And cleanup some whitespaces in pktgen.txt.
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently the testsuite does not have a test case with a backward jump.
The s390x JIT (kernel 4.0) had a bug in that area.
So add one new test case for this now.
Signed-off-by: Michael Holzheu <holzheu@linux.vnet.ibm.com>
Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
This restored previous behaviour. If caller does not want ports to be
filled, we should not break.
Fixes: 06635a35d1 ("flow_dissect: use programable dissector in skb_flow_dissect and friends")
Signed-off-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: David S. Miller <davem@davemloft.net>
Jamal points out that this header also contains kernel internal magic that
cannot be used from userspace for anything meaningful.
Lets remove what the kernel doesn't use anymore and wrap remainder with
__KERNEL__.
Suggested-by: Jamal Hadi Salim <jhs@mojatatu.com>
Suggested-by: Alexei Starovoitov <alexei.starovoitov@gmail.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch tracks the total number of inbound and outbound segments on a
TCP socket. One may use this number to have an idea on connection
quality when compared against the retransmissions.
RFC4898 named these : tcpEStatsPerfSegsIn and tcpEStatsPerfSegsOut
These are a 32bit field each and can be fetched both from TCP_INFO
getsockopt() if one has a handle on a TCP socket, or from inet_diag
netlink facility (iproute2/ss patch will follow)
Note that tp->segs_out was placed near tp->snd_nxt for good data
locality and minimal performance impact, while tp->segs_in was placed
near tp->bytes_received for the same reason.
Join work with Eric Dumazet.
Note that received SYN are accounted on the listener, but sent SYNACK
are not accounted.
Signed-off-by: Marcelo Ricardo Leitner <mleitner@redhat.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
ip -6 addr add dead::1/128 dev eth0
sleep 5
ip -6 route add default via dead::1/128
-> fails
ip -6 addr add dead::1/128 dev eth0
ip -6 route add default via dead::1/128
-> succeeds
reason is that if (nonsensensical) route above is added,
dead::1 is still subject to DAD, so the route lookup will
pick eth0 as outdev due to the prefix route that is added before
DAD work is started.
Add explicit test that checks if nexthop gateway is a local address.
Link: https://bugzilla.redhat.com/show_bug.cgi?id=1167969
Signed-off-by: Florian Westphal <fw@strlen.de>
Acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Joachim Eastwood says:
====================
stmmac: probe code refactoring and clean up part 1
This patch set refactor the code in stmmac_pci_probe and stmmac_pltfr_probe
and moves the common bits into stmmac_dvr_probe. Along the way some clean-
ups are applied to stmmac_pltfr_probe.
The code has been tested on the LPC18xx platform.
I am still working on more refactoring of the platform probe code, hence
part 1, but I need some more time on this.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Since the caller already check the presence of a of_node there
is no need to repeat the check in stmmac_probe_config_dt.
There is also no point in checking the return value of the
of_match_device function since if there wasn't match in the
first place we would never be in this function.
Signed-off-by: Joachim Eastwood <manabian@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since stmmac_dvr_probe takes care of setting driver data and
assign resources to the priv structure there is no need to
access the priv structure from the other probe functions.
This mean that this function can be changed into just return
an int and thus simplifying the callers.
Signed-off-by: Joachim Eastwood <manabian@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Creat a struct that contain all the resources that needs to be
assigned to the priv struct in stmmac_dvr_probe. This makes it
possible to factor out more common code from the other probe
functions and also use this struct to hold the resources as
they are fetched.
Signed-off-by: Joachim Eastwood <manabian@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Move setting of driver data into stmmac_dvr_probe so the
other probe functions don't have to. This will help to
simplify the other probe functions later.
Signed-off-by: Joachim Eastwood <manabian@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Eric Dumazet says:
====================
tcp: improve source port selection
With increase of TCP sockets in hosts, we often hit limitations
caused by port selection, due to randomization and poor strategy.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
inet_csk_get_port() randomization effort tends to spread
sockets on all the available range (ip_local_port_range)
This is unfortunate because SO_REUSEADDR sockets have
less requirements than non SO_REUSEADDR ones.
If an application uses SO_REUSEADDR hint, it is to try to
allow source ports being shared.
