ip6_output: Add ipv6_pinfo null check
Subsystems (3)
PATCH 1/1 ip6_output: Add ipv6_pinfo null check
ipv6_pinfo is initlialized by inet6_sk() which returns NULL. Hence it can cause segmentation fault. Fix this by adding a NULL check.
--- net/ipv6/ip6_output.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
net/ipv6/ip6_output.c +3 -3
@@ -181,10 +181,10 @@ int ip6_output(struct net *net, struct sock *sk, struct sk_buff *skb)bool ip6_autoflowlabel (struct net *net,const struct ipv6_pinfo *np) { -if (!np->autoflowlabel_set) -return ip6_default_np_autolabel (net); -else +if (np && np->autoflowlabel_set)return np->autoflowlabel; +else +ip6_default_np_autolabel (net); }/*
Signature · 1 line
2.17.1
kernel test robot
build botThank you for the patch! Perhaps something to improve:
Hi Gaurav,
Thank you for the patch! Perhaps something to improve:
[auto build test WARNING on sparc-next/master]
[also build test WARNING on ipvs/master linus/master v5.8-rc7 next-20200724] [If your patch is applied to the wrong git tree, kindly drop us a note. And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch]
url: https://github.com/0day-ci/linux/commits/Gaurav-Singh/ip6_output-Add-ipv6_pinfo-null-check/20200727-113949 base: https://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc-next.git master config: csky-defconfig (attached as .config) compiler: csky-linux-gcc (GCC) 9.3.0
reproduce (this is a W=1 build):
wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross chmod +x ~/bin/make.cross
# save the attached .config to linux build tree
COMPILER_INSTALL_PATH=$HOME/0day COMPILER=gcc-9.3.0 make.cross ARCH=csky
If you fix the issue, kindly add following tag as appropriate
All warnings (new ones prefixed by >>):
net/ipv6/ip6_output.c: In function 'ip6_autoflowlabel':
>> net/ipv6/ip6_output.c:188:1: warning: control reaches end of non-void function [-Wreturn-type]
188 | }
| ^vim +188 net/ipv6/ip6_output.c
^1da177e4c3f41 Linus Torvalds 2005-04-16 181
e9191ffb65d8e1 Ben Hutchings 2018-01-22 182 bool ip6_autoflowlabel(struct net *net, const struct ipv6_pinfo *np)
513674b5a2c9c7 Shaohua Li 2017-12-20 183 {
5bdc1ea8a7d229 Gaurav Singh 2020-07-26 184 if (np && np->autoflowlabel_set)
513674b5a2c9c7 Shaohua Li 2017-12-20 185 return np->autoflowlabel;
5bdc1ea8a7d229 Gaurav Singh 2020-07-26 186 else
5bdc1ea8a7d229 Gaurav Singh 2020-07-26 187 ip6_default_np_autolabel(net);513674b5a2c9c7 Shaohua Li 2017-12-20 @188 }513674b5a2c9c7 Shaohua Li 2017-12-20 189
---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all(a)lists.01.org
kernel test robot
build botThank you for the patch! Perhaps something to improve:
Hi Gaurav,
Thank you for the patch! Perhaps something to improve:
[auto build test WARNING on sparc-next/master]
[also build test WARNING on ipvs/master linus/master v5.8-rc7 next-20200724] [If your patch is applied to the wrong git tree, kindly drop us a note. And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch]
url: https://github.com/0day-ci/linux/commits/Gaurav-Singh/ip6_output-Add-ipv6_pinfo-null-check/20200727-113949 base: https://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc-next.git master config: x86_64-randconfig-r011-20200727 (attached as .config) compiler: clang version 12.0.0 (https://github.com/llvm/llvm-project 8dc820393219c7ee440b4ec86c9a201301943276)
reproduce (this is a W=1 build):
wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross chmod +x ~/bin/make.cross
# install x86_64 cross compiling tool for clang build
# apt-get install binutils-x86-64-linux-gnu
# save the attached .config to linux build tree
COMPILER_INSTALL_PATH=$HOME/0day COMPILER=clang make.cross ARCH=x86_64
If you fix the issue, kindly add following tag as appropriate
All warnings (new ones prefixed by >>):
>> net/ipv6/ip6_output.c:188:1: warning: non-void function does not return a value in all control paths [-Wreturn-type]
}
^
1 warning generated.
vim +188 net/ipv6/ip6_output.c
^1da177e4c3f41 Linus Torvalds 2005-04-16 181
e9191ffb65d8e1 Ben Hutchings 2018-01-22 182 bool ip6_autoflowlabel(struct net *net, const struct ipv6_pinfo *np)
513674b5a2c9c7 Shaohua Li 2017-12-20 183 {
5bdc1ea8a7d229 Gaurav Singh 2020-07-26 184 if (np && np->autoflowlabel_set)
513674b5a2c9c7 Shaohua Li 2017-12-20 185 return np->autoflowlabel;
5bdc1ea8a7d229 Gaurav Singh 2020-07-26 186 else
5bdc1ea8a7d229 Gaurav Singh 2020-07-26 187 ip6_default_np_autolabel(net);513674b5a2c9c7 Shaohua Li 2017-12-20 @188 }513674b5a2c9c7 Shaohua Li 2017-12-20 189
---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all(a)lists.01.org
David Miller
maintainerFrom: Gaurav Singh <[email protected]>
Date: Sun, 26 Jul 2020 23:38:10 -0400
Quoted 3 lines
ipv6_pinfois initlialized byinet6_sk()which returns NULL. Hence it can cause segmentation fault. Fix this by adding a NULL check.
