there is a bug in the boundary checking. I think that:

        if (len > rcvbuf->page_len)

should be

        if (len >= rcvbuf->page_len - sizeof(u32) || len > 1024)

Thanks for your feedback. The patch to 2.4.31 that incorporates your suggsted changes is here:

fs/nfs/nfs2xdr.c +2 -2
diff -u linux-2.4.31.orig/fs/nfs/nfs2xdr.c linux-2.4.31/fs/nfs/nfs2xdr.c
@@ -571,8 +571,8 @@
 	strlen = (u32*)kmap(rcvbuf->pages[0]);
 	/* Convert length of symlink */
 	len = ntohl(*strlen);
-	if (len > rcvbuf->page_len)
-		len = rcvbuf->page_len;
+	if (len >= rcvbuf->page_len - sizeof(u32) || len > NFS2_MAXPATHLEN)
+		len = rcvbuf->page_len - sizeof(u32) - 1;
 	*strlen = len;
 	/* NULL terminate the string we got */
 	string = (char *)(strlen + 1);
fs/nfs/nfs3xdr.c +2 -2
diff -u linux-2.4.31.orig/fs/nfs/nfs3xdr.c linux-2.4.31/fs/nfs/nfs3xdr.c
@@ -759,8 +759,8 @@
 	strlen = (u32*)kmap(rcvbuf->pages[0]);
 	/* Convert length of symlink */
 	len = ntohl(*strlen);
-	if (len > rcvbuf->page_len)
-		len = rcvbuf->page_len;
+	if (len >= rcvbuf->page_len - sizeof(u32))
+		len = rcvbuf->page_len - sizeof(u32) - 1;
 	*strlen = len;
 	/* NULL terminate the string we got */
 	string = (char *)(strlen + 1);
Quoted 2 lines

The 2.6 kernel code is also broken, but in a different, but once again, similar fashions.

You mean for length > 1024 ?

fs/nfs/nfs2xdr.c +1 -1
diff -u linux-2.6.13.orig/fs/nfs/nfs2xdr.c linux-2.6.13/fs/nfs/nfs2xdr.c
@@ -557,7 +557,7 @@
 		return -nfs_stat_to_errno(status);
 	/* Convert length of symlink */
 	len = ntohl(*p++);
-	if (len >= rcvbuf->page_len || len <= 0) {
+	if (len >= rcvbuf->page_len || len <= 0 || len > NFS2_MAXPATHLEN) {
 		dprintk(KERN_WARNING "nfs: server returned giant symlink!\n");
 		return -ENAMETOOLONG;
 	}
#12 lore raw
Quoted 12 lines · Assar
Quoted 4 lines

The 2.6 kernel code is also broken, but in a different, but once again, similar fashions.

You mean for length > 1024 ?

diff -u linux-2.6.13.orig/fs/nfs/nfs2xdr.c linux-2.6.13/fs/nfs/nfs2xdr.c

fs/nfs/nfs2xdr.c line 563

Quoted 8 lines
fs/nfs/nfs2xdr.c +1 -1
@@ -557,7 +557,7 @@
 		return -nfs_stat_to_errno(status);
 	/* Convert length of symlink */
 	len = ntohl(*p++);
-	if (len >= rcvbuf->page_len || len <= 0) {
+	if (len >= rcvbuf->page_len || len <= 0 || len > NFS2_MAXPATHLEN) {
 		dprintk(KERN_WARNING "nfs: server returned giant symlink!\n");
 		return -ENAMETOOLONG;
 	}

This code appears to assume that rcvbuf->page_base is zero here, but then uses rcvbuf->page_base when calculating where to place the null byte. It seems to me that it should either use rcvbuf->page_base in both calculations or neither.

Thanx...

ps

author
#13 lore raw
Quoted 4 lines · Peter Staubach

This code appears to assume that rcvbuf->page_base is zero here, but then uses rcvbuf->page_base when calculating where to place the null byte. It seems to me that it should either use rcvbuf->page_base in both calculations or neither.

The meaning of page_len and page_base are not totally clear to me. Is it the case that the data starts at offset page_base and there is page_len bytes of it. If that's the case, I think the code is doing the right thing.

#14 lore raw
Quoted 10 lines · Assar
Quoted 6 lines · Peter Staubach

This code appears to assume that rcvbuf->page_base is zero here, but then uses rcvbuf->page_base when calculating where to place the null byte. It seems to me that it should either use rcvbuf->page_base in both calculations or neither.

Quoted 5 lines

The meaning of page_len and page_base are not totally clear to me. Is it the case that the data starts at offset page_base and there is page_len bytes of it. If that's the case, I think the code is doing the right thing.

I am not clear either, but I would think that kmap_atomic() would return a pointer to the beginning of the page. There is also an assumption that there is only one page. If page_base needs to be used to offset from the address returned by kmap_atomic(), then I wouldn't think that the current test is correct. I think that it needs to take page_base into account when checking for the boundary.

