When DEBUG is defined this error occurs
drivers/net/ethernet/hisilicon/hns/hns_enet.c:1505:36: error:
‘struct net_device’ has no member named ‘ae_handle’;
did you mean ‘rx_handler’?
assert(skb->queue_mapping < ndev->ae_handle->q_num);
^~~~~~~~~
ae_handle is an element of struct hns_nic_priv, so change
ndev to priv.
Signed-off-by: Tom Rix <trix@redhat.com>
Link: https://lore.kernel.org/r/20210117191044.533725-1-trix@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When DEBUG is defined this error occurs
drivers/net/arcnet/com20020_cs.c:70:15: error: ‘com20020_REG_W_ADDR_HI’
undeclared (first use in this function);
did you mean ‘COM20020_REG_W_ADDR_HI’?
ioaddr, com20020_REG_W_ADDR_HI);
^~~~~~~~~~~~~~~~~~~~~~
From reviewing the context, the suggestion is what is meant.
Signed-off-by: Tom Rix <trix@redhat.com>
Acked-by: Joe Perches <joe@perches.com>
Link: https://lore.kernel.org/r/20210117181519.527625-1-trix@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Tariq Toukan says:
====================
TLS device offload for Bond
This series opens TX and RX TLS device offload for bond interfaces.
This allows bond interfaces to benefit from capable lower devices.
We add a new ndo_sk_get_lower_dev() to be used to get the lower dev that
corresponds to a given socket.
The TLS module uses it to interact directly with the lowest device in
chain, and invoke the control operations in tlsdev_ops. This means that the
bond interface doesn't have his own struct tlsdev_ops instance and
derived logic/callbacks.
To keep simple track of the HW and SW TLS contexts, we bind each socket to
a specific lower device for the socket's whole lifetime. This is logically
valid (and similar to the SW kTLS behavior) in the following bond configuration,
so we restrict the offload support to it:
((mode == balance-xor) or (mode == 802.3ad))
and xmit_hash_policy == layer3+4.
In this design, TLS TX/RX offload feature flags of the bond device are
independent from the lower devices. They reflect the current features state,
but are not directly controllable.
This is because the bond driver is bypassed by the call to
ndo_sk_get_lower_dev(), without him knowing who the caller is.
The bond TLS feature flags are set/cleared only according to the configuration
of the mode and xmit_hash_policy.
Bypass is true only for the control flow. Packets in fast path still go through
the bond logic.
The design here differs from the xfrm/ipsec offload, where the bond driver
has his own copy of struct xfrmdev_ops and callbacks.
====================
Link: https://lore.kernel.org/r/20210117145949.8632-1-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
In the tls_dev_event handler, ignore tlsdev_ops requirement for bond
interfaces, they do not exist as the interaction is done directly with
the lower device.
Also, make the validate function pass when it's called with the upper
bond interface.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Boris Pismenny <borisp@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Do not call the tls_dev_ops of upper devices. Instead, ask them
for the proper lowest device and communicate with it directly.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Boris Pismenny <borisp@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Following the description in previous patch (for TX):
As the bond interface is being bypassed by the TLS module, interacting
directly against the lower devs, there is no way for the bond interface
to disable its device offload capabilities, as long as the mode/policy
config allows it.
Hence, the feature flag is not directly controllable, but just reflects
the offload status based on the logic under bond_sk_check().
Here we just declare RX device offload support, and expose it via the
NETIF_F_HW_TLS_RX flag.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Boris Pismenny <borisp@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Implement TLS TX device offload for bonding interfaces.
This allows kTLS sockets running on a bond to benefit from the
device offload on capable lower devices.
To allow a simple and fast maintenance of the TLS context in SW and
lower devices, we bind the TLS socket to a specific lower dev.
To achieve a behavior similar to SW kTLS, we support only balance-xor
and 802.3ad modes, with xmit_hash_policy=layer3+4. This is enforced
in bond_sk_check(), done in a previous patch.
For the above configuration, the SW implementation keeps picking the
same exact lower dev for all the socket's SKBs. The device offload
behaves similarly, making the decision once at the connection creation.
Per socket, the TLS module should work directly with the lowest netdev
in chain, to call the tls_dev_ops operations.
As the bond interface is being bypassed by the TLS module, interacting
directly against the lower devs, there is no way for the bond interface
to disable its device offload capabilities, as long as the mode/policy
config allows it.
Hence, the feature flag is not directly controllable, but just reflects
the current offload status based on the logic under bond_sk_check().
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Boris Pismenny <borisp@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
In preparation for more cases that call netdev_update_features().
While here, move the features logic to the stage where struct bond
is already updated, and pass it as the only parameter to function
bond_set_xfrm_features().
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Boris Pismenny <borisp@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add ndo_sk_get_lower_dev() implementation for bond interfaces.
Support only for the cases where the socket's and SKBs' hash
yields identical value for the whole connection lifetime.
Here we restrict it to L3+4 sockets only, with
xmit_hash_policy==LAYER34 and bond modes xor/802.3ad.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Boris Pismenny <borisp@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Hash logic on L3 will be used in a downstream patch for one more use
case.
Take it to a function for a better code reuse.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Boris Pismenny <borisp@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
ndo_sk_get_lower_dev returns the lower netdev that corresponds to
a given socket.
Additionally, we implement a helper netdev_sk_get_lowest_dev() to get
the lowest one in chain.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Boris Pismenny <borisp@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
tcf_action_init_1() loads tc action modules automatically with
request_module() after parsing the tc action names, and it drops RTNL
lock and re-holds it before and after request_module(). This causes a
lot of troubles, as discovered by syzbot, because we can be in the
middle of batch initializations when we create an array of tc actions.
One of the problem is deadlock:
CPU 0 CPU 1
rtnl_lock();
for (...) {
tcf_action_init_1();
-> rtnl_unlock();
-> request_module();
rtnl_lock();
for (...) {
tcf_action_init_1();
-> tcf_idr_check_alloc();
// Insert one action into idr,
// but it is not committed until
// tcf_idr_insert_many(), then drop
// the RTNL lock in the _next_
// iteration
-> rtnl_unlock();
-> rtnl_lock();
-> a_o->init();
-> tcf_idr_check_alloc();
// Now waiting for the same index
// to be committed
-> request_module();
-> rtnl_lock()
// Now waiting for RTNL lock
}
rtnl_unlock();
}
rtnl_unlock();
This is not easy to solve, we can move the request_module() before
this loop and pre-load all the modules we need for this netlink
message and then do the rest initializations. So the loop breaks down
to two now:
for (i = 1; i <= TCA_ACT_MAX_PRIO && tb[i]; i++) {
struct tc_action_ops *a_o;
a_o = tc_action_load_ops(name, tb[i]...);
ops[i - 1] = a_o;
}
for (i = 1; i <= TCA_ACT_MAX_PRIO && tb[i]; i++) {
act = tcf_action_init_1(ops[i - 1]...);
}
Although this looks serious, it only has been reported by syzbot, so it
seems hard to trigger this by humans. And given the size of this patch,
I'd suggest to make it to net-next and not to backport to stable.
This patch has been tested by syzbot and tested with tdc.py by me.
Fixes: 0fedc63fad ("net_sched: commit action insertions together")
Reported-and-tested-by: syzbot+82752bc5331601cf4899@syzkaller.appspotmail.com
Reported-and-tested-by: syzbot+b3b63b6bff456bd95294@syzkaller.appspotmail.com
Reported-by: syzbot+ba67b12b1ca729912834@syzkaller.appspotmail.com
Cc: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: Cong Wang <cong.wang@bytedance.com>
Tested-by: Jamal Hadi Salim <jhs@mojatatu.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Link: https://lore.kernel.org/r/20210117005657.14810-1-xiyou.wangcong@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The TCP session does not terminate with TCP_USER_TIMEOUT when data
remain untransmitted due to zero window.
