[PATCH 0/4] Miscellaneous fixes

[PATCH 0/4] Miscellaneous fixes

[Andrew Murray]

This series provides miscellaneous fixes and removes some unintended
debug.

Andrew Murray (4):
  smdb.py: fix handling of capped ranges
  smdb.py: fix param tracking now that we use $ instead of p
  user_data: Recognize uaccess for non-x86 architectures
  kchecker: remove debug

 check_get_user_overflow.c | 2 +-
 smatch_data/db/smdb.py    | 6 +++---
 smatch_kernel_user_data.c | 2 +-
 smatch_scripts/kchecker   | 1 -
 4 files changed, 5 insertions(+), 6 deletions(-)

-- 
2.21.0

[PATCH 1/4] smdb.py: fix handling of capped ranges

[Andrew Murray]

The get_next_str function of smdb.py does not understand ranges that
have flags in square brackets. This results in a range such as
"0,4096-18446744073709547520[c]" being interpreted as 0->0 and 4096->0.

Let's fix this by breaking when we reach an opening square bracket.

Signed-off-by: Andrew Murray <andrew.murray@arm.com>
---
 smatch_data/db/smdb.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/smatch_data/db/smdb.py b/smatch_data/db/smdb.py
index c0b310b13b1d..1e089b4804bd 100755
--- a/smatch_data/db/smdb.py
+++ b/smatch_data/db/smdb.py
@@ -267,7 +267,7 @@ def get_next_str(txt):
         if txt[0] == '-':
             parsed += 1
         for char in txt[parsed:]:
-            if char == '-':
+            if char == '-' or char == '[':
                 break
             parsed += 1
         val = txt[:parsed]
-- 
2.21.0

[PATCH 2/4] smdb.py: fix param tracking now that we use $ instead of p

[Andrew Murray]

Since d8084fee2b7c the data source of a parameter is now stored with the
notation: $0 instead of p0 (where the $ or p indicates that the data source
of the parameter is from the callers function parameters, the value next to
it is the parameter number).

Let's update the smdb script to reflect this.

Fixes: d8084fee2b7c ("data_source: change how parameter data source information is recorded")
Signed-off-by: Andrew Murray <andrew.murray@arm.com>
---
 smatch_data/db/smdb.py | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/smatch_data/db/smdb.py b/smatch_data/db/smdb.py
index 1e089b4804bd..9025a95070b5 100755
--- a/smatch_data/db/smdb.py
+++ b/smatch_data/db/smdb.py
@@ -574,8 +574,8 @@ def trace_param_helper(func, param, indent = 0):
     sources = trace_callers(func, param)
     for path in sources:
 
-        if len(path[1]) and path[1][0] == 'p' and path[1][1] == ' ':
-            p = int(path[1][2:])
+        if len(path[1]) and path[1][0] == '$':
+            p = int(re.findall('\d+', path[1][1:])[0])
             trace_param_helper(path[0], p, indent + 2)
         elif len(path[0]) and path[0][0] == '%':
             print "  %s%s" %(" " * indent, path[1])
-- 
2.21.0

[PATCH 3/4] user_data: Recognize uaccess for non-x86 architectures

[Andrew Murray]

We identify some user data by recognizing both use of the get_user
macro and assignment from __val_gu. However on most non-x86
architectures the get_user macro uses the variable __gu_val instead
of __val_gu.

Let's update handle_get_user to recognize that get_user may also
use __gu_val.

Signed-off-by: Andrew Murray <andrew.murray@arm.com>
---
 check_get_user_overflow.c | 2 +-
 smatch_kernel_user_data.c | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/check_get_user_overflow.c b/check_get_user_overflow.c
index 20ec952c885a..fed299ee2300 100644
--- a/check_get_user_overflow.c
+++ b/check_get_user_overflow.c
@@ -108,7 +108,7 @@ static void match_assign(struct expression *expr)
 		return;
 	}
 	name = expr_to_var(expr->right);
-	if (!name || strcmp(name, "__val_gu") != 0)
+	if (!name || (strcmp(name, "__val_gu") != 0 && strcmp(name, "__gu_val")))
 		goto free;
 	set_state_expr(my_max_id, expr->left, &user_data);
 	set_state_expr(my_min_id, expr->left, &user_data);
diff --git a/smatch_kernel_user_data.c b/smatch_kernel_user_data.c
index bc9298fd6312..7c5fb8d7d872 100644
--- a/smatch_kernel_user_data.c
+++ b/smatch_kernel_user_data.c
@@ -542,7 +542,7 @@ static int handle_get_user(struct expression *expr)
 		return 0;
 
 	name = expr_to_var(expr->right);
-	if (!name || strcmp(name, "__val_gu") != 0)
+	if (!name || (strcmp(name, "__val_gu") != 0 && strcmp(name, "__gu_val") != 0))
 		goto free;
 	set_state_expr(my_id, expr->left, alloc_estate_whole(get_type(expr->left)));
 	ret = 1;
-- 
2.21.0

[PATCH 4/4] kchecker: remove debug

[Andrew Murray]

Remove unnecessary debug from kchecker script when cross-building.

Signed-off-by: Andrew Murray <andrew.murray@arm.com>
---
 smatch_scripts/kchecker | 1 -
 1 file changed, 1 deletion(-)

diff --git a/smatch_scripts/kchecker b/smatch_scripts/kchecker
index 2ca9b0f00f49..229250b4da8b 100755
--- a/smatch_scripts/kchecker
+++ b/smatch_scripts/kchecker
@@ -71,7 +71,6 @@ if  echo $oname | grep -q .o$ ; then
 fi
 
 if [[ ! -z $ARCH ]]; then
-	echo "yes"
 	KERNEL_ARCH="ARCH=$ARCH"
 fi
 if [[ ! -z $CROSS_COMPILE ]] ; then
-- 
2.21.0

Re: [PATCH 0/4] Miscellaneous fixes

[Dan Carpenter]

I'm back home now.  I've applied this series.  Thanks!

I don't know if my phone emails made any sense at all.  Let me know if
you need anything else.

regards,
dan carpenter

Re: [PATCH 0/4] Miscellaneous fixes

[Andrew Murray]

On Wed, Jul 17, 2019 at 01:18:14PM +0300, Dan Carpenter wrote:
> I'm back home now.  I've applied this series.  Thanks!

Thanks for this.

> 
> I don't know if my phone emails made any sense at all.  Let me know if
> you need anything else.

They do, thanks for the responses, apologies for my slow reply.

I'll share my progress on the list shortly.

Thanks,

Andrew Murray

> 
> regards,
> dan carpenter
>