Thanx...

ps

#15 lore raw
In reply to Assar
Quoted 35 lines · Assar
Quoted 9 lines · Peter Staubach

Yes, I think that there is a bug in the boundary checking. I think that:

       if (len > rcvbuf->page_len)

should be

       if (len >= rcvbuf->page_len - sizeof(u32) || len > 1024)

Thanks for your feedback. The patch to 2.4.31 that incorporates your suggsted changes is here:

diff -u linux-2.4.31.orig/fs/nfs/nfs2xdr.c linux-2.4.31/fs/nfs/nfs2xdr.c

fs/nfs/nfs2xdr.c +2 -2
@@ -571,8 +571,8 @@
 	strlen = (u32*)kmap(rcvbuf->pages[0]);
 	/* Convert length of symlink */
 	len = ntohl(*strlen);
-	if (len > rcvbuf->page_len)
-		len = rcvbuf->page_len;
+	if (len >= rcvbuf->page_len - sizeof(u32) || len > NFS2_MAXPATHLEN)
+		len = rcvbuf->page_len - sizeof(u32) - 1;
 	*strlen = len;
 	/* NULL terminate the string we got */
 	string = (char *)(strlen + 1);

diff -u linux-2.4.31.orig/fs/nfs/nfs3xdr.c linux-2.4.31/fs/nfs/nfs3xdr.c

fs/nfs/nfs3xdr.c line 766

Quoted 10 lines
fs/nfs/nfs3xdr.c +2 -2
@@ -759,8 +759,8 @@
 	strlen = (u32*)kmap(rcvbuf->pages[0]);
 	/* Convert length of symlink */
 	len = ntohl(*strlen);
-	if (len > rcvbuf->page_len)
-		len = rcvbuf->page_len;
+	if (len >= rcvbuf->page_len - sizeof(u32))
+		len = rcvbuf->page_len - sizeof(u32) - 1;
 	*strlen = len;
 	/* NULL terminate the string we got */
 	string = (char *)(strlen + 1);

One other thing -- it doesn't seem particularly correct to me to just silently truncate the symbolic link contents. If the contents can not be handled correctly because they are too large, then some sort of error should be generated. Silently truncating could lead to interoperability issues when multiple clients handle the contents in different fashions, some truncating, some returning errors, and some handling the long returns.

Thanx...

ps

author
#16 lore raw
Quoted 2 lines · Peter Staubach

One other thing -- it doesn't seem particularly correct to me to just silently truncate the symbolic link contents.

Sure, and 2.6 indeed returns ENAMETOOLONG. I was just trying to close the problem and not change the functionality in 2.4. If the consensus is that we should change it to return an error, I can certainly cook up patches for that.

#17 lore raw
Quoted 8 lines · Assar
Quoted 4 lines · Peter Staubach

One other thing -- it doesn't seem particularly correct to me to just silently truncate the symbolic link contents.

Quoted 5 lines

Sure, and 2.6 indeed returns ENAMETOOLONG. I was just trying to close the problem and not change the functionality in 2.4. If the consensus is that we should change it to return an error, I can certainly cook up patches for that.

Understand. I would recommend that the 2.4 kernel be modified to return an error, since we are already modifying the area anyway.

Thanx...

ps

author
#18 lore raw
Quoted 2 lines · Peter Staubach

Understand. I would recommend that the 2.4 kernel be modified to return an error, since we are already modifying the area anyway.

Marcelo: what's your thought?

If you want to go that route, the patch below should do it.

fs/nfs/nfs2xdr.c +5 -2
diff -u linux-2.4.31.orig/fs/nfs/nfs2xdr.c linux-2.4.31/fs/nfs/nfs2xdr.c
@@ -571,8 +571,11 @@
 	strlen = (u32*)kmap(rcvbuf->pages[0]);
 	/* Convert length of symlink */
 	len = ntohl(*strlen);
-	if (len > rcvbuf->page_len)
-		len = rcvbuf->page_len;
+	if (len >= rcvbuf->page_len - sizeof(u32) || len > NFS2_MAXPATHLEN) {
+		printk(KERN_WARNING "NFS: server returned giant symlink!\n");
+		kunmap(rcvbuf->pages[0]);
+		return -ENAMETOOLONG;
+        }
 	*strlen = len;
 	/* NULL terminate the string we got */
 	string = (char *)(strlen + 1);
fs/nfs/nfs3xdr.c +5 -2
diff -u linux-2.4.31.orig/fs/nfs/nfs3xdr.c linux-2.4.31/fs/nfs/nfs3xdr.c
@@ -759,8 +759,11 @@
 	strlen = (u32*)kmap(rcvbuf->pages[0]);
 	/* Convert length of symlink */
 	len = ntohl(*strlen);
-	if (len > rcvbuf->page_len)
-		len = rcvbuf->page_len;
+	if (len >= rcvbuf->page_len - sizeof(u32)) {
+		printk(KERN_WARNING "NFS: server returned giant symlink!\n");
+		kunmap(rcvbuf->pages[0]);
+		return -ENAMETOOLONG;
+        }
 	*strlen = len;
 	/* NULL terminate the string we got */
 	string = (char *)(strlen + 1);
#19 lore raw
In reply to Assar
Quoted 11 lines · Assar
Quoted 3 lines

If thats the reason, you don't need the "-1" there?

It also writes a 0 byte. I think it looks like this:

---- ------------ -
len string... 0

-

NFS uses XDR to encode C strings. They are encoded as counted byte arrays and are _not_ null terminated. The space containing the string is rounded up to the next 4 byte boundary though and, usually, this space is zero filled. The number of bytes in the string is encoded as a big endian integer in the first four bytes.

Thanx...

ps

author
#20 lore raw
Quoted 6 lines · Peter Staubach

NFS uses XDR to encode C strings. They are encoded as counted byte arrays and are _not_ null terminated. The space containing the string is rounded up to the next 4 byte boundary though and, usually, this space is zero filled.
The number of bytes in the string is encoded as a big endian integer in the first four bytes.

Yes, but fs/nfs/nfs2xdr.c:nfs_xdr_readlinkres on 2.4.31 writes a 0 at the end of string after having received it, which is what started this thread. Look at the end of nfs_xdr_readlinkres.