Skip to content

K
Kernel

openeuler / Kernel
1 年多 前同步成功

通知 8
Star 0
Fork 0
K
Kernel
切换分支/标签
  1. 21 5月, 2019 1 次提交
  3. 23 4月, 2019 3 次提交
  5. 08 4月, 2019 3 次提交
    • F
      xfrm: store xfrm_mode directly, not its address · c9500d7b
      由 Florian Westphal 提交于 
      This structure is now only 4 bytes, so its more efficient
to cache a copy rather than its address.

No significant size difference in allmodconfig vmlinux.

With non-modular kernel that has all XFRM options enabled, this
series reduces vmlinux image size by ~11kb. All xfrm_mode
indirections are gone and all modes are built-in.

before (ipsec-next master):
    text      data      bss         dec   filename
21071494   7233140 11104324    39408958   vmlinux.master

after this series:
21066448   7226772 11104324    39397544   vmlinux.patched

With allmodconfig kernel, the size increase is only 362 bytes,
even all the xfrm config options removed in this series are
modular.

before:
    text      data     bss      dec   filename
15731286   6936912 4046908 26715106   vmlinux.master

after this series:
15731492   6937068  4046908  26715468 vmlinux
Signed-off-by: NFlorian Westphal <fw@strlen.de>
Reviewed-by: NSabrina Dubroca <sd@queasysnail.net>
Signed-off-by: NSteffen Klassert <steffen.klassert@secunet.com>
      c9500d7b
    • F
      xfrm: make xfrm modes builtin · 4c145dce
      由 Florian Westphal 提交于 
      after previous changes, xfrm_mode contains no function pointers anymore
and all modules defining such struct contain no code except an init/exit
functions to register the xfrm_mode struct with the xfrm core.

Just place the xfrm modes core and remove the modules,
the run-time xfrm_mode register/unregister functionality is removed.

Before:

    text    data     bss      dec filename
    7523     200    2364    10087 net/xfrm/xfrm_input.o
   40003     628     440    41071 net/xfrm/xfrm_state.o
15730338 6937080 4046908 26714326 vmlinux

    7389     200    2364    9953  net/xfrm/xfrm_input.o
   40574     656     440   41670  net/xfrm/xfrm_state.o
15730084 6937068 4046908 26714060 vmlinux

The xfrm*_mode_{transport,tunnel,beet} modules are gone.

v2: replace CONFIG_INET6_XFRM_MODE_* IS_ENABLED guards with CONFIG_IPV6
    ones rather than removing them.
Signed-off-by: NFlorian Westphal <fw@strlen.de>
Reviewed-by: NSabrina Dubroca <sd@queasysnail.net>
Signed-off-by: NSteffen Klassert <steffen.klassert@secunet.com>
      4c145dce
    • F
      xfrm: remove afinfo pointer from xfrm_mode · 733a5fac
      由 Florian Westphal 提交于 
      Adds an EXPORT_SYMBOL for afinfo_get_rcu, as it will now be called from
ipv6 in case of CONFIG_IPV6=m.

This change has virtually no effect on vmlinux size, but it reduces
afinfo size and allows followup patch to make xfrm modes const.

v2: mark if (afinfo) tests as likely (Sabrina)
    re-fetch afinfo according to inner_mode in xfrm_prepare_input().
Signed-off-by: NFlorian Westphal <fw@strlen.de>
Reviewed-by: NSabrina Dubroca <sd@queasysnail.net>
Signed-off-by: NSteffen Klassert <steffen.klassert@secunet.com>
      733a5fac
  7. 27 3月, 2019 1 次提交
  9. 18 2月, 2019 1 次提交
    • T
      xfrm: Fix inbound traffic via XFRM interfaces across network namespaces · 660899dd
      由 Tobias Brunner 提交于 
      After moving an XFRM interface to another namespace it stays associated
with the original namespace (net in `struct xfrm_if` and the list keyed
with `xfrmi_net_id`), allowing processes in the new namespace to use
SAs/policies that were created in the original namespace.  For instance,
this allows a keying daemon in one namespace to establish IPsec SAs for
other namespaces without processes there having access to the keys or IKE
credentials.

This worked fine for outbound traffic, however, for inbound traffic the
lookup for the interfaces and the policies used the incorrect namespace
(the one the XFRM interface was moved to).

Fixes: f203b76d ("xfrm: Add virtual xfrm interfaces")
Signed-off-by: NTobias Brunner <tobias@strongswan.org>
Signed-off-by: NSteffen Klassert <steffen.klassert@secunet.com>
      660899dd
  11. 16 1月, 2019 1 次提交
  13. 09 1月, 2019 5 次提交
    • F
      xfrm: policy: fix infinite loop when merging src-nodes · 12750aba
      由 Florian Westphal 提交于 
      With very small change to test script we can trigger softlockup due to
bogus assignment of 'p' (policy to be examined) on restart.