So instead of picking a random port number in ip_local_port_range,
lets try first in first half of the range.
This gives more chances to use upper half of the range for the
sockets with strong requirements (not using SO_REUSEADDR)
Note this patch does not add a new sysctl, and only changes
the way we try to pick port number.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Marcelo Ricardo Leitner <mleitner@redhat.com>
Cc: Flavio Leitner <fbl@redhat.com>
Acked-by: Flavio Leitner <fbl@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We no longer need bsocket atomic counter, as inet_csk_get_port()
calls bind_conflict() regardless of its value, after commit
2b05ad33e1 ("tcp: bind() fix autoselection to share ports")
This patch removes overhead of maintaining this counter and
double inet_csk_get_port() calls under pressure.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Marcelo Ricardo Leitner <mleitner@redhat.com>
Cc: Flavio Leitner <fbl@redhat.com>
Acked-by: Flavio Leitner <fbl@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We currently rely on the setting of SOCK_NOSPACE in the write()
path to ensure that we wake up any epoll edge trigger waiters when
acks return to free space in the write queue. However, if we fail
to allocate even a single skb in the write queue, we could end up
waiting indefinitely.
Fix this by explicitly issuing a wakeup when we detect the condition
of an empty write queue and a return value of -EAGAIN. This allows
userspace to re-try as we expect this to be a temporary failure.
I've tested this approach by artificially making
sk_stream_alloc_skb() return NULL periodically. In that case,
epoll edge trigger waiters will hang indefinitely in epoll_wait()
without this patch.
Signed-off-by: Jason Baron <jbaron@akamai.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Hariprasad Shenai says:
====================
cxgb4: Cleanup and update T4/T4 register ranges
This series cleans and optimizes setup_memwin function and also updates
T4/T5 adapter register ranges by removing incorrect register addresses
This patch series has been created against net-next tree and includes
patches on cxgb4 driver.
We have included all the maintainers of respective drivers. Kindly review
the change and let us know in case of any review comments.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Remove some T4/T5 registers that were included incorrectly.
Signed-off-by: Hariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Shradha Shah says:
====================
sfc: Get/Set MAC address and ndo_[set/get]_vf_* entrypoint functions
This is the second installment of patches towards supporting EF10 SRIOV.
This patch series implements the ndo_get_vf_config, ndo_set_vf_mac,
ndo_set_vf_vlan and ndo_set_vf_spoofcheck function callbacks for EF10.
This patch series also introduces privileges for the MCDI commands
based on which functions are allowed to call them, i.e. Link control
or primary function.
The patch series has been tested with and without CONFIG_SFC_SRIOV.
The ndo function callbacks are tested using ip link.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Add a set_mac_address() NIC-type function for EF10 only, and
use this to set the MAC address on the vadaptor. For Siena and
earlier, the MAC address continues to be set by MC_CMD_SET_MAC;
this is still called on EF10, and including a MAC address in
this command has no effect.
The sriov_mac_address_changed() NIC-type function is no longer
needed on EF10, but it is needed for Siena where it is used to
update the peer address of the PF for VFDI. Change this to use
the new set_mac_address function pointer.
efx_ef10_sriov_mac_address_changed() is no longer called, as VFs
will try to change the MAC address on their vadaptor rather than
trying to change to the context of the PF to alter the vport.
When a VF is running in direct passthrough mode with MAC spoofing
enabled, it will be able to change the MAC address on its vadaptor.
In this case, there is a link to the PF, so find the correct VF in
its ef10_vf array and update the MAC address.
ndo_set_mac_address() can be called during driver unload while
bonding, and in this case the device has already been stopped, so
don't call efx_net_open() to restart it after reconfiguration.
efx->port_enabled is set to false in efx_stop_port(), so it is
indicator of whether the device needs to be restarted.
Signed-off-by: Shradha Shah <sshah@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Exercised with
"ip link set <PF intf> vf <vf_i> state {auto|enable|disable}"
Sets the reporting policy for VF link state to either
- mirror physical link state
- always up
- always down
get VF link state mode in efx_ef10_sriov_get_vf_config
Exercised by
"ip link show <PF intf>";
output will include a line like
vf 0 MAC 12:34:56:78:9a:bc, link-state auto
Signed-off-by: Shradha Shah <sshah@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The max vlan tags that can be offloaded is 2, including any upstream VLAN
aggregator. Currently there is no way for the net driver to know whether
the upstream vswitch (if any) is using vlan tags, so there is no way to
know how many tags we can request.