Please take your time with such changes and actually look at the compiler output, it will warn that you are adding a new problem.
Specifically, the function now fails to return a value at all.
But even more important, how is the situation you are fixing possible? Do you have a sample crash? Can you please describe the code paths and conditions that lead to the problem?
You must also provide a valid and properly formatted Fixes: tag for bug fixes such as this.
Thank you.
Quoted 2 lines · Gaurav Singh
ipv6_pinfois initlialized byinet6_sk()which returns NULL.
Why? It only returns NULL for timewait or request sock, but I don't see how ip6_autoflowlabel() could be called on these sockets. So please explain.
Quoted 2 lines
Hence it can cause segmentation fault. Fix this by adding a NULL check.
Which exact call path? Do you have a full stack trace?
Thanks.
PATCH 1/1 ip6_output: Add ipv6_pinfo null check
In reply to Gaurav SinghAdd return to fix build issue. Haven't reproduced this issue at my end.
My hypothesis is this: In function: ip6_xmit(), we have const struct ipv6_pinfo *np = inet6_sk(sk); which returns NULL.
Further down the function, there's a check:
if (np) hlimit = hp->htop_limit
Further, we have a callip6_flow_hdr(hdr, tclass, ip6_make_flowlabel(net, skb, fl6->flowlabel, ip6_autoflowlabel(net, np), fl6)); .
Hence np = NULL gets passed in
the function ip6_autoflowlabel() which accesses np-> without check which may cause a segment violation.
--- net/ipv6/ip6_output.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
net/ipv6/ip6_output.c +3 -3
@@ -181,10 +181,10 @@ int ip6_output(struct net *net, struct sock *sk, struct sk_buff *skb)bool ip6_autoflowlabel (struct net *net,const struct ipv6_pinfo *np) { -if (!np->autoflowlabel_set) -return ip6_default_np_autolabel (net); -else +if (np && np->autoflowlabel_set)return np->autoflowlabel; +else +return ip6_default_np_autolabel (net); }/*
Signature · 1 line
2.17.1
Quoted 9 lines · Gaurav Singh
Add return to fix build issue. Haven't reproduced this issue at my end.
My hypothesis is this: In function:
ip6_xmit(), we have const structipv6_pinfo*np =inet6_sk(sk); which returns NULL.Further down the function, there's a check:
if (np) hlimit =hp->htop_limit
This check exists before git history, at that time 'sk' could be NULL, hence 'np', so it does not mean it is still necessary now.
I looked at all callers of ip6_xmit(), I don't see how it is called with a non-full socket, neither 'sk' could be NULL after
commit b30bd282cbf5c46247a279a2e8d2aae027d9f1bf
("[IPV6]: ip6_xmit: remove unnecessary NULL ptr check").
Thanks.
Quoted 14 lines · Cong Wang
Quoted 9 lines · Gaurav Singh
Add return to fix build issue. Haven't reproduced this issue at my end.
My hypothesis is this: In function:
ip6_xmit(), we have const structipv6_pinfo*np =inet6_sk(sk); which returns NULL.Further down the function, there's a check:
if (np) hlimit =hp->htop_limitThis check exists before git history, at that time 'sk' could be NULL, hence 'np', so it does not mean it is still necessary now.
Quoted 7 lines
I looked at all callers of
ip6_xmit(), I don't see how it is called with a non-full socket, neither 'sk' could be NULL after
commitb30bd282cbf5c46247a279a2e8d2aae027d9f1bf
("[IPV6]:ip6_xmit: remove unnecessary NULL ptr check").Thanks.
Agreed.
And again, fact that this patch lacks a Fixes: tag speaks for itself.
This means the author expects all reviewers to make a deep analysis.
Please bear with us, and add a Fixes: tag so that we can fully understand what was the bug origin and why a fix is valid.
PATCH 1/1 ip6_output: Add ipv6_pinfo null check
In reply to Gaurav SinghThis PR fixes a possible segmentation violation.
In function: ip6_xmit(), we have
const struct ipv6_pinfo *np = inet6_sk(sk); which returns NULL unconditionally (regardless sk being NULL or not).
In include/linux/ipv6.h:
static inline struct ipv6_pinfo * inet6_sk(const struct sock *__sk)
{
return NULL;
}Further down the function, there's a check:
if (np) hlimit = hp->htop_limit
Thereafter, we have a callip6_flow_hdr(hdr, tclass, ip6_make_flowlabel(net, skb, fl6->flowlabel, ip6_autoflowlabel(net, np), fl6));
Hence np = NULL gets passed in
the function ip6_autoflowlabel() that accesses np without check which may cause a segment violation.
--- net/ipv6/ip6_output.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
net/ipv6/ip6_output.c +3 -3
@@ -181,10 +181,10 @@ int ip6_output(struct net *net, struct sock *sk, struct sk_buff *skb)bool ip6_autoflowlabel (struct net *net,const struct ipv6_pinfo *np) { -if (!np->autoflowlabel_set) -return ip6_default_np_autolabel (net); -else +if (np && np->autoflowlabel_set)return np->autoflowlabel; +else +return ip6_default_np_autolabel (net); }/*
Signature · 1 line
2.17.1
Quoted 7 lines · Gaurav Singh
This PR fixes a possible segmentation violation.
In function:
ip6_xmit(), we have
const structipv6_pinfo*np =inet6_sk(sk); which returns NULL unconditionally (regardless sk being NULL or not).
Quoted 7 lines
In
include/linux/ipv6.h:static inline struct
ipv6_pinfo*inet6_sk(const struct sock *__sk){ return NULL; }
Tell us who will use ip6_autoflowlabel() when CONFIG_IPV6 is disabled. :)