Skip to contentSkip to conversation

kexec: fix mmap return code handling

PATCH Superseded Michael Holzheu 3 participants 5 messages Nov 25, 2015 – Nov 27, 2015

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 kexec: fix mmap return code handling

Hi Simon,

I made a mistake in my mmap patch, sorry for that!

When mmap fails, MAP_FAILED (that is, (void *) -1) is returned. Currently we assume that NULL is returned. Fix this and add the MAP_FAILED check.

Fixes 95741713e790 ("kexec/s390x: use mmap instead of read for slurp_file")
---
 kexec/kexec.c | 2 ++
 1 file changed, 2 insertions(+)
kexec/kexec.c +2 -0
index cf6e03d..02285fb 100644
@@ -568,6 +568,8 @@ static char *slurp_file_generic(const char *filename, off_t *r_size,
 		if (use_mmap) {
 			buf = mmap(NULL, size, PROT_READ|PROT_WRITE,
 				   MAP_PRIVATE, fd, 0);
+			if (buf == MAP_FAILED)
+				buf = NULL;
 			nread = size;
 		} else {
 			buf = slurp_fd(fd, filename, size, &nread);
Signature · 7 lines
2.3.9


_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
#1 lore raw

PATCH v2 1/1 kexec: fix mmap return code handling

Hi Simon again,

After a bit more thinking: In theory mmap() could also return NULL. Therefore the following fix is probably the better one ...
---
Subject: [PATCH] kexec: fix mmap return code handling

When mmap fails, MAP_FAILED (that is, (void *) -1) is returned. Currently we assume that NULL is returned. Fix this and add the MAP_FAILED check.

Fixes 95741713e790 ("kexec/s390x: use mmap instead of read for slurp_file")
kexec/kexec.c +1 -1
index cf6e03d..f0bd527 100644
@@ -573,7 +573,7 @@ static char *slurp_file_generic(const char *filename, off_t *r_size,
 			buf = slurp_fd(fd, filename, size, &nread);
 		}
 	}
-	if (!buf)
+	if ((use_mmap && (buf == MAP_FAILED)) || (!use_mmap && (buf == NULL)))
 		die("Cannot read %s", filename);
 
 	if (nread != size)
Signature · 7 lines
2.3.9


_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
#2 lore raw
Quoted 4 lines · On Thu, 26 Nov 2015 18:32:31 +0100 Michael Holzheu…

Hi Simon again,

After a bit more thinking: In theory mmap() could also return NULL. Therefore the following fix is probably the better one ...

No, if you let the kernel choose the address (i.e. call mmap with NULL addr), it will return at least PAGE_SIZE (and a higher limit is usually enforced by sys.vm.mmap_min_addr sysctl). Admittedly the limit is set in arch-specific code, so theoretically there can be an architecture which sets the limit to 0, but I doubt it, because it would break too many assumptions in user space (for example gcc assumes that dereferencing a NULL pointer terminates a process).

In short, this other fix is just as good as the previous one.

Regards,
Petr Tesarik

Quoted 14 lines

---
Subject: [PATCH] kexec: fix mmap return code handling

When mmap fails, MAP_FAILED (that is, (void *) -1) is returned. Currently we assume that NULL is returned. Fix this and add the MAP_FAILED check.

Fixes 95741713e790 ("kexec/s390x: use mmap instead of read for slurp_file")
kexec/kexec.c +0 -0
index cf6e03d..f0bd527 100644

kexec/kexec.c static char *slurp_file_generic(const char *filename, off_t *r_size, · line 579

Quoted 8 lines
kexec/kexec.c +1 -1
@@ -573,7 +573,7 @@ static char *slurp_file_generic(const char *filename, off_t *r_size,
 			buf = slurp_fd(fd, filename, size, &nread);
 		}
 	}
-	if (!buf)
+	if ((use_mmap && (buf == MAP_FAILED)) || (!use_mmap && (buf == NULL)))
 		die("Cannot read %s", filename);
 
 	if (nread != size)

_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec

#3 lore raw
Quoted 16 lines · On Thu, 26 Nov 2015 19:02:28 +0100 Petr Tesarik <[email protected]> wrote:
Quoted 4 lines · On Thu, 26 Nov 2015 18:32:31 +0100 Michael Holzheu…

Hi Simon again,

After a bit more thinking: In theory mmap() could also return NULL. Therefore the following fix is probably the better one ...

No, if you let the kernel choose the address (i.e. call mmap with NULL addr), it will return at least PAGE_SIZE (and a higher limit is usually enforced by sys.vm.mmap_min_addr sysctl). Admittedly the limit is set in arch-specific code, so theoretically there can be an architecture which sets the limit to 0, but I doubt it, because it would break too many assumptions in user space (for example gcc assumes that dereferencing a NULL pointer terminates a process).

Quoted 1 line

In short, this other fix is just as good as the previous one.

Hi Petr,

Thanks for clarification! I still would vote for the second one ;-)

Michael

_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec

#4 lore raw
Quoted 19 lines · Michael Holzheu
Quoted 16 lines · On Thu, 26 Nov 2015 19:02:28 +0100 Petr Tesarik <[email protected]> wrote:
Quoted 4 lines · On Thu, 26 Nov 2015 18:32:31 +0100 Michael Holzheu…

Hi Simon again,

After a bit more thinking: In theory mmap() could also return NULL. Therefore the following fix is probably the better one ...

No, if you let the kernel choose the address (i.e. call mmap with NULL addr), it will return at least PAGE_SIZE (and a higher limit is usually enforced by sys.vm.mmap_min_addr sysctl). Admittedly the limit is set in arch-specific code, so theoretically there can be an architecture which sets the limit to 0, but I doubt it, because it would break too many assumptions in user space (for example gcc assumes that dereferencing a NULL pointer terminates a process).

Quoted 5 lines
Quoted 1 line

In short, this other fix is just as good as the previous one.

Hi Petr,

Thanks for clarification! I still would vote for the second one ;-)

Thanks, I have applied that version.

_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec