[Qemu-devel] [PULL 0/5] Miscellaneous patches for 2019-05-22

[Qemu-devel] [PULL 0/5] Miscellaneous patches for 2019-05-22

[Markus Armbruster]

The following changes since commit a4f667b6714916683408b983cfe0a615a725775f:

  Merge remote-tracking branch 'remotes/cohuck/tags/s390x-20190521-3' into staging (2019-05-21 16:30:13 +0100)

are available in the Git repository at:

  git://repo.or.cz/qemu/armbru.git tags/pull-misc-2019-05-22

for you to fetch changes up to db3d11ee3f0cb851124830172f0a93c3d77a450a:

  cutils: Simplify how parse_uint() checks for whitespace (2019-05-22 15:00:04 +0200)

----------------------------------------------------------------
Miscellaneous patches for 2019-05-22

----------------------------------------------------------------
Markus Armbruster (5):
      qemu-bridge-helper: Fix misuse of isspace()
      tests/vhost-user-bridge: Fix misuse of isdigit()
      gdbstub: Reject invalid RLE repeat counts
      gdbstub: Fix misuse of isxdigit()
      cutils: Simplify how parse_uint() checks for whitespace

 gdbstub.c                 | 20 ++++++++++++--------
 qemu-bridge-helper.c      |  6 +++---
 tests/vhost-user-bridge.c |  3 ++-
 util/cutils.c             |  2 +-
 4 files changed, 18 insertions(+), 13 deletions(-)

-- 
2.17.2

[Qemu-devel] [PULL 1/5] qemu-bridge-helper: Fix misuse of isspace()

[Markus Armbruster]

parse_acl_file() passes char values to isspace().  Undefined behavior
when the value is negative.  Not a security issue, because the
characters come from trusted $prefix/etc/qemu/bridge.conf and the
files it includes.

Furthermore, isspace()'s locale-dependence means qemu-bridge-helper
uses the user's locale for parsing $prefix/etc/bridge.conf.  Feels
wrong.

Use g_ascii_isspace() instead.  This fixes the undefined behavior, and
makes parsing of $prefix/etc/bridge.conf locale-independent.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20190514180311.16028-2-armbru@redhat.com>
---
 qemu-bridge-helper.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/qemu-bridge-helper.c b/qemu-bridge-helper.c
index 5396fbfbb6..f9940deefd 100644
--- a/qemu-bridge-helper.c
+++ b/qemu-bridge-helper.c
@@ -75,7 +75,7 @@ static int parse_acl_file(const char *filename, ACLList *acl_list)
         char *ptr = line;
         char *cmd, *arg, *argend;
 
-        while (isspace(*ptr)) {
+        while (g_ascii_isspace(*ptr)) {
             ptr++;
         }
 
@@ -99,12 +99,12 @@ static int parse_acl_file(const char *filename, ACLList *acl_list)
 
         *arg = 0;
         arg++;
-        while (isspace(*arg)) {
+        while (g_ascii_isspace(*arg)) {
             arg++;
         }
 
         argend = arg + strlen(arg);
-        while (arg != argend && isspace(*(argend - 1))) {
+        while (arg != argend && g_ascii_isspace(*(argend - 1))) {
             argend--;
         }
         *argend = 0;
-- 
2.17.2

Re: [Qemu-devel] [PULL 1/5] qemu-bridge-helper: Fix misuse of isspace()

[Peter Maydell]

On Wed, 22 May 2019 at 14:49, Markus Armbruster <armbru@redhat.com> wrote:
>
> parse_acl_file() passes char values to isspace().  Undefined behavior
> when the value is negative.  Not a security issue, because the
> characters come from trusted $prefix/etc/qemu/bridge.conf and the
> files it includes.
>
> Furthermore, isspace()'s locale-dependence means qemu-bridge-helper
> uses the user's locale for parsing $prefix/etc/bridge.conf.  Feels
> wrong.
>
> Use g_ascii_isspace() instead.  This fixes the undefined behavior, and
> makes parsing of $prefix/etc/bridge.conf locale-independent.
>
> Signed-off-by: Markus Armbruster <armbru@redhat.com>
> Message-Id: <20190514180311.16028-2-armbru@redhat.com>
> ---
>  qemu-bridge-helper.c | 6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)

Coverity complains about this change (CID 1401706) because
it doesn't have enough information to know that the table
lookup g_ascii_isspace does is always safe:

  tainted_data: Using tainted variable
   (guchar)*(argend - 1) as an index to pointer g_ascii_table.

We know this is OK because we know the table is big enough that
a guchar index can't possibly overrun, but because the table
is declared in the glib header file as
  GLIB_VAR const guint16 * const g_ascii_table;
Coverity has no idea of its size and is being pessimistic.

