Netfilter-Devel Archive on lore.kernel.org
 help / color / Atom feed
* [libnftnl PATCH] src: Fix nftnl_assert() on data_len
@ 2020-02-14 17:24 Phil Sutter
  2020-02-14 17:32 ` Pablo Neira Ayuso
  0 siblings, 1 reply; 12+ messages in thread
From: Phil Sutter @ 2020-02-14 17:24 UTC (permalink / raw)
  To: Pablo Neira Ayuso; +Cc: netfilter-devel

Typical idiom for *_get_u*() getters is to call *_get_data() and make
sure data_len matches what each of them is returning. Yet they shouldn't
trust *_get_data() to write into passed pointer to data_len since for
chains and NFTNL_CHAIN_DEVICES attribute, it does not. Make sure these
assert() calls trigger in those cases.

Signed-off-by: Phil Sutter <phil@nwl.cc>
---
 src/chain.c | 8 ++++----
 src/rule.c  | 6 +++---
 src/set.c   | 4 ++--
 3 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/src/chain.c b/src/chain.c
index b4066e4d4e888..62a9249b57930 100644
--- a/src/chain.c
+++ b/src/chain.c
@@ -385,7 +385,7 @@ const char *nftnl_chain_get_str(const struct nftnl_chain *c, uint16_t attr)
 EXPORT_SYMBOL(nftnl_chain_get_u32);
 uint32_t nftnl_chain_get_u32(const struct nftnl_chain *c, uint16_t attr)
 {
-	uint32_t data_len;
+	uint32_t data_len = 0;
 	const uint32_t *val = nftnl_chain_get_data(c, attr, &data_len);
 
 	nftnl_assert(val, attr, data_len == sizeof(uint32_t));
@@ -396,7 +396,7 @@ uint32_t nftnl_chain_get_u32(const struct nftnl_chain *c, uint16_t attr)
 EXPORT_SYMBOL(nftnl_chain_get_s32);
 int32_t nftnl_chain_get_s32(const struct nftnl_chain *c, uint16_t attr)
 {
-	uint32_t data_len;
+	uint32_t data_len = 0;
 	const int32_t *val = nftnl_chain_get_data(c, attr, &data_len);
 
 	nftnl_assert(val, attr, data_len == sizeof(int32_t));
@@ -407,7 +407,7 @@ int32_t nftnl_chain_get_s32(const struct nftnl_chain *c, uint16_t attr)
 EXPORT_SYMBOL(nftnl_chain_get_u64);
 uint64_t nftnl_chain_get_u64(const struct nftnl_chain *c, uint16_t attr)
 {
-	uint32_t data_len;
+	uint32_t data_len = 0;
 	const uint64_t *val = nftnl_chain_get_data(c, attr, &data_len);
 
 	nftnl_assert(val, attr, data_len == sizeof(int64_t));
@@ -418,7 +418,7 @@ uint64_t nftnl_chain_get_u64(const struct nftnl_chain *c, uint16_t attr)
 EXPORT_SYMBOL(nftnl_chain_get_u8);
 uint8_t nftnl_chain_get_u8(const struct nftnl_chain *c, uint16_t attr)
 {
-	uint32_t data_len;
+	uint32_t data_len = 0;
 	const uint8_t *val = nftnl_chain_get_data(c, attr, &data_len);
 
 	nftnl_assert(val, attr, data_len == sizeof(int8_t));
diff --git a/src/rule.c b/src/rule.c
index 8d7e0681cb42c..ffdbbf8e08140 100644
--- a/src/rule.c
+++ b/src/rule.c
@@ -249,7 +249,7 @@ const char *nftnl_rule_get_str(const struct nftnl_rule *r, uint16_t attr)
 EXPORT_SYMBOL(nftnl_rule_get_u32);
 uint32_t nftnl_rule_get_u32(const struct nftnl_rule *r, uint16_t attr)
 {
-	uint32_t data_len;
+	uint32_t data_len = 0;
 	const uint32_t *val = nftnl_rule_get_data(r, attr, &data_len);
 
 	nftnl_assert(val, attr, data_len == sizeof(uint32_t));
@@ -260,7 +260,7 @@ uint32_t nftnl_rule_get_u32(const struct nftnl_rule *r, uint16_t attr)
 EXPORT_SYMBOL(nftnl_rule_get_u64);
 uint64_t nftnl_rule_get_u64(const struct nftnl_rule *r, uint16_t attr)
 {
-	uint32_t data_len;
+	uint32_t data_len = 0;
 	const uint64_t *val = nftnl_rule_get_data(r, attr, &data_len);
 
 	nftnl_assert(val, attr, data_len == sizeof(uint64_t));
@@ -271,7 +271,7 @@ uint64_t nftnl_rule_get_u64(const struct nftnl_rule *r, uint16_t attr)
 EXPORT_SYMBOL(nftnl_rule_get_u8);
 uint8_t nftnl_rule_get_u8(const struct nftnl_rule *r, uint16_t attr)
 {
-	uint32_t data_len;
+	uint32_t data_len = 0;
 	const uint8_t *val = nftnl_rule_get_data(r, attr, &data_len);
 
 	nftnl_assert(val, attr, data_len == sizeof(uint8_t));
diff --git a/src/set.c b/src/set.c
index 651dcfa56022d..6a6f19bc7fbbf 100644
--- a/src/set.c
+++ b/src/set.c
@@ -303,7 +303,7 @@ const char *nftnl_set_get_str(const struct nftnl_set *s, uint16_t attr)
 EXPORT_SYMBOL(nftnl_set_get_u32);
 uint32_t nftnl_set_get_u32(const struct nftnl_set *s, uint16_t attr)
 {
-	uint32_t data_len;
+	uint32_t data_len = 0;
 	const uint32_t *val = nftnl_set_get_data(s, attr, &data_len);
 