The number of unanswered zero-window probes (tcp_probes_out) is
reset to zero with incoming acks irrespective of the window size,
as described in tcp_probe_timer():
RFC 1122 4.2.2.17 requires the sender to stay open indefinitely
as long as the receiver continues to respond probes. We support
this by default and reset icsk_probes_out with incoming ACKs.
This counter, however, is the wrong one to be used in calculating the
duration that the window remains closed and data remain untransmitted.
Thanks to Jonathan Maxwell <jmaxwell37@gmail.com> for diagnosing the
actual issue.
In this patch a new timestamp is introduced for the socket in order to
track the elapsed time for the zero-window probes that have not been
answered with any non-zero window ack.
Fixes: 9721e709fa ("tcp: simplify window probe aborting on USER_TIMEOUT")
Reported-by: William McCall <william.mccall@gmail.com>
Co-developed-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Enke Chen <enchen@paloaltonetworks.com>
Reviewed-by: Yuchung Cheng <ycheng@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/20210115223058.GA39267@localhost.localdomain
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Xin Long says:
====================
net: make udp tunnel devices support fraglist
Like GRE device, UDP tunnel devices should also support fraglist, so
that some protocol (like SCTP) HW GSO that requires NETIF_F_FRAGLIST
in the dev can work. Especially when the lower device support both
NETIF_F_GSO_UDP_TUNNEL and NETIF_F_GSO_SCTP.
====================
Link: https://lore.kernel.org/r/cover.1610704037.git.lucien.xin@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Like vxlan and geneve, bareudp also needs this dev feature
to support some protocol's HW GSO.
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Some protocol HW GSO requires fraglist supported by the device, like
SCTP. Without NETIF_F_FRAGLIST set in the dev features of geneve, it
would have to do SW GSO before the packets enter the driver, even
when the geneve dev and lower dev (like veth) both have the feature
of NETIF_F_GSO_SCTP.
So this patch is to add it for geneve.
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Some protocol HW GSO requires fraglist supported by the device, like
SCTP. Without NETIF_F_FRAGLIST set in the dev features of vxlan, it
would have to do SW GSO before the packets enter the driver, even
when the vxlan dev and lower dev (like veth) both have the feature
of NETIF_F_GSO_SCTP.
So this patch is to add it for vxlan.
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The multicast route ff00::/8 is created with type RTN_UNICAST:
$ ip -6 -d route
unicast ::1 dev lo proto kernel scope global metric 256 pref medium
unicast fe80::/64 dev eth0 proto kernel scope global metric 256 pref medium
unicast ff00::/8 dev eth0 proto kernel scope global metric 256 pref medium
Set the type to RTN_MULTICAST which is more appropriate.
Fixes: e8478e80e5 ("net/ipv6: Save route type in rt6_info")
Signed-off-by: Matteo Croce <mcroce@microsoft.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The ff00::/8 multicast route is created without specifying the fc_protocol
field, so the default RTPROT_BOOT value is used:
$ ip -6 -d route
unicast ::1 dev lo proto kernel scope global metric 256 pref medium
unicast fe80::/64 dev eth0 proto kernel scope global metric 256 pref medium
unicast ff00::/8 dev eth0 proto boot scope global metric 256 pref medium
As the documentation says, this value identifies routes installed during
boot, but the route is created when interface is set up.
Change the value to RTPROT_KERNEL which is a better value.
Fixes: 1da177e4c3 ("Linux-2.6.12-rc2")
Signed-off-by: Matteo Croce <mcroce@microsoft.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
For additional robustness in the face of Hyper-V errors or malicious
behavior, validate all values that originate from packets that Hyper-V
has sent to the guest. Ensure that invalid values cannot cause indexing
off the end of an array, or subvert an existing validation via integer
overflow. Ensure that outgoing packets do not have any leftover guest
memory that has not been zeroed out.
