Skip to contentSkip to conversation

env: mmc: Correct partition comparison in mmc_offset_try_partition

PATCH Superseded Hoyeonjiki Kim 2 participants 5 messages Nov 12, 2020 – Nov 15, 2020

v2 of 5 Superseded Latest is v5 →

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 env: mmc: Correct partition comparison in mmc_offset_try_partition

The function mmc_offset_try_partition searches MMC partition to save the environment data by name. However, it only compares the first word-size bytes (size of 'const char *'), which may make the function to find unintended partition.

Correct the function not to partially compare the partition name with config "u-boot,mmc-env-partition".

Fixes c9e87ba66540 ("env: Save environment at the end of an MMC partition")
---
 env/mmc.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
env/mmc.c +1 -1
index 4e67180b23..505f7aa2b8 100644
@@ -42,7 +42,7 @@ static inline int mmc_offset_try_partition(const char *str, int copy, s64 *val)
 		if (ret < 0)
 			return ret;
 
-		if (!strncmp((const char *)info.name, str, sizeof(str)))
+		if (!strcmp((const char *)info.name, str))
 			break;
 	}
 
Signature · 1 line
2.25.1
#1 lore raw

PATCH v2 1/1 env: mmc: Correct partition comparison in mmc_offset_try_partition

Dear Hoyeonjiki Kim,

Quoted 20 lines · In message <[email protected]> you wrote:

The function mmc_offset_try_partition searches MMC partition to save the environment data by name. However, it only compares the first word-size bytes (size of 'const char *'), which may make the function to find unintended partition.

Correct the function not to partially compare the partition name with config "u-boot,mmc-env-partition".

Fixes c9e87ba66540 ("env: Save environment at the end of an MMC partition")
---
 env/mmc.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
env/mmc.c +0 -0
index 4e67180b23..505f7aa2b8 100644

env/mmc.c static inline int mmc_offset_try_partition(const char *str, int copy, s64 *val) · line 45

Quoted 5 lines
env/mmc.c +1 -1
@@ -42,7 +42,7 @@ static inline int mmc_offset_try_partition(const char *str, int copy, s64 *val)
 		if (ret < 0)
 			return ret;
 
-		if (!strncmp((const char *)info.name, str, sizeof(str)))
+		if (!strcmp((const char *)info.name, str))

Resend my comment, too. This looks dangerous, please double check!!

Best regards,

Wolfgang Denk

Signature · 5 lines
DENX Software Engineering GmbH,      Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
In Nature there are neither rewards nor punishments, there are conse-
quences.                                            -- R.G. Ingersoll
#2 lore raw

PATCH v2 1/1 env: mmc: Correct partition comparison in mmc_offset_try_partition

Quoted 27 lines · Wolfgang Denk

Dear Hoyeonjiki Kim,

Quoted 23 lines · In message <[email protected]> you wrote:

The function mmc_offset_try_partition searches MMC partition to save the environment data by name. However, it only compares the first word-size bytes (size of 'const char *'), which may make the function to find unintended partition.

Correct the function not to partially compare the partition name with config "u-boot,mmc-env-partition".

Fixes c9e87ba66540 ("env: Save environment at the end of an MMC partition")
---
 env/mmc.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
env/mmc.c +0 -0
index 4e67180b23..505f7aa2b8 100644
@@ -42,7 +42,7 @@ static inline int mmc_offset_try_partition(const char *str, int copy, s64 *val)
              if (ret < 0)
                      return ret;
 
Quoted 4 lines

env/mmc.c static inline int mmc_offset_try_partition(const char *str, int copy, s64 *val) · line 45

Quoted 2 lines
env/mmc.c +1 -1
-             if (!strncmp((const char *)info.name, str, sizeof(str)))
+             if (!strcmp((const char *)info.name, str))

Resend my comment, too. This looks dangerous, please double check!!

Dear Wolfgang Denk,

Thanks for your feedback.

As you referred, strcmp suffers with non-null terminated string(s). I'd also checked if using strcmp can cause some issues and seems it's **guaranteed** that there is no such issue in this context.

