lkml.org 
[lkml]   [2016]   [May]   [26]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patches in this message
/
Date
From
SubjectRe: [Regression?] iptables broken on 32bit with pre-4.7-rc
John Stultz <john.stultz@linaro.org> wrote:
> In updating a 32bit arm device from 4.6 to Linus' current HEAD, I
> noticed I was having some trouble with networking, and realized that
> /proc/net/ip_tables_names was suddenly empty.
>
> Digging through the registration process, it seems we're catching on the:
>
> if (strcmp(t->u.user.name, XT_STANDARD_TARGET) == 0 &&
> target_offset + sizeof(struct xt_standard_target) != next_offset)
> return -EINVAL;
>
> check added in 7ed2abddd20cf ("netfilter: x_tables: check standard
> target size too").
>
> Where next_offset seems to be 4 bytes larger then the the offset +
> standard_target struct size.

I guess its because arm32 needs 8 byte alignment for 64bit
quantities. So we can fix this either via XT_ALIGN()'ing the
target_offset + sizeof() result or by weakening the test to a '>'.

Since we already test proper alignment of start-of-rule in
check_entry_size_and_hooks() I'd suggest we just change the test
to fail only if the next offset is within the min size, i.e.:

diff --git a/net/netfilter/x_tables.c b/net/netfilter/x_tables.c
index c69c892..9643047 100644
--- a/net/netfilter/x_tables.c
+++ b/net/netfilter/x_tables.c
@@ -612,7 +612,7 @@ int xt_compat_check_entry_offsets(const void *base, const char *elems,
return -EINVAL;

if (strcmp(t->u.user.name, XT_STANDARD_TARGET) == 0 &&
- target_offset + sizeof(struct compat_xt_standard_target) != next_offset)
+ target_offset + sizeof(struct compat_xt_standard_target) > next_offset)
return -EINVAL;

/* compat_xt_entry match has less strict aligment requirements,
@@ -694,7 +694,7 @@ int xt_check_entry_offsets(const void *base,
return -EINVAL;

if (strcmp(t->u.user.name, XT_STANDARD_TARGET) == 0 &&
- target_offset + sizeof(struct xt_standard_target) != next_offset)
+ target_offset + sizeof(struct xt_standard_target) > next_offset)
return -EINVAL;

return xt_check_entry_match(elems, base + target_offset,
> I'm not exactly sure how the next_offset value is set, so I'm hoping
> the proper fix is more obvious to one of you.

Its the start of the next rule so it has to be properly aligned
via XT_ALIGN(). Only 32bit system I tested was plain x86 which
only needs 4byte alignment for u64...

Alternative would be something like this:

diff --git a/net/netfilter/x_tables.c b/net/netfilter/x_tables.c
index c69c892..ca16c26 100644
--- a/net/netfilter/x_tables.c
+++ b/net/netfilter/x_tables.c
@@ -612,7 +612,7 @@ int xt_compat_check_entry_offsets(const void *base, const char *elems,
return -EINVAL;

if (strcmp(t->u.user.name, XT_STANDARD_TARGET) == 0 &&
- target_offset + sizeof(struct compat_xt_standard_target) != next_offset)
+ XT_COMPAT_ALIGN(target_offset + sizeof(struct compat_xt_standard_target)) != next_offset)
return -EINVAL;

/* compat_xt_entry match has less strict aligment requirements,
@@ -694,7 +694,7 @@ int xt_check_entry_offsets(const void *base,
return -EINVAL;

if (strcmp(t->u.user.name, XT_STANDARD_TARGET) == 0 &&
- target_offset + sizeof(struct xt_standard_target) != next_offset)
+ XT_ALIGN(target_offset + sizeof(struct xt_standard_target)) != next_offset)
return -EINVAL;

return xt_check_entry_match(elems, base + target_offset,

but afaics the stricter check does not buy anything.

\
 
 \ /
  Last update: 2016-05-26 12:01    [W:0.049 / U:2.236 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site