I've squashed the Coverity issue as a false-positive, but I
mention it here in case you thought it worth trying to write
something in coverity-model.c to provide a better model of
the glib function.

thanks
-- PMM

[Qemu-devel] [PULL 2/5] tests/vhost-user-bridge: Fix misuse of isdigit()

[Markus Armbruster]

vubr_set_host() passes char values to isdigit().  Undefined behavior
when the value is negative.

Fix by using qemu_isdigit() instead.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20190514180311.16028-3-armbru@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
[Missing #include "qemu-common.h" fixed]
---
 tests/vhost-user-bridge.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/tests/vhost-user-bridge.c b/tests/vhost-user-bridge.c
index 0033b61f2e..5b771de7a3 100644
--- a/tests/vhost-user-bridge.c
+++ b/tests/vhost-user-bridge.c
@@ -30,6 +30,7 @@
 #define _FILE_OFFSET_BITS 64
 
 #include "qemu/osdep.h"
+#include "qemu-common.h"
 #include "qemu/atomic.h"
 #include "qemu/iov.h"
 #include "standard-headers/linux/virtio_net.h"
@@ -645,7 +646,7 @@ vubr_host_notifier_setup(VubrDev *dev)
 static void
 vubr_set_host(struct sockaddr_in *saddr, const char *host)
 {
-    if (isdigit(host[0])) {
+    if (qemu_isdigit(host[0])) {
         if (!inet_aton(host, &saddr->sin_addr)) {
             fprintf(stderr, "inet_aton() failed.\n");
             exit(1);
-- 
2.17.2

[Qemu-devel] [PULL 3/5] gdbstub: Reject invalid RLE repeat counts

[Markus Armbruster]

"Debugging with GDB / Appendix E GDB Remote Serial Protocol /
Overview" specifies "The printable characters '#' and '$' or with a
numeric value greater than 126 must not be used."  gdb_read_byte()
only rejects values < 32.  This is wrong.  Impact depends on the caller:

* gdb_handlesig() passes a char.  Incorrectly accepts '#', '$' and
  '\127'.

* gdb_chr_receive() passes an uint8_t.  Additionally accepts
  characters with the most-significant bit set.

Correct the validity check to match the specification.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20190514180311.16028-4-armbru@redhat.com>
---
 gdbstub.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/gdbstub.c b/gdbstub.c
index d54abd17cc..c41eb1de07 100644
--- a/gdbstub.c
+++ b/gdbstub.c
@@ -2064,7 +2064,11 @@ static void gdb_read_byte(GDBState *s, int ch)
             }
             break;
         case RS_GETLINE_RLE:
-            if (ch < ' ') {
+            /*
+             * Run-length encoding is explained in "Debugging with GDB /
+             * Appendix E GDB Remote Serial Protocol / Overview".
+             */
+            if (ch < ' ' || ch == '#' || ch == '$' || ch > 126) {
                 /* invalid RLE count encoding */
                 trace_gdbstub_err_invalid_repeat((uint8_t)ch);
                 s->state = RS_GETLINE;
-- 
2.17.2

[Qemu-devel] [PULL 4/5] gdbstub: Fix misuse of isxdigit()

[Markus Armbruster]

gdb_read_byte() passes its @ch argument to isxdigit().  Undefined
behavior when the value is negative.  Two callers:

* gdb_chr_receive() passes an uint8_t value.  Safe.

* gdb_handlesig() a char value.  Unsafe.  Not a security issue,
  because the characters come from the gdb client, which is trusted.

The obvious fix would be casting @ch to unsigned char.  But note that
gdb_read_byte() already casts @ch to uint8_t in many places.  Uses of
@ch without such a cast:

(1) Compare to a character constant with == or !=

(2) s->linesum += ch

(3) Store ch or ch ^ 0x20 into s->line_buf[]

(4) Check for invalid RLE count:
    ch < ' ' || ch == '#' || ch == '$' || ch > 126

(5) Pass to isxdigit()

(6) Pass to fromhex()

Change the parameter type from int to uint8_t, and drop the now
redundant casts.  Affects the above uses as follows:

(1) No change: the character constants are all non-negative.

(2) Effectively no change: we only ever use s->linesum & 0xff, and
    s->linesum is int.

(3) No change: s->line_buf[] is char[].

(4) No change.

(5) Avoid undefined behavior.