 	nftnl_assert(val, attr, data_len == sizeof(uint32_t));
@@ -314,7 +314,7 @@ uint32_t nftnl_set_get_u32(const struct nftnl_set *s, uint16_t attr)
 EXPORT_SYMBOL(nftnl_set_get_u64);
 uint64_t nftnl_set_get_u64(const struct nftnl_set *s, uint16_t attr)
 {
-	uint32_t data_len;
+	uint32_t data_len = 0;
 	const uint64_t *val = nftnl_set_get_data(s, attr, &data_len);
 
 	nftnl_assert(val, attr, data_len == sizeof(uint64_t));
-- 
2.24.1


^ permalink raw reply	[flat|nested] 12+ messages in thread

* Re: [libnftnl PATCH] src: Fix nftnl_assert() on data_len
  2020-02-14 17:24 [libnftnl PATCH] src: Fix nftnl_assert() on data_len Phil Sutter
@ 2020-02-14 17:32 ` Pablo Neira Ayuso
  2020-02-14 17:34   ` Phil Sutter
  0 siblings, 1 reply; 12+ messages in thread
From: Pablo Neira Ayuso @ 2020-02-14 17:32 UTC (permalink / raw)
  To: Phil Sutter; +Cc: netfilter-devel

On Fri, Feb 14, 2020 at 06:24:17PM +0100, Phil Sutter wrote:
> Typical idiom for *_get_u*() getters is to call *_get_data() and make
> sure data_len matches what each of them is returning. Yet they shouldn't
> trust *_get_data() to write into passed pointer to data_len since for
> chains and NFTNL_CHAIN_DEVICES attribute, it does not. Make sure these
> assert() calls trigger in those cases.

The intention to catch for unset attributes through the assertion,
right?

^ permalink raw reply	[flat|nested] 12+ messages in thread

* Re: [libnftnl PATCH] src: Fix nftnl_assert() on data_len
  2020-02-14 17:32 ` Pablo Neira Ayuso
@ 2020-02-14 17:34   ` Phil Sutter
  2020-02-14 17:42     ` Pablo Neira Ayuso
  0 siblings, 1 reply; 12+ messages in thread
From: Phil Sutter @ 2020-02-14 17:34 UTC (permalink / raw)
  To: Pablo Neira Ayuso; +Cc: netfilter-devel

On Fri, Feb 14, 2020 at 06:32:47PM +0100, Pablo Neira Ayuso wrote:
> On Fri, Feb 14, 2020 at 06:24:17PM +0100, Phil Sutter wrote:
> > Typical idiom for *_get_u*() getters is to call *_get_data() and make
> > sure data_len matches what each of them is returning. Yet they shouldn't
> > trust *_get_data() to write into passed pointer to data_len since for
> > chains and NFTNL_CHAIN_DEVICES attribute, it does not. Make sure these
> > assert() calls trigger in those cases.
> 
> The intention to catch for unset attributes through the assertion,
> right?

No, this is about making sure that no wrong getter is called, e.g.
nftnl_chain_get_u64() with e.g. NFTNL_CHAIN_HOOKNUM attribute which is
only 32bits.

Cheers, Phil

^ permalink raw reply	[flat|nested] 12+ messages in thread

* Re: [libnftnl PATCH] src: Fix nftnl_assert() on data_len
  2020-02-14 17:34   ` Phil Sutter
@ 2020-02-14 17:42     ` Pablo Neira Ayuso
  2020-02-15  0:43       ` Phil Sutter
  0 siblings, 1 reply; 12+ messages in thread
From: Pablo Neira Ayuso @ 2020-02-14 17:42 UTC (permalink / raw)
  To: Phil Sutter, netfilter-devel

On Fri, Feb 14, 2020 at 06:34:50PM +0100, Phil Sutter wrote:
> On Fri, Feb 14, 2020 at 06:32:47PM +0100, Pablo Neira Ayuso wrote:
> > On Fri, Feb 14, 2020 at 06:24:17PM +0100, Phil Sutter wrote:
> > > Typical idiom for *_get_u*() getters is to call *_get_data() and make
> > > sure data_len matches what each of them is returning. Yet they shouldn't
> > > trust *_get_data() to write into passed pointer to data_len since for
> > > chains and NFTNL_CHAIN_DEVICES attribute, it does not. Make sure these
> > > assert() calls trigger in those cases.
> > 
> > The intention to catch for unset attributes through the assertion,
> > right?
> 
> No, this is about making sure that no wrong getter is called, e.g.
> nftnl_chain_get_u64() with e.g. NFTNL_CHAIN_HOOKNUM attribute which is
> only 32bits.

I think it will also catch the case I'm asking. If attribute is unset,
then nftnl_chain_get_data() returns NULL and the assertion checks
data_len, which has not been properly initialized.

