When doing parallel build on a fast machine with bottleneck in i/o,
m_xt.so may start linking faster than dynsyms.list gets populated,
resulting in error:
ld:dynsyms.list:0: syntax error in dynamic list
Fix this by adding dynsyms.list as make dependency to m_xt.so
Described also here:
https://bugs.openwrt.org/index.php?do=details&task_id=3353
Change from v1:
- add dynsysms.list dependancy only when shared libs are enabled
Signed-off-by: Roman Yeryomin <roman@advem.lv>
Fixes: FS#3353
(cherry-picked from commit edd53df16843a0a6380920ed17b88bfe7d26d71b)
Upstream iproute2 detects libbpf using a one-line $CC test-compile, which
normally ignores LDFLAGS. With NLS enabled however, LDFLAGS includes an
"rpath-link" linker option needed to resolve libintl.so. Its absence
causes both the compile and libbpf detection to fail:
ld: warning: libintl.so.8, needed by libbpf.so, not found (try using
-rpath or -rpath-link)
ld: libelf.so.1: undefined reference to `libintl_dgettext'
collect2: error: ld returned 1 exit status
Fix this by directly including $LDFLAGS in the test-compile command.
Reported-by: Ian Cooper <iancooper@hotmail.com>
Signed-off-by: Tony Ambardar <itugrok@yahoo.com>
(cherry picked from commit aab3a04ce87c27996b7ce9ea306db8c6eed86941)
This change was investigated previously [1] but not deemed necessary. With
the recent addition [2] of modern BPF loader support, however, tc gained
dependencies on libelf and libbpf, with a larger installation footprint.
Similar to ip-tiny/ip-full, split tc into tc-full and tc-tiny variants,
where the latter excludes the eBPF loader, uses a smaller executable, and
avoids libelf and libbpf package dependencies. Both variants provide the
'tc' virtual package, with tc-tiny as the default.
The previous tc package included a loadable module for iptables actions.
Separate this out into a common package, tc-mod-iptables, which both
variants depend on. Some package sizes on mips_24kc:
Before:
148343 tc_5.11.0-1_mips_24kc.ipk
After:
144833 tc-full_5.11.0-2_mips_24kc.ipk
138430 tc-tiny_5.11.0-2_mips_24kc.ipk (and no libelf or libbpf)
4115 tc-mod-iptables_5.11.0-2_mips_24kc.ipk
Also fix up some Makefile indentation.
[1] https://github.com/openwrt/openwrt/pull/1627#issuecomment-447619962
[2] b048a305a3d3 ("iproute2: update to 5.11.0")
Signed-off-by: Tony Ambardar <itugrok@yahoo.com>
(cherry picked from commit 72885e9608d9c0fa06abce6bdb6dd054c2116eb7)
The link equalizer sch_teql.ko of package kmod-sched relies on a hotplug
script historically included in iproute2's tc package. In previous
discussion [1], consensus was the hotplug script is best located together
with the module in kmod-sched, but this change was deferred at the time.
Relocate the hotplug script now. This change also simplifies adding a tc
variant for minimal size with reduced functionality.
[1] https://github.com/openwrt/openwrt/pull/1627#issuecomment-447923636
Signed-off-by: Tony Ambardar <itugrok@yahoo.com>
(cherry picked from commit 863ce4f15f74a674f01ab47f445b362c113b113b)
This patch has been submitted upstream to fix an error reported by a few
users. One instance seen using gcc 10.2.0, binutils 2.35.1 and musl 1.1.24:
bpf_glue.c: In function 'get_libbpf_version':
bpf_glue.c:46:11: error: 'PATH_MAX' undeclared (first use in this function);
did you mean 'AF_MAX'?
46 | char buf[PATH_MAX], *s;
| ^~~~~~~~
| AF_MAX
Reported-by: Rui Salvaterra <rsalvaterra@gmail.com>
Signed-off-by: Tony Ambardar <itugrok@yahoo.com>
(cherry picked from commit 10ffefe60299637bc592ec4ef0dde93e0e2e8344)
The latest iproute2 version brings various improvements and fixes:
https://git.kernel.org/pub/scm/network/iproute2/iproute2.git/log/?qt=range&q=v5.10.0..v5.11.0
In particular, ip and tc now use libbpf as the standard way to load BPF
programs, rather than the old, limited custom loader. This allows more
consistent and featureful BPF program handling e.g. support for global
initialized variables.