Here's why:
- the first input string, info.name comes from fn part_get_info which gets the partition info from one of the partition driver.

Each driver will return the partition info with null terminated partition name (actually it must, or every part in U-Boot referring info.name will have potential issues), so the first input string is safe to use in strcmp.

- The second one, str comes from fn fdtdec_get_config_string which gets the 'u-boot,mmc-env-offset' property value from FDT.

When you keep tracking that function, you will meet fdt_get_string which returns error (-FDT_ERR_TRUNCATED) if the property value is non-null terminated. So the second input string also is safe to use.

fdtdec_get_config_string
--> fdt_getprop

      --> fdt_getprop_namelen
          --> fdt_get_property_namelen_
              --> fdt_string_eq_
                  --> fdt_get_stringa

For this reason, I think we can use strcmp in this context.

But if we need to specify that the context will not suffer anyway, there is an option to use strncmp with PART_NAME_LEN as max count param.

PART_NAME_LEN is the size of info.name which is a character buffer.

Please let me know your opinion.

Quoted 11 lines

Best regards,

Wolfgang Denk

--
DENX Software Engineering GmbH, Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de In Nature there are neither rewards nor punishments, there are conse-
quences. -- R.G. Ingersoll

#3 lore raw

PATCH v2 1/1 env: mmc: Correct partition comparison in mmc_offset_try_partition

Dear Hoyeonjiki Kim,

Quoted 4 lines · In message <CAL9K-_jni+850m5Zf7QPPeM+dmSxqSZ5AgirDnzE_2uxrO4Akg@mail.gmail.com>…

As you referred, strcmp suffers with non-null terminated string(s). I'd also checked if using strcmp can cause some issues and seems it's **guaranteed** that there is no such issue in this context.

You ar4e probably right, but the problem with this approach is that what today is a verified context, may tomoroow change - a new use case may be added, which is not aware of this potential problem, and which thus triggers a (foreseeable and avoidable bug).

Quoted 4 lines

But if we need to specify that the context will not suffer anyway, there is an option to use strncmp with PART_NAME_LEN as max count param.

PART_NAME_LEN is the size of info.name which is a character buffer.

If we know we size (and apparewntly we do), we should use this with strncmp(). Just in case...

Best regards,

Wolfgang Denk

Signature · 4 lines
DENX Software Engineering GmbH,      Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
It's not what you do, it's how you do what you do!  - Jordan D. Ulmer
#4 lore raw

PATCH v2 1/1 env: mmc: Correct partition comparison in mmc_offset_try_partition

Dear Wolfgang Denk,

Quoted 5 lines · Wolfgang Denk

Dear Hoyeonjiki Kim,

Quoted 1 line · In message <CAL9K-_jni+850m5Zf7QPPeM+dmSxqSZ5AgirDnzE_2uxrO4Akg@mail.gmail.com>…
Quoted 8 lines
Quoted 3 lines

As you referred, strcmp suffers with non-null terminated string(s). I'd also checked if using strcmp can cause some issues and seems it's **guaranteed** that there is no such issue in this context.

You ar4e probably right, but the problem with this approach is that what today is a verified context, may tomoroow change - a new use case may be added, which is not aware of this potential problem, and which thus triggers a (foreseeable and avoidable bug).

Absolutely. I got your point and agree with that.

Quoted 8 lines
Quoted 5 lines

But if we need to specify that the context will not suffer anyway, there is an option to use strncmp with PART_NAME_LEN as max count param.

PART_NAME_LEN is the size of info.name which is a character buffer.

If we know we size (and apparewntly we do), we should use this with strncmp(). Just in case...

Because in this context, we can use 'sizeof(info.name)' as max count, I think I can bring v3 with strncmp().

Thanks for the feedback.

Best Regards,
Hoyeonjiki Kim

Quoted 10 lines

Best regards,

Wolfgang Denk

--
DENX Software Engineering GmbH, Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de It's not what you do, it's how you do what you do! - Jordan D. Ulmer