Skip to contentSkip to conversation

libfdt: fix undefined behaviour in fdt_splice_()

PATCH Applied Jan Beulich 2 participants 4 messages Mar 5, 2020 – Mar 13, 2020 3 messages not in the archive

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 libfdt: fix undefined behaviour in fdt_splice_()

libfdt: fix undefined behaviour in fdt_splice_()

Along the lines of commit d0b3ab0a0f46 ("libfdt: Fix undefined behaviour in fdt_offset_ptr()"), fdt_splice_() similarly may not use pointer arithmetic to do overflow checks. (The left side of the checks added by d4c7c25c9ed1 ["libfdt: check for potential overrun in _fdt_splice()"] doesn't really lend itself to similar replacement though.)

---
The right side of the checks added by d4c7c25c9ed1 looks suspicious, which is more noticable after the transformation done here:

end - oldlen + newlen < (char *)fdt

end - (char *)fdt + newlen < oldlen

dsize + newlen < oldlen

libfdt/fdt_rw.c +7 -6
@@ -46,7 +46,7 @@ static int fdt_rw_probe_(void *fdt)
 			return err_; \
 	}
 
-static inline int fdt_data_size_(void *fdt)
+static inline unsigned int fdt_data_size_(void *fdt)
 {
 	return fdt_off_dt_strings(fdt) + fdt_size_dt_strings(fdt);
 }
@@ -54,15 +54,16 @@ static inline int fdt_data_size_(void *f
 static int fdt_splice_(void *fdt, void *splicepoint, int oldlen, int newlen)
 {
 	char *p = splicepoint;
-	char *end = (char *)fdt + fdt_data_size_(fdt);
+	unsigned int dsize = fdt_data_size_(fdt);
+	size_t soff = p - (char *)fdt;
 
-	if (((p + oldlen) < p) || ((p + oldlen) > end))
+	if ((oldlen < 0) || (soff + oldlen < soff) || (soff + oldlen > dsize))
 		return -FDT_ERR_BADOFFSET;
-	if ((p < (char *)fdt) || ((end - oldlen + newlen) < (char *)fdt))
+	if ((p < (char *)fdt) || (dsize + newlen < oldlen))
 		return -FDT_ERR_BADOFFSET;
-	if ((end - oldlen + newlen) > ((char *)fdt + fdt_totalsize(fdt)))
+	if (dsize - oldlen + newlen > fdt_totalsize(fdt))
 		return -FDT_ERR_NOSPACE;
-	memmove(p + newlen, p + oldlen, end - p - oldlen);
+	memmove(p + newlen, p + oldlen, ((char *)fdt + dsize) - (p + oldlen));
 	return 0;
 }
 

A message that is not in the archive (sent off-list or not yet received): f2d09e81-7cb8-c5cc-9699-1ac05b0626ff-IBi9RG/[email protected]

Applied
#2 lore raw
Quoted 9 lines · Jan Beulich

libfdt: fix undefined behaviour in fdt_splice_()

Along the lines of commit d0b3ab0a0f46 ("libfdt: Fix undefined behaviour in fdt_offset_ptr()"), fdt_splice_() similarly may not use pointer arithmetic to do overflow checks. (The left side of the checks added by d4c7c25c9ed1 ["libfdt: check for potential overrun in _fdt_splice()"] doesn't really lend itself to similar replacement though.)

Applied, thanks.

Quoted 9 lines

---
The right side of the checks added by d4c7c25c9ed1 looks suspicious, which is more noticable after the transformation done here:

end - oldlen + newlen < (char *)fdt

end - (char *)fdt + newlen < oldlen

dsize + newlen < oldlen

Hrm, yeah, I think this is trying to check for overflow of dsize+newlen, but it's pretty hard to follow.