Previously the two to-be-merged nodes had same address/prefixlength pair,
so no erase/reinsert was necessary, we only had to append the list from
node a to b.

If prefix lengths are different, the node has to be deleted and re-inserted
into the tree, with the updated prefix length.  This was broken; due to
bogus update to 'p' this loops forever.

Add a 'restart' label and use that instead.

While at it, don't perform the unneeded reinserts of the policies that
are already sorted into the 'new' node.

A previous patch in this series made xfrm_policy_inexact_list_reinsert()
use the relative position indicator to sort policies according to age in
case priorities are identical.

Fixes: 6ac098b2 ("xfrm: policy: add 2nd-level saddr trees for inexact policies")
Signed-off-by: NFlorian Westphal <fw@strlen.de>
Signed-off-by: NSteffen Klassert <steffen.klassert@secunet.com>
      12750aba
    • F
      xfrm: policy: fix reinsertion on node merge · 1d38900c
      由 Florian Westphal 提交于 
      "newpos" has wrong scope.  It must be NULL on each iteration of the loop.
Otherwise, when policy is to be inserted at the start, we would instead
insert at point found by the previous loop-iteration instead.

Also, we need to unlink the policy before we reinsert it to the new node,
else we can get next-points-to-self loops.

Because policies are only ordered by priority it is irrelevant which policy
is "more recent" except when two policies have same priority.
(the more recent one is placed after the older one).

In these cases, we can use the ->pos id number to know which one is the
'older': the higher the id, the more recent the policy.

So we only need to unlink all policies from the node that is about to be
removed, and insert them to the replacement node.

Fixes: 9cf545eb ("xfrm: policy: store inexact policies in a tree ordered by destination address")
Signed-off-by: NFlorian Westphal <fw@strlen.de>
Signed-off-by: NSteffen Klassert <steffen.klassert@secunet.com>
      1d38900c
    • F
      xfrm: policy: delete inexact policies from inexact list on hash rebuild · 1548bc4e
      由 Florian Westphal 提交于 
      An xfrm hash rebuild has to reset the inexact policy list before the
policies get re-inserted: A change of hash thresholds will result in
policies to get moved from inexact tree to the policy hash table.

If the thresholds are increased again later, they get moved from hash
table to inexact tree.

We must unlink all policies from the inexact tree before re-insertion.

Otherwise 'migrate' may find policies that are in main hash table a
second time, when it searches the inexact lists.

Furthermore, re-insertion without deletion can cause elements ->next to
point back to itself, causing soft lockups or double-frees.

Reported-by: syzbot+9d971dd21eb26567036b@syzkaller.appspotmail.com
Fixes: 9cf545eb ("xfrm: policy: store inexact policies in a tree ordered by destination address")
Signed-off-by: NFlorian Westphal <fw@strlen.de>
Signed-off-by: NSteffen Klassert <steffen.klassert@secunet.com>
      1548bc4e
    • F
      xfrm: policy: increment xfrm_hash_generation on hash rebuild · 7a474c36
      由 Florian Westphal 提交于 
      Hash rebuild will re-set all the inexact entries, then re-insert them.
Lookups that can occur in parallel will therefore not find any policies.

This was safe when lookups were still guarded by rwlock.
After rcu-ification, lookups check the hash_generation seqcount to detect
when a hash resize takes place.  Hash rebuild missed the needed increment.

Hash resizes and hash rebuilds cannot occur in parallel (both acquire
hash_resize_mutex), so just increment xfrm_hash_generation, like resize.

Fixes: a7c44247 ("xfrm: policy: make xfrm_policy_lookup_bytype lockless")
Signed-off-by: NFlorian Westphal <fw@strlen.de>
Signed-off-by: NSteffen Klassert <steffen.klassert@secunet.com>
      7a474c36
    • F
      xfrm: policy: use hlist rcu variants on inexact insert, part 2 · 355b00d1
      由 Florian Westphal 提交于 
      This function was modeled on the 'exact' insert one, which did not use
the rcu variant either.

When I fixed the 'exact' insert I forgot to propagate this to my
development tree, so the inexact variant retained the bug.

Fixes: 9cf545eb ("xfrm: policy: store inexact policies in a tree ordered by destination address")
Signed-off-by: NFlorian Westphal <fw@strlen.de>
Signed-off-by: NSteffen Klassert <steffen.klassert@secunet.com>
      355b00d1
  15. 20 12月, 2018 1 次提交
  17. 19 12月, 2018 1 次提交
  19. 28 11月, 2018 1 次提交
    • F
      xfrm: policy: fix policy hash rebuild · 88584c30
      由 Florian Westphal 提交于 
      Dan Carpenter reports following static checker warning:
 net/xfrm/xfrm_policy.c:1316 xfrm_hash_rebuild()
 warn: 'dir' is out of bounds '3' vs '2'

 |  1280          /* reset the bydst and inexact table in all directions */
 |  1281          xfrm_hash_reset_inexact_table(net);
 |  1282
 |  1283          for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
 |                              ^^^^^^^^^^^^^^^^^^^^^
 |dir == XFRM_POLICY_MAX at the end of this loop.
 |  1304          /* re-insert all policies by order of creation */
 |  1305          list_for_each_entry_reverse(policy, &net->xfrm.policy_all, walk.all) {
 [..]
 |  1314                                            xfrm_policy_id2dir(policy->index));
 |  1315                  if (!chain) {
 |  1316                          void *p = xfrm_policy_inexact_insert(policy, dir, 0);