Along with the implementation for the ndo_set_vf_vlan callback, this patch
also adds 2 VLAN tags for the driver created VEB switch if possible, that
way it is possible to offload as many tags as are allowed.
Signed-off-by: Shradha Shah <sshah@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently we do an entity reset when we detect an MC reboot.
This messes up SRIOV because it leaves VFs orphaned. The extra
reset is rather redundant anyway, since the MC reboot will have
basically reset everything.
This change replaces the entity reset after MC reboot with a
simpler datapath reset that reallocates resources but doesn't
perform the entity reset.
Signed-off-by: Shradha Shah <sshah@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
rtnetlink calls ndo_get_vf_config when compiling information
about a network interface, so that the VFs associated with a PF
can be listed (eg: ip link show).
Implement a response to this entry point and return PF-set MAC
address for VF in ndo_get_vf_config
Signed-off-by: Shradha Shah <sshah@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Implement a response to this entrypoint.
The ndo_set_vf_mac() entrypoint is only exposed in the driver if
CONFIG_SFC_SRIOV is defined.
Signed-off-by: Shradha Shah <sshah@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In order to avoid MC bugs the flags field needs to be set to 0.
Instead of explicitly clearing out the flags individually, a
better way to do this is to memset the MCDI_BUF to 0.
Signed-off-by: Shradha Shah <sshah@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
A VF's MAC address is set by its parent PF and added to its vport.
To get this MAC address, the VF must use MC_CMD_ VPORT_GET_MAC_ADDRESSES.
In the current scheme, a VF's vport should only have one MAC address,
so warn if this is not the case.
Signed-off-by: Shradha Shah <sshah@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If MCDI timeouts are encountered during efx_ef10_filter_table_remove(),
an FLR will be queued, but efx->filter_state will still be kfree()d.
The queued FLR will then call efx_ef10_filter_table_restore(), which
will try to use efx->filter_state. This previously caused a panic.
This patch adds an rwsem to protect the existence of efx->filter_state,
separately from the spinlock protecting its contents. Users which can
race against efx_ef10_filter_table_remove() should down_read this rwsem.
Signed-off-by: Shradha Shah <sshah@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Initialised in efx_probe_vf and removal is dealt with in
efx_ef10_remove.
vf->efx is needed in future patches to change the MAC address
of the VF via the parent PF, while the driver is bound to the
VF.
Example: ip link set dev vf NUM mac LLADDR
Signed-off-by: Shradha Shah <sshah@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Otherwise the PF and VF can disagree on the VF's MAC address and
this leads to strange behaviour, up to and including kernel panics.
Signed-off-by: Shradha Shah <sshah@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Added function efx_ef10_get_vf_index to store the vf_index
in nic_data during probe
vf_index is needed in future patches to access a particular
VF in the VF data structure.
Moved efx_ef10_probe_pf and efx_ef10_probe_vf in order to
used efx_ef10_remove
Signed-off-by: Shradha Shah <sshah@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
MC_CMD_SET_MAC is privileged and can only by called by the link
control function.
This patch adds efx_ef10_mac_reconfigure_vf which avoids the call
to MC_CMD_SET_MAC by the Virtual function
Signed-off-by: Shradha Shah <sshah@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There is one primary function per adaptor, one link control function
per port and the rest as categorised as general.
This patch adds privileges to the MCDI commands based on which
functions are allowed to call them.
Signed-off-by: Shradha Shah <sshah@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This also matches with the sibling call netdev_alloc_skb_ip_align() made in
rx fast path.
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
30 usecs (or really, 1 jiffy) can go by pretty fast.
Move the set of the timeout immediately before the loop.
Remove the unnecessary max(1ul, usecs_to_jiffies(30)) as
usecs_to_jiffies with a non-zero constant is guaranteed
to be non-zero.
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Simon Horman says:
====================
rocker: transaction fixes
this series addresses what appear to be errors in the handling of
prepare and then commit transactions in the rocker driver.