^ permalink raw reply	[flat|nested] 12+ messages in thread

* Re: [libnftnl PATCH] src: Fix nftnl_assert() on data_len
  2020-02-14 17:42     ` Pablo Neira Ayuso
@ 2020-02-15  0:43       ` Phil Sutter
  2020-02-15 13:17         ` Pablo Neira Ayuso
  0 siblings, 1 reply; 12+ messages in thread
From: Phil Sutter @ 2020-02-15  0:43 UTC (permalink / raw)
  To: Pablo Neira Ayuso; +Cc: netfilter-devel

Hi Pablo,

On Fri, Feb 14, 2020 at 06:42:00PM +0100, Pablo Neira Ayuso wrote:
> On Fri, Feb 14, 2020 at 06:34:50PM +0100, Phil Sutter wrote:
> > On Fri, Feb 14, 2020 at 06:32:47PM +0100, Pablo Neira Ayuso wrote:
> > > On Fri, Feb 14, 2020 at 06:24:17PM +0100, Phil Sutter wrote:
> > > > Typical idiom for *_get_u*() getters is to call *_get_data() and make
> > > > sure data_len matches what each of them is returning. Yet they shouldn't
> > > > trust *_get_data() to write into passed pointer to data_len since for
> > > > chains and NFTNL_CHAIN_DEVICES attribute, it does not. Make sure these
> > > > assert() calls trigger in those cases.
> > > 
> > > The intention to catch for unset attributes through the assertion,
> > > right?
> > 
> > No, this is about making sure that no wrong getter is called, e.g.
> > nftnl_chain_get_u64() with e.g. NFTNL_CHAIN_HOOKNUM attribute which is
> > only 32bits.
> 
> I think it will also catch the case I'm asking. If attribute is unset,
> then nftnl_chain_get_data() returns NULL and the assertion checks
> data_len, which has not been properly initialized.

With nftnl_assert() being (shortened):

| #define nftnl_assert(val, attr, expr) \
|  ((!val || expr) ? \
|  (void)0 : __nftnl_assert_fail(attr, __FILE__, __LINE__))

Check for 'expr' (which is passed as 'data_len == sizeof(<something>)')
will only happen if 'val' is not NULL. Callers then return like so:

| return val ? *val : 0;

This means that if you pass an unset attribute to the getter, it will
simply return 0.

Cheers, Phil

^ permalink raw reply	[flat|nested] 12+ messages in thread

* Re: [libnftnl PATCH] src: Fix nftnl_assert() on data_len
  2020-02-15  0:43       ` Phil Sutter
@ 2020-02-15 13:17         ` Pablo Neira Ayuso
  2020-02-15 22:58           ` Phil Sutter
  0 siblings, 1 reply; 12+ messages in thread
From: Pablo Neira Ayuso @ 2020-02-15 13:17 UTC (permalink / raw)
  To: Phil Sutter, netfilter-devel


[-- Attachment #1: Type: text/plain, Size: 2029 bytes --]

On Sat, Feb 15, 2020 at 01:43:11AM +0100, Phil Sutter wrote:
> Hi Pablo,
> 
> On Fri, Feb 14, 2020 at 06:42:00PM +0100, Pablo Neira Ayuso wrote:
> > On Fri, Feb 14, 2020 at 06:34:50PM +0100, Phil Sutter wrote:
> > > On Fri, Feb 14, 2020 at 06:32:47PM +0100, Pablo Neira Ayuso wrote:
> > > > On Fri, Feb 14, 2020 at 06:24:17PM +0100, Phil Sutter wrote:
> > > > > Typical idiom for *_get_u*() getters is to call *_get_data() and make
> > > > > sure data_len matches what each of them is returning. Yet they shouldn't
> > > > > trust *_get_data() to write into passed pointer to data_len since for
> > > > > chains and NFTNL_CHAIN_DEVICES attribute, it does not. Make sure these
> > > > > assert() calls trigger in those cases.
> > > > 
> > > > The intention to catch for unset attributes through the assertion,
> > > > right?
> > > 
> > > No, this is about making sure that no wrong getter is called, e.g.
> > > nftnl_chain_get_u64() with e.g. NFTNL_CHAIN_HOOKNUM attribute which is
> > > only 32bits.
> > 
> > I think it will also catch the case I'm asking. If attribute is unset,
> > then nftnl_chain_get_data() returns NULL and the assertion checks
> > data_len, which has not been properly initialized.
> 
> With nftnl_assert() being (shortened):
> 
> | #define nftnl_assert(val, attr, expr) \
> |  ((!val || expr) ? \
> |  (void)0 : __nftnl_assert_fail(attr, __FILE__, __LINE__))
> 
> Check for 'expr' (which is passed as 'data_len == sizeof(<something>)')
> will only happen if 'val' is not NULL. Callers then return like so:
> 
> | return val ? *val : 0;
> 
> This means that if you pass an unset attribute to the getter, it will
> simply return 0.

Thanks for explaining, Phil. If the problem is just
NFTNL_CHAIN_DEVICES and NFTNL_FLOWTABLE_DEVICES, probably this is just
fine? So zero data-length is reversed for arrays and update
nftnl_assert() to skip data_len == 0, ie.

> | #define nftnl_assert(val, attr, expr) \
> |  ((!val || data_len == 0 || expr) ? \
> |  (void)0 : __nftnl_assert_fail(attr, __FILE__, __LINE__))

[-- Attachment #2: x.patch --]
[-- Type: text/x-diff, Size: 1261 bytes --]

diff --git a/include/utils.h b/include/utils.h
index 8af5a8e973fa..53999c982c56 100644
--- a/include/utils.h
+++ b/include/utils.h
@@ -30,7 +30,7 @@ void __noreturn __abi_breakage(const char *file, int line, const char *reason);
 void __nftnl_assert_fail(uint16_t attr, const char *filename, int line);
 