(6) No change: only reached when isxdigit(ch)

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20190514180311.16028-5-armbru@redhat.com>
---
 gdbstub.c | 14 +++++++-------
 1 file changed, 7 insertions(+), 7 deletions(-)

diff --git a/gdbstub.c b/gdbstub.c
index c41eb1de07..b129df4e59 100644
--- a/gdbstub.c
+++ b/gdbstub.c
@@ -1987,7 +1987,7 @@ void gdb_do_syscall(gdb_syscall_complete_cb cb, const char *fmt, ...)
     va_end(va);
 }
 
-static void gdb_read_byte(GDBState *s, int ch)
+static void gdb_read_byte(GDBState *s, uint8_t ch)
 {
     uint8_t reply;
 
@@ -2001,7 +2001,7 @@ static void gdb_read_byte(GDBState *s, int ch)
         } else if (ch == '+') {
             trace_gdbstub_io_got_ack();
         } else {
-            trace_gdbstub_io_got_unexpected((uint8_t)ch);
+            trace_gdbstub_io_got_unexpected(ch);
         }
 
         if (ch == '+' || ch == '$')
@@ -2024,7 +2024,7 @@ static void gdb_read_byte(GDBState *s, int ch)
                 s->line_sum = 0;
                 s->state = RS_GETLINE;
             } else {
-                trace_gdbstub_err_garbage((uint8_t)ch);
+                trace_gdbstub_err_garbage(ch);
             }
             break;
         case RS_GETLINE:
@@ -2070,11 +2070,11 @@ static void gdb_read_byte(GDBState *s, int ch)
              */
             if (ch < ' ' || ch == '#' || ch == '$' || ch > 126) {
                 /* invalid RLE count encoding */
-                trace_gdbstub_err_invalid_repeat((uint8_t)ch);
+                trace_gdbstub_err_invalid_repeat(ch);
                 s->state = RS_GETLINE;
             } else {
                 /* decode repeat length */
-                int repeat = (unsigned char)ch - ' ' + 3;
+                int repeat = ch - ' ' + 3;
                 if (s->line_buf_index + repeat >= sizeof(s->line_buf) - 1) {
                     /* that many repeats would overrun the command buffer */
                     trace_gdbstub_err_overrun();
@@ -2096,7 +2096,7 @@ static void gdb_read_byte(GDBState *s, int ch)
         case RS_CHKSUM1:
             /* get high hex digit of checksum */
             if (!isxdigit(ch)) {
-                trace_gdbstub_err_checksum_invalid((uint8_t)ch);
+                trace_gdbstub_err_checksum_invalid(ch);
                 s->state = RS_GETLINE;
                 break;
             }
@@ -2107,7 +2107,7 @@ static void gdb_read_byte(GDBState *s, int ch)
         case RS_CHKSUM2:
             /* get low hex digit of checksum */
             if (!isxdigit(ch)) {
-                trace_gdbstub_err_checksum_invalid((uint8_t)ch);
+                trace_gdbstub_err_checksum_invalid(ch);
                 s->state = RS_GETLINE;
                 break;
             }
-- 
2.17.2

[Qemu-devel] [PULL 5/5] cutils: Simplify how parse_uint() checks for whitespace

[Markus Armbruster]

Use qemu_isspace() so we don't have to cast to unsigned char.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20190514180311.16028-7-armbru@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
---
 util/cutils.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/util/cutils.c b/util/cutils.c
index d682c90901..9aacc422ca 100644
--- a/util/cutils.c
+++ b/util/cutils.c
@@ -683,7 +683,7 @@ int parse_uint(const char *s, unsigned long long *value, char **endptr,
     }
 
     /* make sure we reject negative numbers: */
-    while (isspace((unsigned char)*s)) {
+    while (qemu_isspace(*s)) {
         s++;
     }
     if (*s == '-') {
-- 
2.17.2

Re: [Qemu-devel] [PULL 0/5] Miscellaneous patches for 2019-05-22

[Peter Maydell]

On Wed, 22 May 2019 at 14:51, Markus Armbruster <armbru@redhat.com> wrote:
>
> The following changes since commit a4f667b6714916683408b983cfe0a615a725775f:
>
>   Merge remote-tracking branch 'remotes/cohuck/tags/s390x-20190521-3' into staging (2019-05-21 16:30:13 +0100)
>
> are available in the Git repository at:
>
>   git://repo.or.cz/qemu/armbru.git tags/pull-misc-2019-05-22
>
> for you to fetch changes up to db3d11ee3f0cb851124830172f0a93c3d77a450a:
>
>   cutils: Simplify how parse_uint() checks for whitespace (2019-05-22 15:00:04 +0200)
>
> ----------------------------------------------------------------
> Miscellaneous patches for 2019-05-22
>
> ----------------------------------------------------------------
> Markus Armbruster (5):
>       qemu-bridge-helper: Fix misuse of isspace()
>       tests/vhost-user-bridge: Fix misuse of isdigit()
>       gdbstub: Reject invalid RLE repeat counts
>       gdbstub: Fix misuse of isxdigit()
>       cutils: Simplify how parse_uint() checks for whitespace
>
>  gdbstub.c                 | 20 ++++++++++++--------
>  qemu-bridge-helper.c      |  6 +++---
>  tests/vhost-user-bridge.c |  3 ++-
>  util/cutils.c             |  2 +-
>  4 files changed, 18 insertions(+), 13 deletions(-)


Applied, thanks.

Please update the changelog at https://wiki.qemu.org/ChangeLog/4.1
for any user-visible changes.

-- PMM