In all cases the problem is that data structures visible outside of
the transaction are modified during the prepare phase.
In the case of the first two patches this results in the kernel reporting a
BUG. I have noted test-cases in the change logs.
The third patch is also a bug fix, as noted by Toshiaki Makita,
however I have not been able to reliably reproduce the problem and
thus have not provided a test case.
The last patch is a correctness fix that does not fix a bug
that manifests as far as I can tell.
Changes: v3->v4
* All patches
- Add Jiri Pirko's ack
* "rocker: do not make neighbour entry changes when preparing transactions"
- Setting of entry values in all transaction phases
as suggested by Toshiaki Makita
* "rocker: make rocker_port_internal_vlan_id_{get,put}() non-transactional"
- Remove Fixes tag as I believe this is a correctness rather than a bug fix
Changes: v2->v3
* "rocker: do not make neighbour entry changes when preparing transactions"
- Correct inverted logic
- Added ack from Scott Feldman
Changes: v1->v2
* "rocker: do not make neighbour entry changes when preparing transactions"
- Revised changelog to reflect information from Toshiaki Makita
that there is a bug that can manifest
- Update address and ttl regardless of the value of the transaction state
* All other patches
- Added acks from Scott Feldman
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
The motivation for this is that rocker_port_internal_vlan_id_{get,put} appear
to only partially implement the transaction model: memory allocation
and freeing is transactional, but hash and bitmap manipulation is not.
The latter could be fixed, however, as it is not currently exercised
due to trans always being SWITCHDEV_TRANS_NONE it seems cleaner
to make rocker_port_internal_vlan_id_get non-transactional.
This problem was introduced by c4f20321d9 ("rocker: support
prepare-commit transaction model").
Found by inspection.
I do not believe that this change should have any run-time effect.
Acked-by: Scott Feldman <sfeldma@gmail.com>
Acked-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
rocker_port_ipv4_nh() and in turn rocker_port_ipv4_neigh() may be
be called with trans == SWITCHDEV_TRANS_PREPARE and then
trans == SWITCHDEV_TRANS_COMMIT from switchdev_port_obj_set() via
fib_table_insert().
The first time that rocker_port_ipv4_nh() is called, with
trans == SWITCHDEV_TRANS_PREPARE, _rocker_neigh_add() adds a new entry to
the neigh table.
And the second time rocker_port_ipv4_nh() is called, with
trans == SWITCHDEV_TRANS_COMMIT, that entry is found. This causes
rocker_port_ipv4_nh() to believe it is not adding an entry and thus it
frees "entry", which is still present in rocker driver's neigh table.
This problem does not appear to affect deletion as my analysis is that
deletion is always performed with trans == SWITCHDEV_TRANS_NONE.
For completeness _rocker_neigh_{add,del,prepare} are updated not to
manipulate fib table entries if trans == SWITCHDEV_TRANS_PREPARE.
Fixes: c4f20321d9 ("rocker: support prepare-commit transaction model")
Reported-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
Acked-by: Scott Feldman <sfeldma@gmail.com>
Acked-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: Simon Horman <simon.horman@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use the generic mechanism to declare a bitmap instead of unsigned long.
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Alexei Starovoitov says:
====================
bpf: introduce bpf_tail_call() helper
introduce bpf_tail_call(ctx, &jmp_table, index) helper function
which can be used from BPF programs like:
int bpf_prog(struct pt_regs *ctx)
{
...
bpf_tail_call(ctx, &jmp_table, index);
...
}
that is roughly equivalent to:
int bpf_prog(struct pt_regs *ctx)
{
...
if (jmp_table[index])
return (*jmp_table[index])(ctx);
...
}
The important detail that it's not a normal call, but a tail call.
The kernel stack is precious, so this helper reuses the current
stack frame and jumps into another BPF program without adding
extra call frame.
It's trivially done in interpreter and a bit trickier in JITs.
Use cases:
- simplify complex programs
- dispatch into other programs
(for example: index in jump table can be syscall number or network protocol)
- build dynamic chains of programs
The chain of tail calls can form unpredictable dynamic loops therefore
tail_call_cnt is used to limit the number of calls and currently is set to 32.
patch 1 - support bpf_tail_call() in interpreter
patch 2 - support in x64 JIT
We've discussed what's neccessary to support it in arm64/s390 JITs
and it looks fine.
patch 3 - sample example for tracing
patch 4 - sample example for networking
More details in every patch.
This set went through several iterations of reviews/fixes and older
attempts can be seen:
https://git.kernel.org/cgit/linux/kernel/git/ast/bpf.git/log/?h=tail_call_v[123456]
- tail_call_v1 does it without touching JITs but introduces overhead
for all programs that don't use this helper function.
- tail_call_v2 still has some overhead and x64 JIT does full stack
unwind (prologue skipping optimization wasn't there)
- tail_call_v3 reuses 'call' instruction encoding and has interpreter
overhead for every normal call
- tail_call_v4 fixes above architectural shortcomings and v5,v6 fix few
more bugs
This last tail_call_v6 approach seems to be the best.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Usage:
$ sudo ./sockex3
IP src.port -> dst.port bytes packets
127.0.0.1.42010 -> 127.0.0.1.12865 1568 8
127.0.0.1.59526 -> 127.0.0.1.33778 11422636 173070
127.0.0.1.33778 -> 127.0.0.1.59526 11260224828 341974
127.0.0.1.12865 -> 127.0.0.1.42010 1832 12
IP src.port -> dst.port bytes packets
127.0.0.1.42010 -> 127.0.0.1.12865 1568 8
127.0.0.1.59526 -> 127.0.0.1.33778 23198092 351486
127.0.0.1.33778 -> 127.0.0.1.59526 22972698518 698616
127.0.0.1.12865 -> 127.0.0.1.42010 1832 12
this example is similar to sockex2 in a way that it accumulates per-flow
statistics, but it does packet parsing differently.
sockex2 inlines full packet parser routine into single bpf program.
This sockex3 example have 4 independent programs that parse vlan, mpls, ip, ipv6
and one main program that starts the process.
bpf_tail_call() mechanism allows each program to be small and be called
on demand potentially multiple times, so that many vlan, mpls, ip in ip,
gre encapsulations can be parsed. These and other protocol parsers can
be added or removed at runtime. TLVs can be parsed in similar manner.
Note, tail_call_cnt dynamic check limits the number of tail calls to 32.
Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
bpf_tail_call() arguments:
ctx - context pointer
jmp_table - one of BPF_MAP_TYPE_PROG_ARRAY maps used as the jump table
index - index in the jump table
In this implementation x64 JIT bypasses stack unwind and jumps into the
callee program after prologue, so the callee program reuses the same stack.
The logic can be roughly expressed in C like:
u32 tail_call_cnt;
void *jumptable[2] = { &&label1, &&label2 };
int bpf_prog1(void *ctx)
{
label1:
...
}
int bpf_prog2(void *ctx)
{
label2:
...
}
int bpf_prog1(void *ctx)
{
...
if (tail_call_cnt++ < MAX_TAIL_CALL_CNT)
goto *jumptable[index]; ... and pass my 'ctx' to callee ...
... fall through if no entry in jumptable ...
}
Note that 'skip current program epilogue and next program prologue' is
an optimization. Other JITs don't have to do it the same way.
>From safety point of view it's valid as well, since programs always
initialize the stack before use, so any residue in the stack left by
the current program is not going be read. The same verifier checks are
done for the calls from the kernel into all bpf programs.
Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
introduce bpf_tail_call(ctx, &jmp_table, index) helper function
which can be used from BPF programs like:
int bpf_prog(struct pt_regs *ctx)
{
...
bpf_tail_call(ctx, &jmp_table, index);
...
}
that is roughly equivalent to:
int bpf_prog(struct pt_regs *ctx)
{
...
if (jmp_table[index])
return (*jmp_table[index])(ctx);
...
}
The important detail that it's not a normal call, but a tail call.
The kernel stack is precious, so this helper reuses the current
stack frame and jumps into another BPF program without adding
extra call frame.
It's trivially done in interpreter and a bit trickier in JITs.
In case of x64 JIT the bigger part of generated assembler prologue
is common for all programs, so it is simply skipped while jumping.