 #define nftnl_assert(val, attr, expr)			\
-  ((!val || expr)					\
+  ((!val || data_len == 0 || expr)			\
    ? (void)0						\
    : __nftnl_assert_fail(attr, __FILE__, __LINE__))
 
diff --git a/src/chain.c b/src/chain.c
index b4066e4d4e88..94a9e43a1754 100644
--- a/src/chain.c
+++ b/src/chain.c
@@ -364,6 +364,7 @@ const void *nftnl_chain_get_data(const struct nftnl_chain *c, uint16_t attr,
 		*data_len = strlen(c->dev) + 1;
 		return c->dev;
 	case NFTNL_CHAIN_DEVICES:
+		*data_len = 0;
 		return &c->dev_array[0];
 	}
 	return NULL;
diff --git a/src/flowtable.c b/src/flowtable.c
index 1e235d0ba50f..635322d7fa56 100644
--- a/src/flowtable.c
+++ b/src/flowtable.c
@@ -230,6 +230,7 @@ const void *nftnl_flowtable_get_data(const struct nftnl_flowtable *c,
 		*data_len = sizeof(int32_t);
 		return &c->family;
 	case NFTNL_FLOWTABLE_DEVICES:
+		*data_len = 0;
 		return &c->dev_array[0];
 	case NFTNL_FLOWTABLE_SIZE:
 		*data_len = sizeof(int32_t);

^ permalink raw reply	[flat|nested] 12+ messages in thread

* Re: [libnftnl PATCH] src: Fix nftnl_assert() on data_len
  2020-02-15 13:17         ` Pablo Neira Ayuso
@ 2020-02-15 22:58           ` Phil Sutter
  2020-02-18 13:42             ` Pablo Neira Ayuso
  0 siblings, 1 reply; 12+ messages in thread
From: Phil Sutter @ 2020-02-15 22:58 UTC (permalink / raw)
  To: Pablo Neira Ayuso; +Cc: netfilter-devel

Hi Pablo,

On Sat, Feb 15, 2020 at 02:17:13PM +0100, Pablo Neira Ayuso wrote:
> On Sat, Feb 15, 2020 at 01:43:11AM +0100, Phil Sutter wrote:
> > Hi Pablo,
> > 
> > On Fri, Feb 14, 2020 at 06:42:00PM +0100, Pablo Neira Ayuso wrote:
> > > On Fri, Feb 14, 2020 at 06:34:50PM +0100, Phil Sutter wrote:
> > > > On Fri, Feb 14, 2020 at 06:32:47PM +0100, Pablo Neira Ayuso wrote:
> > > > > On Fri, Feb 14, 2020 at 06:24:17PM +0100, Phil Sutter wrote:
> > > > > > Typical idiom for *_get_u*() getters is to call *_get_data() and make
> > > > > > sure data_len matches what each of them is returning. Yet they shouldn't
> > > > > > trust *_get_data() to write into passed pointer to data_len since for
> > > > > > chains and NFTNL_CHAIN_DEVICES attribute, it does not. Make sure these
> > > > > > assert() calls trigger in those cases.
> > > > > 
> > > > > The intention to catch for unset attributes through the assertion,
> > > > > right?
> > > > 
> > > > No, this is about making sure that no wrong getter is called, e.g.
> > > > nftnl_chain_get_u64() with e.g. NFTNL_CHAIN_HOOKNUM attribute which is
> > > > only 32bits.
> > > 
> > > I think it will also catch the case I'm asking. If attribute is unset,
> > > then nftnl_chain_get_data() returns NULL and the assertion checks
> > > data_len, which has not been properly initialized.
> > 
> > With nftnl_assert() being (shortened):
> > 
> > | #define nftnl_assert(val, attr, expr) \
> > |  ((!val || expr) ? \
> > |  (void)0 : __nftnl_assert_fail(attr, __FILE__, __LINE__))
> > 
> > Check for 'expr' (which is passed as 'data_len == sizeof(<something>)')
> > will only happen if 'val' is not NULL. Callers then return like so:
> > 
> > | return val ? *val : 0;
> > 
> > This means that if you pass an unset attribute to the getter, it will
> > simply return 0.
> 
> Thanks for explaining, Phil. If the problem is just
> NFTNL_CHAIN_DEVICES and NFTNL_FLOWTABLE_DEVICES, probably this is just
> fine? So zero data-length is reversed for arrays and update
> nftnl_assert() to skip data_len == 0, ie.
> 
> > | #define nftnl_assert(val, attr, expr) \
> > |  ((!val || data_len == 0 || expr) ? \
> > |  (void)0 : __nftnl_assert_fail(attr, __FILE__, __LINE__))

Your proposed patch would allow to call e.g.:

| nftnl_chain_get_u32(c, NFTNL_CHAIN_DEVICES)

This would return (uint32_t)*(&c->dev_array[0]), I highly doubt we
should allow this. Unless I miss something, it is certainly a
programming error if someone calls any of the nftnl_chain_get_{u,s}*
getters on NFTNL_CHAIN_DEVICES attribute. So aborting with error message
in nftnl_assert() is not only OK but actually helpful, no?

