Misc Cleanups.
Subsystems (2)
David Miller
maintainerPATCH 1/2 Misc Cleanups.
Forwarding to netfilter-devel where this belongs...
Hi,
In this patch, these three fixes were made in net/ipv4/netfilter/ip_queue.c:
1) No need to perform data_len = 0 in the switch command, since data_len is initialized to 0 in the beginning of the method
,ipq_build_packet_message().
2) We can reach nlmsg_failure only from one place; skb is sure to be NULL when getting there; since skb is NULL, there is no need to check this fact and call kfree_skb().
3) Add #ifdef CONFIG_PROC_FS when removing the VFS entry, proc_net_remove(&init_net, IPQ_PROC_FS_NAME);
Regards,
Rami Rosen
net/ipv4/netfilter/ip_queue.c +2 -3
@@ -156,7 +156,6 @@ ipq_build_packet_message(struct nf_queue_entry *entry, int *errp)case IPQ_COPY_META :case IPQ_COPY_NONE : size =NLMSG_SPACE (sizeof (*pmsg)); - data_len =0 ;break ;case IPQ_COPY_PACKET :
@@ -224,8 +223,6 @@ ipq_build_packet_message(struct nf_queue_entry *entry, int *errp)return skb; nlmsg_failure: -if (skb) -kfree_skb (skb); *errp = -EINVAL ;printk (KERN_ERR "ip_queue: error creating packet message\n" );return NULL ;
@@ -610,7 +607,9 @@ cleanup_sysctl:unregister_sysctl_table (ipq_sysctl_header);#endif unregister_netdevice_notifier (&ipq_dev_notifier); +#ifdef CONFIG_PROC_FS proc_net_remove (&init_net,IPQ_PROC_FS_NAME ); +#endif cleanup_ipqnl:__maybe_unused netlink_kernel_release (ipqnl);mutex_lock (&ipqnl_mutex);
Quoted 1 line · David Miller
Forwarding to netfilter-devel where this belongs...
Thanks.
Quoted 14 lines
In this patch, these three fixes were made in
net/ipv4/netfilter/ip_queue.c:1) No need to perform
data_len= 0 in the switch command, sincedata_lenis initialized to 0 in the beginning of the method
,ipq_build_packet_message().2) We can reach
nlmsg_failureonly from one place; skb is sure to be NULL when getting there; since skb is NULL, there is no need to check this fact and callkfree_skb().3) Add #ifdef
CONFIG_PROC_FSwhen removing the VFS entry,proc_net_remove(&init_net, IPQ_PROC_FS_NAME);
Quoted 5 lines
Regards,
Rami RosenSigned-off-by Rami Rosen <[email protected]>
1) also affects ip6_queue and nfnetlink_queue
2) also affects ip6_queue
3) is unnecessary since proc_net_remove is a NOP without CONFIG_PROC_FS
Please update your patch to also change ip6_queue and nfnetlink_queue where applicable. Thanks.
PATCH 1/1 Misc Cleanups.
Hi,
- Thanks for your comments; I was not aware that this issues occur
in other places too; attached here is another patch, fixing where applicable in ip6_queue and nfnetlink_queue.
1) in net/ipv6/netfilter/ip6_queue.c
- No need to perform data_len = 0 in the switch command, since data_len is initialized to 0 in the beginning of theipq_build_packet_message() method
- We can reach nlmsg_failure only from one place; skb is sure to be NULL when getting there; since skb is NULL, there is no need to check this fact and call kfree_skb().
2) in net/netfilter/nfnetlink_queue.c:
- No need to perform data_len = 0 in the switch command, since data_len is initialized to 0 in the beginning of thenfqnl_build_packet_message() method
(Note: here, as opposed to previous patch, nlmsg_failure must check skb and free it if it is not NULL, so the call to kfree_skb() is needed , so it is not removed)
Regards,
Rami Rosen
Quoted 30 lines · Patrick McHardy
Quoted 2 lines · David Miller
Forwarding to netfilter-devel where this belongs...
Thanks.
Quoted 18 lines
In this patch, these three fixes were made in
net/ipv4/netfilter/ip_queue.c:1) No need to perform
data_len= 0 in the switch command, sincedata_len
is initialized to 0 in the beginning of the method
,ipq_build_packet_message().2) We can reach
nlmsg_failureonly from one place; skb is sure to
be NULL
when getting there; since skb is NULL, there is no need to check
this fact
and callkfree_skb().3) Add #ifdef
CONFIG_PROC_FSwhen removing the VFS entry,proc_net_remove(&init_net, IPQ_PROC_FS_NAME);Quoted 5 lines
Regards,
Rami RosenSigned-off-by Rami Rosen <[email protected]>
Quoted 10 lines
1) also affects
ip6_queueandnfnetlink_queue
2) also affectsip6_queue
3) is unnecessary sinceproc_net_removeis a NOP withoutCONFIG_PROC_FSPlease update your patch to also change
ip6_queueandnfnetlink_queuewhere applicable. Thanks.
net/ipv6/netfilter/ip6_queue.c +0 -3
@@ -159,7 +159,6 @@ ipq_build_packet_message(struct nf_queue_entry *entry, int *errp)case IPQ_COPY_META :case IPQ_COPY_NONE : size =NLMSG_SPACE (sizeof (*pmsg)); - data_len =0 ;break ;case IPQ_COPY_PACKET :
@@ -226,8 +225,6 @@ ipq_build_packet_message(struct nf_queue_entry *entry, int *errp)return skb; nlmsg_failure: -if (skb) -kfree_skb (skb); *errp = -EINVAL ;printk (KERN_ERR "ip6_queue: error creating packet message\n" );return NULL ;
net/netfilter/nfnetlink_queue.c +0 -1
@@ -243,7 +243,6 @@ nfqnl_build_packet_message(struct nfqnl_instance *queue,switch ((enum nfqnl_config_mode )queue->copy_mode) {case NFQNL_COPY_META :case NFQNL_COPY_NONE : - data_len =0 ;break ;case NFQNL_COPY_PACKET :
Quoted 13 lines · Rami Rosen
Hi,
- Thanks for your comments; I was not aware that this issues occur
in other places too; attached here is another patch, fixing where applicable inip6_queueandnfnetlink_queue.1) in
net/ipv6/netfilter/ip6_queue.c
- No need to performdata_len= 0 in the switch command, sincedata_lenis initialized to 0 in the beginning of theipq_build_packet_message()method
- We can reachnlmsg_failureonly from one place; skb is sure to be NULL when getting there; since skb is NULL, there is no need to check this fact and callkfree_skb().
Quoted 8 lines
2) in
net/netfilter/nfnetlink_queue.c:
- No need to performdata_len= 0 in the switch command, sincedata_lenis initialized to 0 in the beginning of thenfqnl_build_packet_message()method(Note: here, as opposed to previous patch,
nlmsg_failuremust check skb and free it if it is not NULL, so the call tokfree_skb()is needed , so it is not removed)
I've queued this patch and the previous one (without the procfs bits) as one combined patch for 2.6.27, thanks.