Reported-by: Juan Vazquez <juvazq@microsoft.com>
Signed-off-by: Andrea Parri (Microsoft) <parri.andrea@gmail.com>
Link: https://lore.kernel.org/r/20210114202628.119541-1-parri.andrea@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Johannes Berg says:
====================
Various fixes:
* kernel-doc parsing fixes
* incorrect debugfs string checks
* locking fix in regulatory
* some encryption-related fixes
* tag 'mac80211-for-net-2021-01-18.2' of git://git.kernel.org/pub/scm/linux/kernel/git/jberg/mac80211:
mac80211: check if atf has been disabled in __ieee80211_schedule_txq
mac80211: do not drop tx nulldata packets on encrypted links
mac80211: fix encryption key selection for 802.3 xmit
mac80211: fix fast-rx encryption check
mac80211: fix incorrect strlen of .write in debugfs
cfg80211: fix a kerneldoc markup
cfg80211: Save the regulatory domain with a lock
cfg80211/mac80211: fix kernel-doc for SAR APIs
====================
Link: https://lore.kernel.org/r/20210118204750.7243-1-johannes@sipsolutions.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
mv88e6xxx_port_vlan_join checks whether the VTU already contains an
entry for the given vid (via mv88e6xxx_vtu_getnext), and if so, merely
changes the relevant .member[] element and loads the updated entry
into the VTU.
However, at least for the mv88e6250, the on-stack struct
mv88e6xxx_vtu_entry vlan never has its .state[] array explicitly
initialized, neither in mv88e6xxx_port_vlan_join() nor inside the
getnext implementation. So the new entry has random garbage for the
STU bits, breaking VLAN filtering.
When the VTU entry is initially created, those bits are all zero, and
we should make sure to keep them that way when the entry is updated.
Fixes: 92307069a9 (net: dsa: mv88e6xxx: Avoid VTU corruption on 6097)
Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Tobias Waldekranz <tobias@waldekranz.com>
Tested-by: Tobias Waldekranz <tobias@waldekranz.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Alex Elder says:
====================
net: ipa: interconnect improvements
The main outcome of this series is to allow the number of
interconnects used by the IPA to differ from the three that
are implemented now. With this series in place, any number
of interconnects can now be used, all specified in the
configuration data for a specific platform.
A few minor interconnect-related cleanups are implemented as well.
====================
Link: https://lore.kernel.org/r/20210115125050.20555-1-elder@linaro.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Currently we assume that the IPA hardware has exactly three
interconnects. But that won't be guaranteed for all platforms,
so allow any number of interconnects to be specified in the
configuration data.
For each platform, define an array of interconnect data entries
(still associated with the IPA clock structure), and record the
number of entries initialized in that array.
Loop over all entries in this array when initializing, enabling,
disabling, or tearing down the set of interconnects.
With this change we no longer need the ipa_interconnect_id
enumerated type, so get rid of it.
Signed-off-by: Alex Elder <elder@linaro.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Pass an the address of an IPA interconnect structure and its
configuration data to ipa_interconnect_init_one() and have that
function initialize all the structure's fields. Change the function
to simply return an error code.
Introduce ipa_interconnect_exit_one() to encapsulate the cleanup of
an IPA interconnect structure.
Signed-off-by: Alex Elder <elder@linaro.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add the name to the configuration data for each interconnect. Use
this information rather than a constant string during initialization.
Signed-off-by: Alex Elder <elder@linaro.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add fields in the ipa_interconnect structure to hold the average and
peak bandwidth values for the interconnect. Pass the configuring
data for interconnects to ipa_interconnect_init() so these values
can be recorded, and use them when enabling the interconnects.
There's no longer any need to keep a copy of the interconnect data
after initialization.
Signed-off-by: Alex Elder <elder@linaro.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Rather than having separate pointers for the memory, imem, and
config interconnect paths, maintain an array of ipa_interconnect
structures each of which contains a pointer to a path.