Cheers, Phil

^ permalink raw reply	[flat|nested] 12+ messages in thread

* Re: [libnftnl PATCH] src: Fix nftnl_assert() on data_len
  2020-02-15 22:58           ` Phil Sutter
@ 2020-02-18 13:42             ` Pablo Neira Ayuso
  2020-02-18 18:18               ` Phil Sutter
  0 siblings, 1 reply; 12+ messages in thread
From: Pablo Neira Ayuso @ 2020-02-18 13:42 UTC (permalink / raw)
  To: Phil Sutter, netfilter-devel

Hi Phil,

On Sat, Feb 15, 2020 at 11:58:55PM +0100, Phil Sutter wrote:
> Hi Pablo,
> 
> On Sat, Feb 15, 2020 at 02:17:13PM +0100, Pablo Neira Ayuso wrote:
> > On Sat, Feb 15, 2020 at 01:43:11AM +0100, Phil Sutter wrote:
> > > Hi Pablo,
> > > 
> > > On Fri, Feb 14, 2020 at 06:42:00PM +0100, Pablo Neira Ayuso wrote:
> > > > On Fri, Feb 14, 2020 at 06:34:50PM +0100, Phil Sutter wrote:
> > > > > On Fri, Feb 14, 2020 at 06:32:47PM +0100, Pablo Neira Ayuso wrote:
> > > > > > On Fri, Feb 14, 2020 at 06:24:17PM +0100, Phil Sutter wrote:
> > > > > > > Typical idiom for *_get_u*() getters is to call *_get_data() and make
> > > > > > > sure data_len matches what each of them is returning. Yet they shouldn't
> > > > > > > trust *_get_data() to write into passed pointer to data_len since for
> > > > > > > chains and NFTNL_CHAIN_DEVICES attribute, it does not. Make sure these
> > > > > > > assert() calls trigger in those cases.
> > > > > > 
> > > > > > The intention to catch for unset attributes through the assertion,
> > > > > > right?
> > > > > 
> > > > > No, this is about making sure that no wrong getter is called, e.g.
> > > > > nftnl_chain_get_u64() with e.g. NFTNL_CHAIN_HOOKNUM attribute which is
> > > > > only 32bits.
> > > > 
> > > > I think it will also catch the case I'm asking. If attribute is unset,
> > > > then nftnl_chain_get_data() returns NULL and the assertion checks
> > > > data_len, which has not been properly initialized.
> > > 
> > > With nftnl_assert() being (shortened):
> > > 
> > > | #define nftnl_assert(val, attr, expr) \
> > > |  ((!val || expr) ? \
> > > |  (void)0 : __nftnl_assert_fail(attr, __FILE__, __LINE__))
> > > 
> > > Check for 'expr' (which is passed as 'data_len == sizeof(<something>)')
> > > will only happen if 'val' is not NULL. Callers then return like so:
> > > 
> > > | return val ? *val : 0;
> > > 
> > > This means that if you pass an unset attribute to the getter, it will
> > > simply return 0.
> > 
> > Thanks for explaining, Phil. If the problem is just
> > NFTNL_CHAIN_DEVICES and NFTNL_FLOWTABLE_DEVICES, probably this is just
> > fine? So zero data-length is reversed for arrays and update
> > nftnl_assert() to skip data_len == 0, ie.
> > 
> > > | #define nftnl_assert(val, attr, expr) \
> > > |  ((!val || data_len == 0 || expr) ? \
> > > |  (void)0 : __nftnl_assert_fail(attr, __FILE__, __LINE__))
> 
> Your proposed patch would allow to call e.g.:
> 
> | nftnl_chain_get_u32(c, NFTNL_CHAIN_DEVICES)
> 
> This would return (uint32_t)*(&c->dev_array[0]), I highly doubt we
> should allow this. Unless I miss something, it is certainly a
> programming error if someone calls any of the nftnl_chain_get_{u,s}*
> getters on NFTNL_CHAIN_DEVICES attribute. So aborting with error message
> in nftnl_assert() is not only OK but actually helpful, no?

Indeed, good point.

I don't think nftnl_flowtable_set_data() is good for these two device
array.

I just sent a patch, I forgot to finish the _set_array() and
_get_array() helpers for the flowtable, the definition in the header
file prooves this.

Can we introduce these new interfaces? Then, update nftables to use it.
Then, at some point, set *data_len = 0 for these array datatypes. Yes,
it's a bit longer term, but better fix this interface. But setting all
these data_len to zero when in most cases it is going to be thereafter
properly set to the datatype length is...

Would this work for you? I know it is not so short term.

^ permalink raw reply	[flat|nested] 12+ messages in thread

* Re: [libnftnl PATCH] src: Fix nftnl_assert() on data_len
  2020-02-18 13:42             ` Pablo Neira Ayuso
@ 2020-02-18 18:18               ` Phil Sutter
  2020-02-18 21:06                 ` Pablo Neira Ayuso
  0 siblings, 1 reply; 12+ messages in thread
From: Phil Sutter @ 2020-02-18 18:18 UTC (permalink / raw)
  To: Pablo Neira Ayuso; +Cc: netfilter-devel

