KVM: x86: Wait for IPIs to be delivered when handling Hyper-V TLB flush hypercall
Prior to commit 0baedd792713 ("KVM: x86: make Hyper-V PV TLB flush use
tlb_flush_guest()"), kvm_hv_flush_tlb() was using 'KVM_REQ_TLB_FLUSH |
KVM_REQUEST_NO_WAKEUP' when making a request to flush TLBs on other vCPUs
and KVM_REQ_TLB_FLUSH is/was defined as:
(0 | KVM_REQUEST_WAIT | KVM_REQUEST_NO_WAKEUP)
so KVM_REQUEST_WAIT was lost. Hyper-V TLFS, however, requires that
"This call guarantees that by the time control returns back to the
caller, the observable effects of all flushes on the specified virtual
processors have occurred." and without KVM_REQUEST_WAIT there's a small
chance that the vCPU making the TLB flush will resume running before
all IPIs get delivered to other vCPUs and a stale mapping can get read
there.
Fix the issue by adding KVM_REQUEST_WAIT flag to KVM_REQ_TLB_FLUSH_GUEST:
kvm_hv_flush_tlb() is the sole caller which uses it for
kvm_make_all_cpus_request()/kvm_make_vcpus_request_mask() where
KVM_REQUEST_WAIT makes a difference.
Linus Torvalds [Thu, 9 Dec 2021 21:20:59 +0000 (13:20 -0800)]
Merge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma
Pull rdma fixes from Jason Gunthorpe:
"Quite a few small bug fixes old and new, also Doug Ledford is retiring
now, we thank him for his work. Details:
- Use after free in rxe
- mlx5 DM regression
- hns bugs triggred by device reset
- Two fixes for CONFIG_DEBUG_PREEMPT
- Several longstanding corner case bugs in hfi1
- Two irdma data path bugs in rare cases and some memory issues"
* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma:
RDMA/irdma: Don't arm the CQ more than two times if no CE for this CQ
RDMA/irdma: Report correct WC errors
RDMA/irdma: Fix a potential memory allocation issue in 'irdma_prm_add_pble_mem()'
RDMA/irdma: Fix a user-after-free in add_pble_prm
IB/hfi1: Fix leak of rcvhdrtail_dummy_kvaddr
IB/hfi1: Fix early init panic
IB/hfi1: Insure use of smp_processor_id() is preempt disabled
IB/hfi1: Correct guard on eager buffer deallocation
RDMA/rtrs: Call {get,put}_cpu_ptr to silence a debug kernel warning
RDMA/hns: Do not destroy QP resources in the hw resetting phase
RDMA/hns: Do not halt commands during reset until later
Remove Doug Ledford from MAINTAINERS
RDMA/mlx5: Fix releasing unallocated memory in dereg MR flow
RDMA: Fix use-after-free in rxe_queue_cleanup
Linus Torvalds [Thu, 9 Dec 2021 19:26:44 +0000 (11:26 -0800)]
Merge tag 'net-5.16-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
Pull networking fixes from Jakub Kicinski:
"Including fixes from bpf, can and netfilter.
Current release - regressions:
- bpf, sockmap: re-evaluate proto ops when psock is removed from
sockmap
Current release - new code bugs:
- bpf: fix bpf_check_mod_kfunc_call for built-in modules
- ice: fixes for TC classifier offloads
- vrf: don't run conntrack on vrf with !dflt qdisc
Previous releases - regressions:
- bpf: fix the off-by-two error in range markings
- seg6: fix the iif in the IPv6 socket control block
- devlink: fix netns refcount leak in devlink_nl_cmd_reload()
- dsa: mv88e6xxx: fix "don't use PHY_DETECT on internal PHY's"
- dsa: mv88e6xxx: allow use of PHYs on CPU and DSA ports
Previous releases - always broken:
- ethtool: do not perform operations on net devices being
unregistered
- udp: use datalen to cap max gso segments
- ice: fix races in stats collection
- fec: only clear interrupt of handling queue in fec_enet_rx_queue()
- m_can: pci: fix incorrect reference clock rate
- m_can: disable and ignore ELO interrupt
- mvpp2: fix XDP rx queues registering
Misc:
- treewide: add missing includes masked by cgroup -> bpf.h
dependency"
* tag 'net-5.16-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (82 commits)
net: dsa: mv88e6xxx: allow use of PHYs on CPU and DSA ports
net: wwan: iosm: fixes unable to send AT command during mbim tx
net: wwan: iosm: fixes net interface nonfunctional after fw flash
net: wwan: iosm: fixes unnecessary doorbell send
net: dsa: felix: Fix memory leak in felix_setup_mmio_filtering
MAINTAINERS: s390/net: remove myself as maintainer
net/sched: fq_pie: prevent dismantle issue
net: mana: Fix memory leak in mana_hwc_create_wq
seg6: fix the iif in the IPv6 socket control block
nfp: Fix memory leak in nfp_cpp_area_cache_add()
nfc: fix potential NULL pointer deref in nfc_genl_dump_ses_done
nfc: fix segfault in nfc_genl_dump_devices_done
udp: using datalen to cap max gso segments
net: dsa: mv88e6xxx: error handling for serdes_power functions
can: kvaser_usb: get CAN clock frequency from device
can: kvaser_pciefd: kvaser_pciefd_rx_error_frame(): increase correct stats->{rx,tx}_errors counter
net: mvpp2: fix XDP rx queues registering
vmxnet3: fix minimum vectors alloc issue
net, neigh: clear whole pneigh_entry at alloc time
net: dsa: mv88e6xxx: fix "don't use PHY_DETECT on internal PHY's"
...