Signed-off-by: Alex Elder <elder@linaro.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
If disabling interconnects fails there's not a lot we can do. The
only two callers of ipa_interconnect_disable() ignore the return
value, so just give the function a void return type.
Print an error message if disabling any of the interconnects is not
successful. Return (and print) only the first error seen.
Signed-off-by: Alex Elder <elder@linaro.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Use "bandwidth" rather than "rate" in describing the average and
peak values to use for IPA interconnects. They should have been
named that way to begin with.
Signed-off-by: Alex Elder <elder@linaro.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The blamed commit was too aggressive, and it made ocelot_netdevice_event
react only to network interface events emitted for the ocelot switch
ports.
In fact, only the PRECHANGEUPPER should have had that check.
When we ignore all events that are not for us, we miss the fact that the
upper of the LAG changes, and the bonding interface gets enslaved to a
bridge. This is an operation we could offload under certain conditions.
Fixes: 7afb3e575e ("net: mscc: ocelot: don't handle netdev events for other netdevs")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
Link: https://lore.kernel.org/r/20210118135210.2666246-1-olteanv@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Pull spi fixes from Mark Brown:
"A few more bug fixes for SPI, both driver specific ones. The caching
in the Cadence driver is to avoid a deadlock trying to retrieve the
cached value later at runtime"
* tag 'spi-fix-v5.11-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi:
spi: cadence: cache reference clock rate during probe
spi: fsl: Fix driver breakage when SPI_CS_HIGH is not set in spi->mode
Pull ia64 build fix from Mike Rapoport:
"Fix an ia64 build failure caused by memory model changes"
* tag 'fixes-2021-01-18' of git://git.kernel.org/pub/scm/linux/kernel/git/rppt/memblock:
ia64: fix build failure caused by memory model changes
Pull crypto fixes from Herbert Xu:
"A Kconfig dependency issue with omap-sham and a divide by zero in xor
on some platforms"
* 'linus' of git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6:
crypto: omap-sham - Fix link error without crypto-engine
crypto: xor - Fix divide error in do_xor_speed()
Pull perf tools fixes from Arnaldo Carvalho de Melo:
- Fix 'CPU too large' error in Intel PT
- Correct event attribute sizes in 'perf inject'
- Sync build_bug.h and kvm.h kernel copies
- Fix bpf.h header include directive in 5sec.c 'perf trace' bpf example
- libbpf tests fixes
- Fix shadow stat 'perf test' for non-bash shells
- Take cgroups into account for shadow stats in 'perf stat'
* tag 'perf-tools-fixes-2021-01-17' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux:
perf inject: Correct event attribute sizes
perf intel-pt: Fix 'CPU too large' error
perf stat: Take cgroups into account for shadow stats
perf stat: Introduce struct runtime_stat_data
libperf tests: Fail when failing to get a tracepoint id
libperf tests: If a test fails return non-zero
libperf tests: Avoid uninitialized variable warning
perf test: Fix shadow stat test for non-bash shells
tools headers: Syncronize linux/build_bug.h with the kernel sources
tools headers UAPI: Sync kvm.h headers with the kernel sources
perf bpf examples: Fix bpf.h header include directive in 5sec.c example
Pull powerpc fixes from Michael Ellerman:
"One fix for a lack of alignment in our linker script, that can lead to
crashes depending on configuration etc.
One fix for the 32-bit VDSO after the C VDSO conversion.
Thanks to Andreas Schwab, Ariel Marcovitch, and Christophe Leroy"
* tag 'powerpc-5.11-4' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
powerpc/vdso: Fix clock_gettime_fallback for vdso32
powerpc: Fix alignment bug within the init sections
Pull misc vfs fixes from Al Viro:
"Several assorted fixes.