Also fix a longstanding problem with iproute2 builds where unneeded DSO
dependencies are added to most utilities, bloating their installation
footprint. From research and testing, explicitly using a "--as-needed"
linker flag avoids the issue. Update accordingly and drop extra package
dependencies from Makefile.
Additional build and packaging updates include:
- install missing development header to iproute2/bpf_elf.h
- propagate OpenWrt verbose flag during build
- update and refresh patches
Compile and run tested: QEMU/malta-mips32be on kernels 5.4 & 5.10.
All iproute2 packages were built and installed to the test image. Some
regression testing using ip-full and tc was successfully performed to
exercise several kmods, tc modules, and simple BPF programs.
Signed-off-by: Tony Ambardar <itugrok@yahoo.com>
(cherry picked from commit b048a305a3d3cbb5375e32757e70aa129b4c7c4f)
Update iproute2 to latest stable 5.9; for the changes see https://lwn.net/Articles/834755/
Refresh patches
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
Acked-by: Hauke Mehrtens <hauke@huake-m.de>
Feature detection doesn't recognize ipset v7 use on kernel v5.x systems
and thus disables the tc ematch function em_ipset.
- backport patch:
* 002-configure-support-ipset-v7.patch:
650591a7a70c configure: support ipset version 7 with kernel version 5
Fixes: 4e0c54bc5bc8 ("kernel: add support for kernel 5.4")
Signed-off-by: Tony Ambardar <itugrok@yahoo.com>
Recent iproute2 5.x versions modified the symbols resolved for plugins,
causing "tc .. action xt .." to fail. Update the list of symbols to fix.
Fixes: b61495409bb3 ("iproute2: tc: reduce size of dynamic symbol table")
Signed-off-by: Tony Ambardar <itugrok@yahoo.com>
In iproute2, some of object, header, source files are used in
qos-gargoyle.
qos-gargoyle will find these files in build_dir.
But if CONFIG_AUTOREMOVE is enable in SDK, an error "Header not found." will be raised.
So one way is that copying these files to staging_dir.
Allows proper usage of the ss tool. Otherwise, several errors and bad
data gets thrown:
Cannot open netlink socket: Protocol not supported
Cannot open netlink socket: Protocol not supported
Cannot open netlink socket: Protocol not supported
Cannot open netlink socket: Protocol not supported
Cannot open netlink socket: Protocol not supported
Cannot open netlink socket: Protocol not supported
Cannot open netlink socket: Protocol not supported
Originally reported here: https://github.com/openwrt/packages/issues/8232
Signed-off-by: Rosen Penev <rosenp@gmail.com>
This reverts commit a6da3f9ef746101b84a6f530f5a40de28341b69a.
The libcap isn't as optional as the commit messages suggests. A hard
dependency to the libcap package is added, which is only available in
the external packages feed. Therefore it is impossible to package
ip-full without having the external packages feed up and running, which
is a regression to the former behaviour.
Signed-off-by: Mathias Kresin <dev@kresin.me>
Acked-by: Hans Dedecker <dedeckeh@gmail.com>
I'm having another attempt at trying to getting the 'store dscp into
conntrack connmark' functionality into upstream kernel, since the
restore function (act_ctinfo) has been accepted.
The syntax has changed from 'savedscp' to 'set-dscpmark' since that
conforms more closely with existing functionality.
Signed-off-by: Kevin Darbyshire-Bryant <ldir@darbyshire-bryant.me.uk>
Update iproute2 to latest stable version, see https://lwn.net/Articles/805654/
for the changes in 5.4.0
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
Preserve optionality of libcap by having configuration script follow the
HAVE_CAP environment variable, used similarly to the HAVE_ELF variable.
Signed-off-by: Alin Nastac <alin.nastac@gmail.com>
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com> [PKG_RELEASE increase/refresh patches]
This changes the default PKG_BUILD_DIR to take BUILD_VARIANT into
account (if set), so that packages do not need to manually override
PKG_BUILD_DIR just to handle variants.
This also updates most base packages with variants to use the updated
default PKG_BUILD_DIR.