Hi Pablo,

On Tue, Feb 18, 2020 at 02:42:27PM +0100, Pablo Neira Ayuso wrote:
> On Sat, Feb 15, 2020 at 11:58:55PM +0100, Phil Sutter wrote:
> > On Sat, Feb 15, 2020 at 02:17:13PM +0100, Pablo Neira Ayuso wrote:
> > > On Sat, Feb 15, 2020 at 01:43:11AM +0100, Phil Sutter wrote:
> > > > On Fri, Feb 14, 2020 at 06:42:00PM +0100, Pablo Neira Ayuso wrote:
> > > > > On Fri, Feb 14, 2020 at 06:34:50PM +0100, Phil Sutter wrote:
> > > > > > On Fri, Feb 14, 2020 at 06:32:47PM +0100, Pablo Neira Ayuso wrote:
> > > > > > > On Fri, Feb 14, 2020 at 06:24:17PM +0100, Phil Sutter wrote:
> > > > > > > > Typical idiom for *_get_u*() getters is to call *_get_data() and make
> > > > > > > > sure data_len matches what each of them is returning. Yet they shouldn't
> > > > > > > > trust *_get_data() to write into passed pointer to data_len since for
> > > > > > > > chains and NFTNL_CHAIN_DEVICES attribute, it does not. Make sure these
> > > > > > > > assert() calls trigger in those cases.
> > > > > > > 
> > > > > > > The intention to catch for unset attributes through the assertion,
> > > > > > > right?
> > > > > > 
> > > > > > No, this is about making sure that no wrong getter is called, e.g.
> > > > > > nftnl_chain_get_u64() with e.g. NFTNL_CHAIN_HOOKNUM attribute which is
> > > > > > only 32bits.
> > > > > 
> > > > > I think it will also catch the case I'm asking. If attribute is unset,
> > > > > then nftnl_chain_get_data() returns NULL and the assertion checks
> > > > > data_len, which has not been properly initialized.
> > > > 
> > > > With nftnl_assert() being (shortened):
> > > > 
> > > > | #define nftnl_assert(val, attr, expr) \
> > > > |  ((!val || expr) ? \
> > > > |  (void)0 : __nftnl_assert_fail(attr, __FILE__, __LINE__))
> > > > 
> > > > Check for 'expr' (which is passed as 'data_len == sizeof(<something>)')
> > > > will only happen if 'val' is not NULL. Callers then return like so:
> > > > 
> > > > | return val ? *val : 0;
> > > > 
> > > > This means that if you pass an unset attribute to the getter, it will
> > > > simply return 0.
> > > 
> > > Thanks for explaining, Phil. If the problem is just
> > > NFTNL_CHAIN_DEVICES and NFTNL_FLOWTABLE_DEVICES, probably this is just
> > > fine? So zero data-length is reversed for arrays and update
> > > nftnl_assert() to skip data_len == 0, ie.
> > > 
> > > > | #define nftnl_assert(val, attr, expr) \
> > > > |  ((!val || data_len == 0 || expr) ? \
> > > > |  (void)0 : __nftnl_assert_fail(attr, __FILE__, __LINE__))
> > 
> > Your proposed patch would allow to call e.g.:
> > 
> > | nftnl_chain_get_u32(c, NFTNL_CHAIN_DEVICES)
> > 
> > This would return (uint32_t)*(&c->dev_array[0]), I highly doubt we
> > should allow this. Unless I miss something, it is certainly a
> > programming error if someone calls any of the nftnl_chain_get_{u,s}*
> > getters on NFTNL_CHAIN_DEVICES attribute. So aborting with error message
> > in nftnl_assert() is not only OK but actually helpful, no?
> 
> Indeed, good point.
> 
> I don't think nftnl_flowtable_set_data() is good for these two device
> array.

Well, right now it serves as a backend for all attribute setters, and
your patch continues in that tradition. So while it may be a bit
"rustic", I'd say it's good enough for its purpose. :)

> I just sent a patch, I forgot to finish the _set_array() and
> _get_array() helpers for the flowtable, the definition in the header
> file prooves this.
> 
> Can we introduce these new interfaces? Then, update nftables to use it.
> Then, at some point, set *data_len = 0 for these array datatypes. Yes,
> it's a bit longer term, but better fix this interface. But setting all
> these data_len to zero when in most cases it is going to be thereafter
> properly set to the datatype length is...
> 
> Would this work for you? I know it is not so short term.

While I think your patch is the right way to providing a sanitized
access to the array attributes, I don't think it's really related to
what my original patch was fixing, which is:

Right now we are preventing users from passing wrong attribute types to
getters by checking the attribute length. This does not work for
NFTNL_CHAIN_DEVICES or NFTNL_FLOWTABLE_DEVICES because they don't set
data_len. Hence the expression in nftnl_asser() call:

| nftnl_assert(val, attr, data_len == sizeof(<something>));

Will lead to comparing with garbage from stack. This may in most cases
fail as expected, but there's no guarantee.

Your patch allows to use "a better" getter/setter for those problematic
attributes, but it doesn't prevent the above from happening.

My first approach was to make nftnl_chain_get_data() and
nftnl_flowtable_get_data() set:

| *data_len = 0;

for the problematic attributes, but the value is not really correct - a
"more correct" value, e.g.:

| *data_len = c->dev_array_len * sizeof(char *);

Could lead to a pass in getter sanitizing by accident although e.g.
nftnl_chain_get_u64() is completely unfit even if c->dev_array_len was
1.

So I decided to go the safe way and initialize data_len variables to zero
instead which has the benefit of catching new attributes added later as
well.

If you don't like the approach of initializing all data_len variables, I
would rather suggest to go with setting '*data_len = 0' in _get_data()
routines as described above. This has the same effect but it's just a
two lines change. What do you think?

Cheers, Phil

^ permalink raw reply	[flat|nested] 12+ messages in thread

* Re: [libnftnl PATCH] src: Fix nftnl_assert() on data_len
  2020-02-18 18:18               ` Phil Sutter
@ 2020-02-18 21:06                 ` Pablo Neira Ayuso
  2020-02-18 23:02                   ` Phil Sutter
  0 siblings, 1 reply; 12+ messages in thread
From: Pablo Neira Ayuso @ 2020-02-18 21:06 UTC (permalink / raw)
  To: Phil Sutter, netfilter-devel


[-- Attachment #1: Type: text/plain, Size: 5894 bytes --]