I still think that audit ->d_name race is better fixed this way for
the benefit of backports, with any possibly fancier variants done on
top of it"
* 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
dump_common_audit_data(): fix racy accesses to ->d_name
iov_iter: fix the uaccess area in copy_compat_iovec_from_user
umount(2): move the flag validity checks first
So technically there is nothing wrong with adding a pinned page to the
swap cache, but the pinning obviously means that the page can't actually
be free'd right now anyway, so it's a bit pointless.
However, the real problem is not with it being a bit pointless: the real
issue is that after we've added it to the swap cache, we'll try to unmap
the page. That will succeed, because the code in mm/rmap.c doesn't know
or care about pinned pages.
Even the unmapping isn't fatal per se, since the page will stay around
in memory due to the pinning, and we do hold the connection to it using
the swap cache. But when we then touch it next and take a page fault,
the logic in do_swap_page() will map it back into the process as a
possibly read-only page, and we'll then break the page association on
the next COW fault.
Honestly, this issue could have been fixed in any of those other places:
(a) we could refuse to unmap a pinned page (which makes conceptual
sense), or (b) we could make sure to re-map a pinned page writably in
do_swap_page(), or (c) we could just make do_wp_page() not COW the
pinned page (which was what we historically did before that "mm:
do_wp_page() simplification" commit).
But while all of them are equally valid models for breaking this chain,
not putting pinned pages into the swap cache in the first place is the
simplest one by far.
It's also the safest one: the reason why do_wp_page() was changed in the
first place was that getting the "can I re-use this page" wrong is so
fraught with errors. If you do it wrong, you end up with an incorrectly
shared page.
As a result, using "page_maybe_dma_pinned()" in either do_wp_page() or
do_swap_page() would be a serious bug since it is only a (very good)
heuristic. Re-using the page requires a hard black-and-white rule with
no room for ambiguity.
In contrast, saying "this page is very likely dma pinned, so let's not
add it to the swap cache and try to unmap it" is an obviously safe thing
to do, and if the heuristic might very rarely be a false positive, no
harm is done.
Fixes: 09854ba94c ("mm: do_wp_page() simplification")
Reported-and-tested-by: Martin Raiber <martin@urbackup.org>
Cc: Pavel Begunkov <asml.silence@gmail.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Peter Xu <peterx@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
With commit 4df4cb9e99, the Hyper-V direct-mode STIMER is actually
initialized before LAPIC is initialized: see
apic_intr_mode_init()
x86_platform.apic_post_init()
hyperv_init()
hv_stimer_alloc()
apic_bsp_setup()
setup_local_APIC()
setup_local_APIC() temporarily disables LAPIC, initializes it and
re-eanble it. The direct-mode STIMER depends on LAPIC, and when it's
registered, it can be programmed immediately and the timer can fire
very soon:
hv_stimer_init
clockevents_config_and_register
clockevents_register_device
tick_check_new_device
tick_setup_device
tick_setup_periodic(), tick_setup_oneshot()
clockevents_program_event
When the timer fires in the hypervisor, if the LAPIC is in the
disabled state, new versions of Hyper-V ignore the event and don't inject
the timer interrupt into the VM, and hence the VM hangs when it boots.
Note: when the VM starts/reboots, the LAPIC is pre-enabled by the
firmware, so the window of LAPIC being temporarily disabled is pretty
small, and the issue can only happen once out of 100~200 reboots for
a 40-vCPU VM on one dev host, and on another host the issue doesn't
reproduce after 2000 reboots.
The issue is more noticeable for kdump/kexec, because the LAPIC is
disabled by the first kernel, and stays disabled until the kdump/kexec
kernel enables it. This is especially an issue to a Generation-2 VM
(for which Hyper-V doesn't emulate the PIT timer) when CONFIG_HZ=1000
(rather than CONFIG_HZ=250) is used.
Fix the issue by moving hv_stimer_alloc() to a later place where the
LAPIC timer is initialized.