Signed-off-by: Jeffery To <jeffery.to@gmail.com>
Follow upstream changes - header file changes only
no functional or executable changes, hence no package bump
required
Signed-off-by: Kevin Darbyshire-Bryant <ldir@darbyshire-bryant.me.uk>
Add the userspace control portion of the backported kernelspace
act_ctinfo.
ctinfo is a tc action restoring data stored in conntrack marks to
various fields. At present it has two independent modes of operation,
restoration of DSCP into IPv4/v6 diffserv and restoration of conntrack
marks into packet skb marks.
It understands a number of parameters specific to this action in
additional to the usual action syntax. Each operating mode is
independent of the other so all options are optional, however not
specifying at least one mode is a bit pointless.
Usage: ... ctinfo [dscp mask [statemask]] [cpmark [mask]] [zone ZONE]
[CONTROL] [index <INDEX>]
DSCP mode
dscp enables copying of a DSCP stored in the conntrack mark into the
ipv4/v6 diffserv field. The mask is a 32bit field and specifies where
in the conntrack mark the DSCP value is located. It must be 6
contiguous bits long. eg. 0xfc000000 would restore the DSCP from the
upper 6 bits of the conntrack mark.
The DSCP copying may be optionally controlled by a statemask. The
statemask is a 32bit field, usually with a single bit set and must not
overlap the dscp mask. The DSCP restore operation will only take place
if the corresponding bit/s in conntrack mark ANDed with the statemask
yield a non zero result.
eg. dscp 0xfc000000 0x01000000 would retrieve the DSCP from the top 6
bits, whilst using bit 25 as a flag to do so. Bit 26 is unused in this
example.
CPMARK mode
cpmark enables copying of the conntrack mark to the packet skb mark. In
this mode it is completely equivalent to the existing act_connmark
action. Additional functionality is provided by the optional mask
parameter, whereby the stored conntrack mark is logically ANDed with the
cpmark mask before being stored into skb mark. This allows shared usage
of the conntrack mark between applications.
eg. cpmark 0x00ffffff would restore only the lower 24 bits of the
conntrack mark, thus may be useful in the event that the upper 8 bits
are used by the DSCP function.
Usage: ... ctinfo [dscp mask [statemask]] [cpmark [mask]] [zone ZONE]
[CONTROL] [index <INDEX>]
where :
dscp MASK is the bitmask to restore DSCP
STATEMASK is the bitmask to determine conditional restoring
cpmark MASK mask applied to restored packet mark
ZONE is the conntrack zone
CONTROL := reclassify | pipe | drop | continue | ok |
goto chain <CHAIN_INDEX>
Signed-off-by: Kevin Darbyshire-Bryant <ldir@darbyshire-bryant.me.uk>
Update iproute2 to 5.0.0
Remove upstream patch 001-tc-fix-undefined-XATTR_SIZE_MAX
Alter patch 170-ip_tiny as support for IPX and DECnet is dropped
Update patch 010-cake-fwmark to match upstream commit
Signed-off-by: Deng Qingfang <dengqf6@mail2.sysu.edu.cn>
In the case of SHARED_LIBS=y, don't use -export-dynamic to place *all*
symbols into the dynamic symbol table. Instead, use --dynamic-list to
export a smaller set of symbols similar to that defined in static-syms.h
in the case of SHARED_LIBS=n, avoiding an 11 KB tc package size increase.
The symbol set is based on that required by the only plugin, m_xt.so.
Also increment PKG_RELEASE.
Signed-off-by: Tony Ambardar <itugrok@yahoo.com>
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com> [PKG_RELEASE fixup]
This enables using the tc module m_xt.so, which uses the act_ipt kernel
module to allow tc actions based on iptables targets. e.g.
tc filter add dev eth0 parent 1: prio 10 protocol ip \
u32 match u32 0 0 action xt -j DSCP --set-dscp-class BE
Make the SHARED_LIBS parameter configurable and based on tc package
selection.
Fix a problem using the tc m_xt.so plugin as also described in
https://bugs.debian.org/868059:
Sync include/xtables.h from iptables to make sure the right offset is
used when accessing structure members defined in libxtables. One could
get “Extension does not know id …” otherwise. (See also: #868059)
Patch to sync the included xtables.h with system iptables 1.6.x. This
continues to work with iptables 1.8.2.
Signed-off-by: Tony Ambardar <itugrok@yahoo.com>
Add build and runtime dependencies on libelf, allowing tc and ip-full
to load BPF and XDP object files respectively.
Define package 'tc' as a singleton package variant, which can be used to
enable additional functionality limited only to tc. Also set ip-tiny
as the default 'ip' variant.
Preserve optionality of libelf by having configuration script follow the
HAVE_ELF environment variable, used similarly to the HAVE_MNL variable.
Signed-off-by: Tony Ambardar <itugrok@yahoo.com>
This reverts commit e6d84fa8864e7bbbcec139dd09a2922a06f5b2a0 as it breaks the
installation of the iproute2 utilities ip-bridge, ss, nstat, devlink and rdma
for the ip-full variant
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
This reverts commit 26681fa6a6fcbec0024906eb8367e9a3160521fb as it breaks the
installation of the iproute2 utilities ip-bridge, ss, nstat, devlink and rdma
for the ip-full variant
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
This reverts commit fc80ef3613465e60257a5ddf0674debe45b09180 as it breaks the
installation of the iproute2 utilities ip-bridge, ss, nstat, devlink and
rdma for the ip-full variant
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
This reverts commit 248797834bf21916ddf663edc96d86ee5377850e as it breaks the
installation of the iproute2 utilities ip-bridge, ss, nstat, devlink and rdma
for the ip-full variant
Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
In the case of SHARED_LIBS=y, don't use -export-dynamic to place *all*
symbols into the dynamic symbol table. Instead, use --dynamic-list to
export a smaller set of symbols similar to that defined in static-syms.h
in the case of SHARED_LIBS=n, avoiding an 11 KB tc package size increase.
Also increment PKG_RELEASE.
Signed-off-by: Tony Ambardar <itugrok@yahoo.com>
This enables using the tc module m_xt.so, which uses the act_ipt kernel
module to allow tc actions based on iptables targets. e.g.
tc filter add dev eth0 parent 1: prio 10 protocol ip \
u32 match u32 0 0 action xt -j DSCP --set-dscp-class BE
Make the SHARED_LIBS parameter configurable and based on tc package
selection.
Fix a problem using the tc m_xt.so plugin as also described in
https://bugs.debian.org/868059:
Sync include/xtables.h from iptables to make sure the right offset is
used when accessing structure members defined in libxtables. One could
get “Extension does not know id …” otherwise. (See also: #868059)
Patch to sync the included xtables.h with system iptables 1.6.x. This
continues to work with iptables 1.8.2.
Signed-off-by: Tony Ambardar <itugrok@yahoo.com>
Simplify build and runtime dependencies on libelf, which allows tc and ip
to load BPF and XDP object files respectively.
Preserve optionality of libelf by having configuration script follow the
HAVE_ELF environment variable, used similarly to the HAVE_MNL variable.
Signed-off-by: Tony Ambardar <itugrok@yahoo.com>
Replace the old 'tc' with a singleton package variant which will be used
to enable additional functionality and limit it only to tc. Non-variant
packages will only be installed during 'tiny' variant builds, hence will
be configured without extra features, thus preserving previously limited
functionality and reduced package sizes.
Also set ip-tiny as the default variant, and install 'tiny' versions of
development libraries.
Signed-off-by: Tony Ambardar <itugrok@yahoo.com>
Compile-based feature detection (e.g. xtables, ipset support) was broken
due to silent compilation errors in the configure script, caused by a
Makefile variable KERNEL_INCLUDE referring to kernel build headers. Use
userspace headers by setting the same "user_headers" kernel include path
as used for the iptables build.
Remove redundant or unused Build/Configure definitions from package
Makefile, including KERNEL_INCLUDE, LIBC_INCLUDE and DBM includes.
Don't pass LDFLAGS within MAKE_FLAGS as this interferes with LDFLAGS in
tc/Makefile and masks a link parameter ("-Wl,-export-dynamic"). Instead,
use standard TARGET_LDFLAGS.
Replace EXTRA_CCOPTS in MAKE_FLAGS with cleaner TARGET_CPPFLAGS, and also
drop now unneeded patch 150-extra-ccopts.patch.
Enable defining XT_LIB_DIR from Makefile, needed to set the iptables
modules directory to something other than /lib/xtables, and also add
libxtables dependency. Both are needed with working xtables detection.
Note that libxtables is also pulled in by iptables, firewall or luci, so
this change has no size impact in most cases.
Signed-off-by: Tony Ambardar <itugrok@yahoo.com>