Quoted 35 lines
libfdt/fdt_rw.c +7 -6
@@ -46,7 +46,7 @@ static int fdt_rw_probe_(void *fdt)
 			return err_; \
 	}
 
-static inline int fdt_data_size_(void *fdt)
+static inline unsigned int fdt_data_size_(void *fdt)
 {
 	return fdt_off_dt_strings(fdt) + fdt_size_dt_strings(fdt);
 }
@@ -54,15 +54,16 @@ static inline int fdt_data_size_(void *f
 static int fdt_splice_(void *fdt, void *splicepoint, int oldlen, int newlen)
 {
 	char *p = splicepoint;
-	char *end = (char *)fdt + fdt_data_size_(fdt);
+	unsigned int dsize = fdt_data_size_(fdt);
+	size_t soff = p - (char *)fdt;
 
-	if (((p + oldlen) < p) || ((p + oldlen) > end))
+	if ((oldlen < 0) || (soff + oldlen < soff) || (soff + oldlen > dsize))
 		return -FDT_ERR_BADOFFSET;
-	if ((p < (char *)fdt) || ((end - oldlen + newlen) < (char *)fdt))
+	if ((p < (char *)fdt) || (dsize + newlen < oldlen))
 		return -FDT_ERR_BADOFFSET;
-	if ((end - oldlen + newlen) > ((char *)fdt + fdt_totalsize(fdt)))
+	if (dsize - oldlen + newlen > fdt_totalsize(fdt))
 		return -FDT_ERR_NOSPACE;
-	memmove(p + newlen, p + oldlen, end - p - oldlen);
+	memmove(p + newlen, p + oldlen, ((char *)fdt + dsize) - (p + oldlen));
 	return 0;
 }
 
Signature · 4 lines
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson

A message that is not in the archive (sent off-list or not yet received): [email protected]

#4 lore raw

David,

Quoted 9 lines · David Gibson
Quoted 8 lines · Jan Beulich

libfdt: fix undefined behaviour in fdt_splice_()

Along the lines of commit d0b3ab0a0f46 ("libfdt: Fix undefined behaviour in fdt_offset_ptr()"), fdt_splice_() similarly may not use pointer arithmetic to do overflow checks. (The left side of the checks added by d4c7c25c9ed1 ["libfdt: check for potential overrun in _fdt_splice()"] doesn't really lend itself to similar replacement though.)

Quoted 3 lines
Quoted 1 line

Applied, thanks.

would you mind pointing me at the tree that you applied this to? From looking around I got the impression that

https://git.kernel.org/pub/scm/utils/dtc/dtc.git

would be the master tree, but I can't see any commit there newer than from 8 days ago. The reason I'm asking is because I found the issue originally in the Xen tree, which had imported libfdt from the DTC project, and in order to submit the (backported) fix there I'd like to be able to name the master tree's commit hash.

Thanks, Jan

A message that is not in the archive (sent off-list or not yet received): 4eaa4349-edbf-eb53-637d-b62a20befddd-IBi9RG/[email protected]

#6 lore raw
Quoted 19 lines · Jan Beulich

David,

Quoted 9 lines · David Gibson
Quoted 8 lines · Jan Beulich

libfdt: fix undefined behaviour in fdt_splice_()

Along the lines of commit d0b3ab0a0f46 ("libfdt: Fix undefined behaviour in fdt_offset_ptr()"), fdt_splice_() similarly may not use pointer arithmetic to do overflow checks. (The left side of the checks added by d4c7c25c9ed1 ["libfdt: check for potential overrun in _fdt_splice()"] doesn't really lend itself to similar replacement though.)

Quoted 3 lines
Quoted 1 line

Applied, thanks.

would you mind pointing me at the tree that you applied this to?

Quoted 1 line

From looking around I got the impression that

Quoted 8 lines

https://git.kernel.org/pub/scm/utils/dtc/dtc.git

would be the master tree, but I can't see any commit there newer than from 8 days ago. The reason I'm asking is because I found the issue originally in the Xen tree, which had imported libfdt from the DTC project, and in order to submit the (backported) fix there I'd like to be able to name the master tree's commit hash.

Oh, sorry, I just forgot to push it out from my local tree. It should be there now.

Signature · 4 lines
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson