diff mbox series

[3/5] fs/fs_parse: Fix 3 issues for validate_constant_table()

Message ID 20250410-fix_fs-v1-3-7c14ccc8ebaa@quicinc.com (mailing list archive)
State New
Headers show
Series fs: bug fixes | expand

Commit Message

Zijun Hu April 10, 2025, 11:45 a.m. UTC 
From: Zijun Hu <quic_zijuhu@quicinc.com>

Constant table array array[] which must end with a empty entry and fix
below issues for validate_constant_table(array, ARRAY_SIZE(array), ...):

- Always return wrong value for good constant table array which ends
  with a empty entry.

- Imprecise error message for missorted case.

- Potential NULL pointer dereference.

Fixes: 31d921c7fb96 ("vfs: Add configuration parser helpers")
Signed-off-by: Zijun Hu <quic_zijuhu@quicinc.com>
---
 fs/fs_parser.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

Comments

Christian Brauner April 11, 2025, 2:37 p.m. UTC | #1 
On Thu, Apr 10, 2025 at 07:45:29PM +0800, Zijun Hu wrote:
> From: Zijun Hu <quic_zijuhu@quicinc.com>
> 
> Constant table array array[] which must end with a empty entry and fix
> below issues for validate_constant_table(array, ARRAY_SIZE(array), ...):
> 
> - Always return wrong value for good constant table array which ends
>   with a empty entry.
> 
> - Imprecise error message for missorted case.
> 
> - Potential NULL pointer dereference.

I really dislike "potential NULL deref" without an explanation. Please
explain how this supposed NULL deref can happen.

> Fixes: 31d921c7fb96 ("vfs: Add configuration parser helpers")
> Signed-off-by: Zijun Hu <quic_zijuhu@quicinc.com>
> ---
>  fs/fs_parser.c | 7 +++++--
>  1 file changed, 5 insertions(+), 2 deletions(-)
> 
> diff --git a/fs/fs_parser.c b/fs/fs_parser.c
> index e635a81e17d965df78ffef27f6885cd70996c6dd..ef7876340a917876bc40df9cdde9232204125a75 100644
> --- a/fs/fs_parser.c
> +++ b/fs/fs_parser.c
> @@ -399,6 +399,9 @@ bool validate_constant_table(const struct constant_table *tbl, size_t tbl_size,
>  	}
>  
>  	for (i = 0; i < tbl_size; i++) {
> +		if (!tbl[i].name && (i + 1 == tbl_size))
> +			break;
> +
>  		if (!tbl[i].name) {
>  			pr_err("VALIDATE C-TBL[%zu]: Null\n", i);
>  			good = false;
> @@ -411,13 +414,13 @@ bool validate_constant_table(const struct constant_table *tbl, size_t tbl_size,
>  				good = false;
>  			}
>  			if (c > 0) {
> -				pr_err("VALIDATE C-TBL[%zu]: Missorted %s>=%s\n",
> +				pr_err("VALIDATE C-TBL[%zu]: Missorted %s>%s\n",
>  				       i, tbl[i-1].name, tbl[i].name);
>  				good = false;
>  			}
>  		}
>  
> -		if (tbl[i].value != special &&
> +		if (tbl[i].name && tbl[i].value != special &&
>  		    (tbl[i].value < low || tbl[i].value > high)) {
>  			pr_err("VALIDATE C-TBL[%zu]: %s->%d const out of range (%d-%d)\n",
>  			       i, tbl[i].name, tbl[i].value, low, high);
> 
> -- 
> 2.34.1
>
Zijun Hu April 11, 2025, 2:48 p.m. UTC | #2 
On 2025/4/11 22:37, Christian Brauner wrote:
>> - Potential NULL pointer dereference.
> I really dislike "potential NULL deref" without an explanation. Please
> explain how this supposed NULL deref can happen.
> 

okay.

>> Fixes: 31d921c7fb96 ("vfs: Add configuration parser helpers")
>> Signed-off-by: Zijun Hu <quic_zijuhu@quicinc.com>
>> ---
>>  fs/fs_parser.c | 7 +++++--
>>  1 file changed, 5 insertions(+), 2 deletions(-)
>>
>> diff --git a/fs/fs_parser.c b/fs/fs_parser.c
>> index e635a81e17d965df78ffef27f6885cd70996c6dd..ef7876340a917876bc40df9cdde9232204125a75 100644
>> --- a/fs/fs_parser.c
>> +++ b/fs/fs_parser.c
>> @@ -399,6 +399,9 @@ bool validate_constant_table(const struct constant_table *tbl, size_t tbl_size,
>>  	}
>>  
>>  	for (i = 0; i < tbl_size; i++) {
>> +		if (!tbl[i].name && (i + 1 == tbl_size))
>> +			break;
>> +
>>  		if (!tbl[i].name) {
>>  			pr_err("VALIDATE C-TBL[%zu]: Null\n", i);
>>  			good = false;
>> @@ -411,13 +414,13 @@ bool validate_constant_table(const struct constant_table *tbl, size_t tbl_size,
>>  				good = false;
>>  			}
>>  			if (c > 0) {
>> -				pr_err("VALIDATE C-TBL[%zu]: Missorted %s>=%s\n",
>> +				pr_err("VALIDATE C-TBL[%zu]: Missorted %s>%s\n",
>>  				       i, tbl[i-1].name, tbl[i].name);
>>  				good = false;
>>  			}
>>  		}
>>  
>> -		if (tbl[i].value != special &&
>> +		if (tbl[i].name && tbl[i].value != special &&
>>  		    (tbl[i].value < low || tbl[i].value > high)) {
>>  			pr_err("VALIDATE C-TBL[%zu]: %s->%d const out of range (%d-%d)\n",
>>  			       i, tbl[i].name, tbl[i].value, low, high);

for good constant table which ends with empty entry. for original logic,
when loop reach the last empty entry.  above pr_err() may access NULL
pointer tbl[i].name.


i find out this validate_constant_table() also has no callers.
fix it or remove it ?
Jan Kara April 14, 2025, 12:39 p.m. UTC | #3 
On Fri 11-04-25 22:48:40, Zijun Hu wrote:
> On 2025/4/11 22:37, Christian Brauner wrote:
> >> - Potential NULL pointer dereference.
> > I really dislike "potential NULL deref" without an explanation. Please
> > explain how this supposed NULL deref can happen.
> > 
> 
> okay.
> 
> >> Fixes: 31d921c7fb96 ("vfs: Add configuration parser helpers")
> >> Signed-off-by: Zijun Hu <quic_zijuhu@quicinc.com>
> >> ---
> >>  fs/fs_parser.c | 7 +++++--
> >>  1 file changed, 5 insertions(+), 2 deletions(-)
> >>
> >> diff --git a/fs/fs_parser.c b/fs/fs_parser.c
> >> index e635a81e17d965df78ffef27f6885cd70996c6dd..ef7876340a917876bc40df9cdde9232204125a75 100644
> >> --- a/fs/fs_parser.c
> >> +++ b/fs/fs_parser.c
> >> @@ -399,6 +399,9 @@ bool validate_constant_table(const struct constant_table *tbl, size_t tbl_size,
> >>  	}
> >>  
> >>  	for (i = 0; i < tbl_size; i++) {
> >> +		if (!tbl[i].name && (i + 1 == tbl_size))
> >> +			break;
> >> +
> >>  		if (!tbl[i].name) {
> >>  			pr_err("VALIDATE C-TBL[%zu]: Null\n", i);
> >>  			good = false;
> >> @@ -411,13 +414,13 @@ bool validate_constant_table(const struct constant_table *tbl, size_t tbl_size,
> >>  				good = false;
> >>  			}
> >>  			if (c > 0) {
> >> -				pr_err("VALIDATE C-TBL[%zu]: Missorted %s>=%s\n",
> >> +				pr_err("VALIDATE C-TBL[%zu]: Missorted %s>%s\n",
> >>  				       i, tbl[i-1].name, tbl[i].name);
> >>  				good = false;
> >>  			}
> >>  		}
> >>  
> >> -		if (tbl[i].value != special &&
> >> +		if (tbl[i].name && tbl[i].value != special &&
> >>  		    (tbl[i].value < low || tbl[i].value > high)) {
> >>  			pr_err("VALIDATE C-TBL[%zu]: %s->%d const out of range (%d-%d)\n",
> >>  			       i, tbl[i].name, tbl[i].value, low, high);
> 
> for good constant table which ends with empty entry. for original logic,
> when loop reach the last empty entry.  above pr_err() may access NULL
> pointer tbl[i].name.
> 
> 
> i find out this validate_constant_table() also has no callers.
> fix it or remove it ?

Yeah, just drop it. In the kernel we are pretty aggressive in dropping
unused code :)

								Honza
diff mbox series

Patch

diff --git a/fs/fs_parser.c b/fs/fs_parser.c
index e635a81e17d965df78ffef27f6885cd70996c6dd..ef7876340a917876bc40df9cdde9232204125a75 100644
--- a/fs/fs_parser.c
+++ b/fs/fs_parser.c
@@ -399,6 +399,9 @@  bool validate_constant_table(const struct constant_table *tbl, size_t tbl_size,
 	}
 
 	for (i = 0; i < tbl_size; i++) {
+		if (!tbl[i].name && (i + 1 == tbl_size))
+			break;
+
 		if (!tbl[i].name) {
 			pr_err("VALIDATE C-TBL[%zu]: Null\n", i);
 			good = false;
@@ -411,13 +414,13 @@  bool validate_constant_table(const struct constant_table *tbl, size_t tbl_size,
 				good = false;
 			}
 			if (c > 0) {
-				pr_err("VALIDATE C-TBL[%zu]: Missorted %s>=%s\n",
+				pr_err("VALIDATE C-TBL[%zu]: Missorted %s>%s\n",
 				       i, tbl[i-1].name, tbl[i].name);
 				good = false;
 			}
 		}
 
-		if (tbl[i].value != special &&
+		if (tbl[i].name && tbl[i].value != special &&
 		    (tbl[i].value < low || tbl[i].value > high)) {
 			pr_err("VALIDATE C-TBL[%zu]: %s->%d const out of range (%d-%d)\n",
 			       i, tbl[i].name, tbl[i].value, low, high);