On Tue, Feb 18, 2020 at 07:18:51PM +0100, Phil Sutter wrote:
> Hi Pablo,
> 
> On Tue, Feb 18, 2020 at 02:42:27PM +0100, Pablo Neira Ayuso wrote:
> > On Sat, Feb 15, 2020 at 11:58:55PM +0100, Phil Sutter wrote:
> > > On Sat, Feb 15, 2020 at 02:17:13PM +0100, Pablo Neira Ayuso wrote:
> > > > On Sat, Feb 15, 2020 at 01:43:11AM +0100, Phil Sutter wrote:
> > > > > On Fri, Feb 14, 2020 at 06:42:00PM +0100, Pablo Neira Ayuso wrote:
> > > > > > On Fri, Feb 14, 2020 at 06:34:50PM +0100, Phil Sutter wrote:
> > > > > > > On Fri, Feb 14, 2020 at 06:32:47PM +0100, Pablo Neira Ayuso wrote:
> > > > > > > > On Fri, Feb 14, 2020 at 06:24:17PM +0100, Phil Sutter wrote:
> > > > > > > > > Typical idiom for *_get_u*() getters is to call *_get_data() and make
> > > > > > > > > sure data_len matches what each of them is returning. Yet they shouldn't
> > > > > > > > > trust *_get_data() to write into passed pointer to data_len since for
> > > > > > > > > chains and NFTNL_CHAIN_DEVICES attribute, it does not. Make sure these
> > > > > > > > > assert() calls trigger in those cases.
> > > > > > > > 
> > > > > > > > The intention to catch for unset attributes through the assertion,
> > > > > > > > right?
> > > > > > > 
> > > > > > > No, this is about making sure that no wrong getter is called, e.g.
> > > > > > > nftnl_chain_get_u64() with e.g. NFTNL_CHAIN_HOOKNUM attribute which is
> > > > > > > only 32bits.
> > > > > > 
> > > > > > I think it will also catch the case I'm asking. If attribute is unset,
> > > > > > then nftnl_chain_get_data() returns NULL and the assertion checks
> > > > > > data_len, which has not been properly initialized.
> > > > > 
> > > > > With nftnl_assert() being (shortened):
> > > > > 
> > > > > | #define nftnl_assert(val, attr, expr) \
> > > > > |  ((!val || expr) ? \
> > > > > |  (void)0 : __nftnl_assert_fail(attr, __FILE__, __LINE__))
> > > > > 
> > > > > Check for 'expr' (which is passed as 'data_len == sizeof(<something>)')
> > > > > will only happen if 'val' is not NULL. Callers then return like so:
> > > > > 
> > > > > | return val ? *val : 0;
> > > > > 
> > > > > This means that if you pass an unset attribute to the getter, it will
> > > > > simply return 0.
> > > > 
> > > > Thanks for explaining, Phil. If the problem is just
> > > > NFTNL_CHAIN_DEVICES and NFTNL_FLOWTABLE_DEVICES, probably this is just
> > > > fine? So zero data-length is reversed for arrays and update
> > > > nftnl_assert() to skip data_len == 0, ie.
> > > > 
> > > > > | #define nftnl_assert(val, attr, expr) \
> > > > > |  ((!val || data_len == 0 || expr) ? \
> > > > > |  (void)0 : __nftnl_assert_fail(attr, __FILE__, __LINE__))
> > > 
> > > Your proposed patch would allow to call e.g.:
> > > 
> > > | nftnl_chain_get_u32(c, NFTNL_CHAIN_DEVICES)
> > > 
> > > This would return (uint32_t)*(&c->dev_array[0]), I highly doubt we
> > > should allow this. Unless I miss something, it is certainly a
> > > programming error if someone calls any of the nftnl_chain_get_{u,s}*
> > > getters on NFTNL_CHAIN_DEVICES attribute. So aborting with error message
> > > in nftnl_assert() is not only OK but actually helpful, no?
> > 
> > Indeed, good point.
> > 
> > I don't think nftnl_flowtable_set_data() is good for these two device
> > array.
> 
> Well, right now it serves as a backend for all attribute setters, and
> your patch continues in that tradition. So while it may be a bit
> "rustic", I'd say it's good enough for its purpose. :)
> 
> > I just sent a patch, I forgot to finish the _set_array() and
> > _get_array() helpers for the flowtable, the definition in the header
> > file prooves this.
> > 
> > Can we introduce these new interfaces? Then, update nftables to use it.
> > Then, at some point, set *data_len = 0 for these array datatypes. Yes,
> > it's a bit longer term, but better fix this interface. But setting all
> > these data_len to zero when in most cases it is going to be thereafter
> > properly set to the datatype length is...
> > 
> > Would this work for you? I know it is not so short term.
> 
> While I think your patch is the right way to providing a sanitized
> access to the array attributes, I don't think it's really related to
> what my original patch was fixing, which is:
> 
> Right now we are preventing users from passing wrong attribute types to
> getters by checking the attribute length. This does not work for
> NFTNL_CHAIN_DEVICES or NFTNL_FLOWTABLE_DEVICES because they don't set
> data_len. Hence the expression in nftnl_asser() call:
> 
> | nftnl_assert(val, attr, data_len == sizeof(<something>));
> 
> Will lead to comparing with garbage from stack. This may in most cases
> fail as expected, but there's no guarantee.
> 
> Your patch allows to use "a better" getter/setter for those problematic
> attributes, but it doesn't prevent the above from happening.
>
> My first approach was to make nftnl_chain_get_data() and
> nftnl_flowtable_get_data() set:
> 
> | *data_len = 0;
> 
> for the problematic attributes, but the value is not really correct - a
> "more correct" value, e.g.:
> 
> | *data_len = c->dev_array_len * sizeof(char *);
> 
> Could lead to a pass in getter sanitizing by accident although e.g.
> nftnl_chain_get_u64() is completely unfit even if c->dev_array_len was
> 1.
> 
> So I decided to go the safe way and initialize data_len variables to zero
> instead which has the benefit of catching new attributes added later as
> well.
> 
> If you don't like the approach of initializing all data_len variables, I
> would rather suggest to go with setting '*data_len = 0' in _get_data()
> routines as described above. This has the same effect but it's just a
> two lines change. What do you think?