Fix this by updating 'dir' based on current policy.  Otherwise, the
inexact policies won't be found anymore during lookup, as they get
hashed to a bogus bin.
Reported-by: NDan Carpenter <dan.carpenter@oracle.com>
Fixes: cc1bb845 ("xfrm: policy: return NULL when inexact search needed")
Signed-off-by: NFlorian Westphal <fw@strlen.de>
Signed-off-by: NSteffen Klassert <steffen.klassert@secunet.com>
      88584c30
  21. 16 11月, 2018 2 次提交
  23. 09 11月, 2018 10 次提交
  25. 11 10月, 2018 1 次提交
  27. 11 9月, 2018 1 次提交
  29. 26 7月, 2018 1 次提交
  31. 20 7月, 2018 1 次提交
    • B
      xfrm: Remove xfrmi interface ID from flowi · bc56b334
      由 Benedict Wong 提交于 
      In order to remove performance impact of having the extra u32 in every
single flowi, this change removes the flowi_xfrm struct, prefering to
take the if_id as a method parameter where needed.

In the inbound direction, if_id is only needed during the
__xfrm_check_policy() function, and the if_id can be determined at that
point based on the skb. As such, xfrmi_decode_session() is only called
with the skb in __xfrm_check_policy().

In the outbound direction, the only place where if_id is needed is the
xfrm_lookup() call in xfrmi_xmit2(). With this change, the if_id is
directly passed into the xfrm_lookup_with_ifid() call. All existing
callers can still call xfrm_lookup(), which uses a default if_id of 0.

This change does not change any behavior of XFRMIs except for improving
overall system performance via flowi size reduction.

This change has been tested against the Android Kernel Networking Tests:

https://android.googlesource.com/kernel/tests/+/master/net/testSigned-off-by: NBenedict Wong <benedictwong@google.com>
Signed-off-by: NSteffen Klassert <steffen.klassert@secunet.com>
      bc56b334
  33. 11 7月, 2018 1 次提交
    • A
      xfrm: use time64_t for in-kernel timestamps · 386c5680
      由 Arnd Bergmann 提交于 
      The lifetime managment uses '__u64' timestamps on the user space
interface, but 'unsigned long' for reading the current time in the kernel
with get_seconds().

While this is probably safe beyond y2038, it will still overflow in 2106,
and the get_seconds() call is deprecated because fo that.

This changes the xfrm time handling to use time64_t consistently, along
with reading the time using the safer ktime_get_real_seconds(). It still
suffers from problems that can happen from a concurrent settimeofday()
call or (to a lesser degree) a leap second update, but since the time
stamps are part of the user API, there is nothing we can do to prevent
that.
Signed-off-by: NArnd Bergmann <arnd@arndb.de>
Signed-off-by: NSteffen Klassert <steffen.klassert@secunet.com>
      386c5680
  35. 25 6月, 2018 1 次提交
    • F
      xfrm: policy: remove pcpu policy cache · e4db5b61
      由 Florian Westphal 提交于 
      Kristian Evensen says:
  In a project I am involved in, we are running ipsec (Strongswan) on
  different mt7621-based routers. Each router is configured as an
  initiator and has around ~30 tunnels to different responders (running
  on misc. devices). Before the flow cache was removed (kernel 4.9), we
  got a combined throughput of around 70Mbit/s for all tunnels on one
  router. However, we recently switched to kernel 4.14 (4.14.48), and
  the total throughput is somewhere around 57Mbit/s (best-case). I.e., a
  drop of around 20%. Reverting the flow cache removal restores, as
  expected, performance levels to that of kernel 4.9.

When pcpu xdst exists, it has to be validated first before it can be
used.

A negative hit thus increases cost vs. no-cache.

As number of tunnels increases, hit rate decreases so this pcpu caching
isn't a viable strategy.

Furthermore, the xdst cache also needs to run with BH off, so when
removing this the bh disable/enable pairs can be removed too.

Kristian tested a 4.14.y backport of this change and reported
increased performance:

  In our tests, the throughput reduction has been reduced from around -20%
  to -5%. We also see that the overall throughput is independent of the
  number of tunnels, while before the throughput was reduced as the number
  of tunnels increased.
Reported-by: NKristian Evensen <kristian.evensen@gmail.com>
Signed-off-by: NFlorian Westphal <fw@strlen.de>
Signed-off-by: NSteffen Klassert <steffen.klassert@secunet.com>
      e4db5b61
  37. 23 6月, 2018 4 次提交