[ upstream commit be95a845cc4402272994ce290e3ad928aff06cb9 ]
In addition to commit b2157399cc98 ("bpf: prevent out-of-bounds
speculation") also change the layout of struct bpf_map such that
false sharing of fast-path members like max_entries is avoided
when the maps reference counter is altered. Therefore enforce
them to be placed into separate cachelines.
pahole dump after change:
struct bpf_map {
const struct bpf_map_ops * ops; /* 0 8 */
struct bpf_map * inner_map_meta; /* 8 8 */
void * security; /* 16 8 */
enum bpf_map_type map_type; /* 24 4 */
u32 key_size; /* 28 4 */
u32 value_size; /* 32 4 */
u32 max_entries; /* 36 4 */
u32 map_flags; /* 40 4 */
u32 pages; /* 44 4 */
u32 id; /* 48 4 */
int numa_node; /* 52 4 */
bool unpriv_array; /* 56 1 */
/* XXX 7 bytes hole, try to pack */
/* --- cacheline 1 boundary (64 bytes) --- */
struct user_struct * user; /* 64 8 */
atomic_t refcnt; /* 72 4 */
atomic_t usercnt; /* 76 4 */
struct work_struct work; /* 80 32 */
char name[16]; /* 112 16 */
/* --- cacheline 2 boundary (128 bytes) --- */
/* size: 128, cachelines: 2, members: 17 */
/* sum members: 121, holes: 1, sum holes: 7 */
};
Now all entries in the first cacheline are read only throughout
the life time of the map, set up once during map creation. Overall
struct size and number of cachelines doesn't change from the
reordering. struct bpf_map is usually first member and embedded
in map structs in specific map implementations, so also avoid those
members to sit at the end where it could potentially share the
cacheline with first map values e.g. in the array since remote
CPUs could trigger map updates just as well for those (easily
dirtying members like max_entries intentionally as well) while
having subsequent values in cache.
Quoting from Google's Project Zero blog [1]:
Additionally, at least on the Intel machine on which this was
tested, bouncing modified cache lines between cores is slow,
apparently because the MESI protocol is used for cache coherence
[8]. Changing the reference counter of an eBPF array on one
physical CPU core causes the cache line containing the reference
counter to be bounced over to that CPU core, making reads of the
reference counter on all other CPU cores slow until the changed
reference counter has been written back to memory. Because the
length and the reference counter of an eBPF array are stored in
the same cache line, this also means that changing the reference
counter on one physical CPU core causes reads of the eBPF array's
length to be slow on other physical CPU cores (intentional false
sharing).
While this doesn't 'control' the out-of-bounds speculation through
masking the index as in commit b2157399cc98, triggering a manipulation
of the map's reference counter is really trivial, so lets not allow
to easily affect max_entries from it.
Splitting to separate cachelines also generally makes sense from
a performance perspective anyway in that fast-path won't have a
cache miss if the map gets pinned, reused in other progs, etc out
of control path, thus also avoids unintentional false sharing.
[1] https://googleprojectzero.blogspot.ch/2018/01/reading-privileged-memory-with-side.html
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 4ee806d51176ba7b8ff1efd81f271d7252e03a1d ] When a tcp socket is closed, if it detects that its net namespace is exiting, close immediately and do not wait for FIN sequence. For normal sockets, a reference is taken to their net namespace, so it will never exit while the socket is open. However, kernel sockets do not take a reference to their net namespace, so it may begin exiting while the kernel socket is still open. In this case if the kernel socket is a tcp socket, it will stay open trying to complete its close sequence. The sock's dst(s) hold a reference to their interface, which are all transferred to the namespace's loopback interface when the real interfaces are taken down. When the namespace tries to take down its loopback interface, it hangs waiting for all references to the loopback interface to release, which results in messages like: unregister_netdevice: waiting for lo to become free. Usage count = 1 These messages continue until the socket finally times out and closes. Since the net namespace cleanup holds the net_mutex while calling its registered pernet callbacks, any new net namespace initialization is blocked until the current net namespace finishes exiting. After this change, the tcp socket notices the exiting net namespace, and closes immediately, releasing its dst(s) and their reference to the loopback interface, which lets the net namespace continue exiting. Link: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1711407 Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=97811 Signed-off-by: Dan Streetman <ddstreet@canonical.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit cd9ff4de0107c65d69d02253bb25d6db93c3dbc1 ] Map all lookup neigh keys to INADDR_ANY for loopback/point-to-point devices to avoid making an entry for every remote ip the device needs to talk to. This used the be the old behavior but became broken in a263b3093641f (ipv4: Make neigh lookups directly in output packet path) and later removed in 0bb4087cbec0 (ipv4: Fix neigh lookup keying over loopback/point-to-point devices) because it was broken. Signed-off-by: Jim Westfall <jwestfall@surrealistic.net> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit d9b3fca27385eafe61c3ca6feab6cb1e7dc77482 upstream. tcp_hdrlen is wasteful if you already have a pointer to struct tcphdr. This splits the size calculation into a helper function that can be used if a struct tcphdr is already available. Signed-off-by: Craig Gallek <kraig@google.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit e9191ffb65d8e159680ce0ad2224e1acbde6985c ]
Commit 513674b5a2c9 ("net: reevalulate autoflowlabel setting after
sysctl setting") removed the initialisation of
ipv6_pinfo::autoflowlabel and added a second flag to indicate
whether this field or the net namespace default should be used.