If I apply the patch that I'm attaching, then I use the wrong datatype
helper:

        nftnl_flowtable_get_u32(nlo, NFTNL_FLOWTABLE_DEVICES);

And I can see:

libnftnl: attribute 6 assertion failed in flowtable.c:274

[-- Attachment #2: x.patch --]
[-- Type: text/x-diff, Size: 772 bytes --]

diff --git a/src/chain.c b/src/chain.c
index e25eb0f5934b..e98af1360912 100644
--- a/src/chain.c
+++ b/src/chain.c
@@ -371,6 +371,7 @@ const void *nftnl_chain_get_data(const struct nftnl_chain *c, uint16_t attr,
 		*data_len = strlen(c->dev) + 1;
 		return c->dev;
 	case NFTNL_CHAIN_DEVICES:
+		*data_len = 0;
 		return &c->dev_array[0];
 	}
 	return NULL;
diff --git a/src/flowtable.c b/src/flowtable.c
index 6e18dde8242e..18a3c98ea62d 100644
--- a/src/flowtable.c
+++ b/src/flowtable.c
@@ -237,6 +237,7 @@ const void *nftnl_flowtable_get_data(const struct nftnl_flowtable *c,
 		*data_len = sizeof(int32_t);
 		return &c->family;
 	case NFTNL_FLOWTABLE_DEVICES:
+		*data_len = 0;
 		return &c->dev_array[0];
 	case NFTNL_FLOWTABLE_SIZE:
 		*data_len = sizeof(int32_t);

^ permalink raw reply	[flat|nested] 12+ messages in thread

* Re: [libnftnl PATCH] src: Fix nftnl_assert() on data_len
  2020-02-18 21:06                 ` Pablo Neira Ayuso
@ 2020-02-18 23:02                   ` Phil Sutter
  2020-02-19  9:32                     ` Pablo Neira Ayuso
  0 siblings, 1 reply; 12+ messages in thread
From: Phil Sutter @ 2020-02-18 23:02 UTC (permalink / raw)
  To: Pablo Neira Ayuso; +Cc: netfilter-devel

Hi Pablo,

On Tue, Feb 18, 2020 at 10:06:11PM +0100, Pablo Neira Ayuso wrote:
[...]
> If I apply the patch that I'm attaching, then I use the wrong datatype
> helper:
> 
>         nftnl_flowtable_get_u32(nlo, NFTNL_FLOWTABLE_DEVICES);
> 
> And I can see:
> 
> libnftnl: attribute 6 assertion failed in flowtable.c:274

Yes, that was what I meant with alternative (simpler) approach. Should I
submit this change formally or do you want to do it?

Thanks, Phil

^ permalink raw reply	[flat|nested] 12+ messages in thread

* Re: [libnftnl PATCH] src: Fix nftnl_assert() on data_len
  2020-02-18 23:02                   ` Phil Sutter
@ 2020-02-19  9:32                     ` Pablo Neira Ayuso
  0 siblings, 0 replies; 12+ messages in thread
From: Pablo Neira Ayuso @ 2020-02-19  9:32 UTC (permalink / raw)
  To: Phil Sutter, netfilter-devel

On Wed, Feb 19, 2020 at 12:02:39AM +0100, Phil Sutter wrote:
> Hi Pablo,
> 
> On Tue, Feb 18, 2020 at 10:06:11PM +0100, Pablo Neira Ayuso wrote:
> [...]
> > If I apply the patch that I'm attaching, then I use the wrong datatype
> > helper:
> > 
> >         nftnl_flowtable_get_u32(nlo, NFTNL_FLOWTABLE_DEVICES);
> > 
> > And I can see:
> > 
> > libnftnl: attribute 6 assertion failed in flowtable.c:274
> 
> Yes, that was what I meant with alternative (simpler) approach. Should I
> submit this change formally or do you want to do it?

Please submit and push it out, thanks.

^ permalink raw reply	[flat|nested] 12+ messages in thread

end of thread, back to index

Thread overview: 12+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-02-14 17:24 [libnftnl PATCH] src: Fix nftnl_assert() on data_len Phil Sutter
2020-02-14 17:32 ` Pablo Neira Ayuso
2020-02-14 17:34   ` Phil Sutter
2020-02-14 17:42     ` Pablo Neira Ayuso
2020-02-15  0:43       ` Phil Sutter
2020-02-15 13:17         ` Pablo Neira Ayuso
2020-02-15 22:58           ` Phil Sutter
2020-02-18 13:42             ` Pablo Neira Ayuso
2020-02-18 18:18               ` Phil Sutter
2020-02-18 21:06                 ` Pablo Neira Ayuso
2020-02-18 23:02                   ` Phil Sutter
2020-02-19  9:32                     ` Pablo Neira Ayuso

Netfilter-Devel Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/netfilter-devel/0 netfilter-devel/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 netfilter-devel netfilter-devel/ https://lore.kernel.org/netfilter-devel \
		netfilter-devel@vger.kernel.org
	public-inbox-index netfilter-devel

Example config snippet for mirrors

Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.kernel.vger.netfilter-devel


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git