Other JITs can do similar prologue-skipping optimization or
do stack unwind before jumping into the next program.
bpf_tail_call() arguments:
ctx - context pointer
jmp_table - one of BPF_MAP_TYPE_PROG_ARRAY maps used as the jump table
index - index in the jump table
Since all BPF programs are idenitified by file descriptor, user space
need to populate the jmp_table with FDs of other BPF programs.
If jmp_table[index] is empty the bpf_tail_call() doesn't jump anywhere
and program execution continues as normal.
New BPF_MAP_TYPE_PROG_ARRAY map type is introduced so that user space can
populate this jmp_table array with FDs of other bpf programs.
Programs can share the same jmp_table array or use multiple jmp_tables.
The chain of tail calls can form unpredictable dynamic loops therefore
tail_call_cnt is used to limit the number of calls and currently is set to 32.
Use cases:
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
==========
- simplify complex programs by splitting them into a sequence of small programs
- dispatch routine
For tracing and future seccomp the program may be triggered on all system
calls, but processing of syscall arguments will be different. It's more
efficient to implement them as:
int syscall_entry(struct seccomp_data *ctx)
{
bpf_tail_call(ctx, &syscall_jmp_table, ctx->nr /* syscall number */);
... default: process unknown syscall ...
}
int sys_write_event(struct seccomp_data *ctx) {...}
int sys_read_event(struct seccomp_data *ctx) {...}
syscall_jmp_table[__NR_write] = sys_write_event;
syscall_jmp_table[__NR_read] = sys_read_event;
For networking the program may call into different parsers depending on
packet format, like:
int packet_parser(struct __sk_buff *skb)
{
... parse L2, L3 here ...
__u8 ipproto = load_byte(skb, ... offsetof(struct iphdr, protocol));
bpf_tail_call(skb, &ipproto_jmp_table, ipproto);
... default: process unknown protocol ...
}
int parse_tcp(struct __sk_buff *skb) {...}
int parse_udp(struct __sk_buff *skb) {...}
ipproto_jmp_table[IPPROTO_TCP] = parse_tcp;
ipproto_jmp_table[IPPROTO_UDP] = parse_udp;
- for TC use case, bpf_tail_call() allows to implement reclassify-like logic
- bpf_map_update_elem/delete calls into BPF_MAP_TYPE_PROG_ARRAY jump table
are atomic, so user space can build chains of BPF programs on the fly
Implementation details:
=======================
- high performance of bpf_tail_call() is the goal.
It could have been implemented without JIT changes as a wrapper on top of
BPF_PROG_RUN() macro, but with two downsides:
. all programs would have to pay performance penalty for this feature and
tail call itself would be slower, since mandatory stack unwind, return,
stack allocate would be done for every tailcall.
. tailcall would be limited to programs running preempt_disabled, since
generic 'void *ctx' doesn't have room for 'tail_call_cnt' and it would
need to be either global per_cpu variable accessed by helper and by wrapper
or global variable protected by locks.
In this implementation x64 JIT bypasses stack unwind and jumps into the
callee program after prologue.
- bpf_prog_array_compatible() ensures that prog_type of callee and caller
are the same and JITed/non-JITed flag is the same, since calling JITed
program from non-JITed is invalid, since stack frames are different.
Similarly calling kprobe type program from socket type program is invalid.
- jump table is implemented as BPF_MAP_TYPE_PROG_ARRAY to reuse 'map'
abstraction, its user space API and all of verifier logic.
It's in the existing arraymap.c file, since several functions are
shared with regular array map.
Signed-off-by: Alexei Starovoitov <ast@plumgrid.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Reduce ifdef pollution slightly, no functional change. We can simply
remove the extra alternative definition of handle_ing() and nf_ingress().
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
In commit 8e4d980ac2 ("tcp: fix behavior for epoll edge trigger")
we fixed a possible hang of TCP sockets under memory pressure,
by allowing sk_stream_alloc_skb() to use sk_forced_mem_schedule()
if no packet is in socket write queue.
It turns out there are other cases where we want to force memory
schedule :
tcp_fragment() & tso_fragment() need to split a big TSO packet into
two smaller ones. If we block here because of TCP memory pressure,
we can effectively block TCP socket from sending new data.
If no further ACK is coming, this hang would be definitive, and socket
has no chance to effectively reduce its memory usage.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