The getsockopt() handling for this case was not updated, so it
currently returns 0 for all sockets for which IPV6_AUTOFLOWLABEL is
not explicitly enabled. Fix it to return the effective value, whether
that has been set at the socket or net namespace level.
Fixes: 513674b5a2c9 ("net: reevalulate autoflowlabel setting after sysctl ...")
Signed-off-by: Ben Hutchings <ben.hutchings@codethink.co.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 7e040726850a106587485c21bdacc0bfc8a0cbed upstream. [resend due to me forgetting to cc: linux-api the first time around I posted these back on Feb 23] From: Greg Kroah-Hartman <gregkh@linuxfoundation.org> For some reason these values are not in the uapi header file, so any libc has to define it themselves. To prevent them from needing to do this, just have the kernel provide the correct values. Reported-by: Elliott Hughes <enh@google.com> Signed-off-by: Greg Hackmann <ghackmann@google.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 5132ede0fe8092b043dae09a7cc32b8ae7272baa upstream.
This reverts commit 6cfb521ac0d5b97470883ff9b7facae264b7ab12.
Turns out distros do not want to make retpoline as part of their "ABI",
so this patch should not have been merged. Sorry Andi, this was my
fault, I suggested it when your original patch was the "correct" way of
doing this instead.
Reported-by: Jiri Kosina <jikos@kernel.org>
Fixes: 6cfb521ac0d5 ("module: Add retpoline tag to VERMAGIC")
Acked-by: Andi Kleen <ak@linux.intel.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: David Woodhouse <dwmw@amazon.co.uk>
Cc: rusty@rustcorp.com.au
Cc: arjan.van.de.ven@intel.com
Cc: jeyu@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 92b4423e3a0bc5d43ecde4bcad871f8b5ba04efd upstream. This is a forward-port of the original patch from Andrzej Hajda, he said: "IS_ERR_VALUE should be used only with unsigned long type. Otherwise it can work incorrectly. To achieve this function xt_percpu_counter_alloc is modified to return unsigned long, and its result is assigned to temporary variable to perform error checking, before assigning to .pcnt field. The patch follows conclusion from discussion on LKML [1][2]. [1]: http://permalink.gmane.org/gmane.linux.kernel/2120927 [2]: http://permalink.gmane.org/gmane.linux.kernel/2150581" Original patch from Andrzej is here: http://patchwork.ozlabs.org/patch/582970/ This patch has clashed with input validation fixes for x_tables. Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Acked-by: Michal Kubecek <mkubecek@suse.cz>
commit f4dc77713f8016d2e8a3295e1c9c53a21f296def upstream.
The dummy ruleset I used to test the original validation change was broken,
most rules were unreachable and were not tested by mark_source_chains().
In some cases rulesets that used to load in a few seconds now require
several minutes.
sample ruleset that shows the behaviour:
echo "*filter"
for i in $(seq 0 100000);do
printf ":chain_%06x - [0:0]\n" $i
done
for i in $(seq 0 100000);do
printf -- "-A INPUT -j chain_%06x\n" $i
printf -- "-A INPUT -j chain_%06x\n" $i
printf -- "-A INPUT -j chain_%06x\n" $i
done
echo COMMIT
[ pipe result into iptables-restore ]
This ruleset will be about 74mbyte in size, with ~500k searches
though all 500k[1] rule entries. iptables-restore will take forever
(gave up after 10 minutes)
Instead of always searching the entire blob for a match, fill an
array with the start offsets of every single ipt_entry struct,
then do a binary search to check if the jump target is present or not.
After this change ruleset restore times get again close to what one
gets when reverting 36472341017529e (~3 seconds on my workstation).
[1] every user-defined rule gets an implicit RETURN, so we get
300k jumps + 100k userchains + 100k returns -> 500k rule entries
Fixes: 36472341017529e ("netfilter: x_tables: validate targets of jumps")
Reported-by: Jeff Wu <wujiafu@gmail.com>
Tested-by: Jeff Wu <wujiafu@gmail.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Acked-by: Michal Kubecek <mkubecek@suse.cz>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit fac51482577d5e05bbb0efa8d602a3c2111098bf upstream. With CONFIG_OF enabled on x86, we get the following error on boot: " Failed to find cpu0 device node Unable to detect cache hierarchy from DT for CPU 0 " and the cacheinfo fails to get populated in the corresponding sysfs entries. This is because cache_setup_of_node looks for of_node for setting up the shared cpu_map without checking that it's already populated in the architecture specific callback. In order to indicate that the shared cpu_map is already populated, this patch introduces a boolean `cpu_map_populated` in struct cpu_cacheinfo that can be used by the generic code to skip cache_shared_cpu_map_setup. This patch also sets that boolean for x86. Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Sudeep Holla <sudeep.holla@arm.com> Signed-off-by: Mian Yousaf Kaukab <yousaf.kaukab@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>