Fixes: 4df4cb9e99 ("x86/hyperv: Initialize clockevents earlier in CPU onlining")
Signed-off-by: Dexuan Cui <decui@microsoft.com>
Reviewed-by: Michael Kelley <mikelley@microsoft.com>
Link: https://lore.kernel.org/r/20210116223136.13892-1-decui@microsoft.com
Signed-off-by: Wei Liu <wei.liu@kernel.org>
The change of ia64's default memory model to SPARSEMEM causes defconfig
build to fail:
CC kernel/async.o
In file included from include/linux/numa.h:25,
from include/linux/async.h:13,
from kernel/async.c:47:
arch/ia64/include/asm/sparsemem.h:14:40: warning: "PAGE_SHIFT" is not defined, evaluates to 0 [-Wundef]
14 | #if ((CONFIG_FORCE_MAX_ZONEORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS)
| ^~~~~~~~~~
In file included from include/linux/gfp.h:6,
from include/linux/xarray.h:14,
from include/linux/radix-tree.h:19,
from include/linux/idr.h:15,
from include/linux/kernfs.h:13,
from include/linux/sysfs.h:16,
from include/linux/kobject.h:20,
from include/linux/energy_model.h:7,
from include/linux/device.h:16,
from include/linux/async.h:14,
from kernel/async.c:47:
include/linux/mmzone.h:1156:2: error: #error Allocator MAX_ORDER exceeds SECTION_SIZE
1156 | #error Allocator MAX_ORDER exceeds SECTION_SIZE
| ^~~~~
The error cause is the missing definition of PAGE_SHIFT in the calculation
of SECTION_SIZE_BITS.
Add include of <asm/page.h> to arch/ia64/include/asm/sparsemem.h to solve
the problem.
Fixes: 214496cb18 ("ia64: make SPARSEMEM default and disable DISCONTIGMEM")
Reported-by: Guenter Roeck <linux@roeck-us.net>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Mike Rapoport <rppt@linux.ibm.com>
It makes more sense to clear NETIF_F_SG instead of set it when
calling skb_segment() in sctp_gso_segment(), since SCTP GSO is
using head_skb's fraglist, of which all frags are linear skbs.
This will make SCTP GSO code more understandable.
Suggested-by: Alexander Duyck <alexander.duyck@gmail.com>
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
After commit 89319d3801 ("net: Add frag_list support to skb_segment"),
it goes to process frag_list when !hsize in skb_segment(). However, when
using skb frag_list, sg normally should not be set. In this case, hsize
will be set with len right before !hsize check, then it won't go to
frag_list processing code.
So the right thing to do is move the hsize check to the else block, so
that it won't affect the !hsize check for frag_list processing.
v1->v2:
- change to do "hsize <= 0" check instead of "!hsize", and also move
"hsize < 0" into else block, to save some cycles, as Alex suggested.
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Reviewed-by: Alexander Duyck <alexanderduyck@fb.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Commit 3226b158e6 ("net: avoid 32 x truesize under-estimation for
tiny skbs") ensured that skbs with data size lower than 1025 bytes
will be kmalloc'ed to avoid excessive page cache fragmentation and
memory consumption.
However, the fix adressed only __napi_alloc_skb() (primarily for
virtio_net and napi_get_frags()), but the issue can still be achieved
through __netdev_alloc_skb(), which is still used by several drivers.
Drivers often allocate a tiny skb for headers and place the rest of
the frame to frags (so-called copybreak).
Mirror the condition to __netdev_alloc_skb() to handle this case too.
Since v1 [0]:
- fix "Fixes:" tag;
- refine commit message (mention copybreak usecase).
[0] https://lore.kernel.org/netdev/20210114235423.232737-1-alobakin@pm.me
Fixes: a1c7fff7e1 ("net: netdev_alloc_skb() use build_skb()")
Signed-off-by: Alexander Lobakin <alobakin@pm.me>
Link: https://lore.kernel.org/r/20210115150354.85967-1-alobakin@pm.me
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