Linus Torvalds [Thu, 9 Dec 2021 19:08:19 +0000 (11:08 -0800)]
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/hid/hid
Pull HID fixes from Jiri Kosina:
- fixes for various drivers which assume that a HID device is on USB
transport, but that might not necessarily be the case, as the device
can be faked by uhid. (Greg, Benjamin Tissoires)
- fix for spurious wakeups on certain Lenovo notebooks (Thomas
Weißschuh)
- a few other device-specific quirks
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/hid/hid:
HID: Ignore battery for Elan touchscreen on Asus UX550VE
HID: intel-ish-hid: ipc: only enable IRQ wakeup when requested
HID: google: add eel USB id
HID: add USB_HID dependancy to hid-prodikeys
HID: add USB_HID dependancy to hid-chicony
HID: bigbenff: prevent null pointer dereference
HID: sony: fix error path in probe
HID: add USB_HID dependancy on some USB HID drivers
HID: check for valid USB device for many HID drivers
HID: wacom: fix problems when device is not a valid USB device
HID: add hid_is_usb() function to make it simpler for USB detection
HID: quirks: Add quirk for the Microsoft Surface 3 type-cover
On sc7180-trogdor class devices with 'fw_devlink=permissive' and KASAN
enabled, you'll see a Use-After-Free reported at bootup.
The root of the problem is that dwc3_qcom_of_register_core() is adding
a devm-allocated "tx-fifo-resize" property to its device tree node
using of_add_property().
The issue is that of_add_property() makes a _permanent_ addition to
the device tree that lasts until reboot. That means allocating memory
for the property using "devm" managed memory is a terrible idea since
that memory will be freed upon probe deferral or device unbinding.
Let's revert the patch since the system is still functional without
it. The fact that of_add_property() makes a permanent change is extra
fodder for those folks who were aruging that the device tree isn't
really the right way to pass information between parts of the
driver. It is an exercise left to the reader to submit a patch
re-adding the new feature in a way that makes everyone happier.
Eric Biggers [Thu, 9 Dec 2021 01:04:55 +0000 (17:04 -0800)]
aio: fix use-after-free due to missing POLLFREE handling
signalfd_poll() and binder_poll() are special in that they use a
waitqueue whose lifetime is the current task, rather than the struct
file as is normally the case. This is okay for blocking polls, since a
blocking poll occurs within one task; however, non-blocking polls
require another solution. This solution is for the queue to be cleared
before it is freed, by sending a POLLFREE notification to all waiters.
Unfortunately, only eventpoll handles POLLFREE. A second type of
non-blocking poll, aio poll, was added in kernel v4.18, and it doesn't
handle POLLFREE. This allows a use-after-free to occur if a signalfd or
binder fd is polled with aio poll, and the waitqueue gets freed.
Fix this by making aio poll handle POLLFREE.
A patch by Ramji Jiyani <[email protected]>
(https://lore.kernel.org/r/20211027011834.2497484[email protected])
tried to do this by making aio_poll_wake() always complete the request
inline if POLLFREE is seen. However, that solution had two bugs.
First, it introduced a deadlock, as it unconditionally locked the aio
context while holding the waitqueue lock, which inverts the normal
locking order. Second, it didn't consider that POLLFREE notifications
are missed while the request has been temporarily de-queued.
The second problem was solved by my previous patch. This patch then
properly fixes the use-after-free by handling POLLFREE in a
deadlock-free way. It does this by taking advantage of the fact that
freeing of the waitqueue is RCU-delayed, similar to what eventpoll does.
Eric Biggers [Thu, 9 Dec 2021 01:04:54 +0000 (17:04 -0800)]
aio: keep poll requests on waitqueue until completed
Currently, aio_poll_wake() will always remove the poll request from the
waitqueue. Then, if aio_poll_complete_work() sees that none of the
polled events are ready and the request isn't cancelled, it re-adds the
request to the waitqueue. (This can easily happen when polling a file
that doesn't pass an event mask when waking up its waitqueue.)
This is fundamentally broken for two reasons:
1. If a wakeup occurs between vfs_poll() and the request being
re-added to the waitqueue, it will be missed because the request
wasn't on the waitqueue at the time. Therefore, IOCB_CMD_POLL
might never complete even if the polled file is ready.
2. When the request isn't on the waitqueue, there is no way to be
notified that the waitqueue is being freed (which happens when its
lifetime is shorter than the struct file's). This is supposed to
happen via the waitqueue entries being woken up with POLLFREE.
Therefore, leave the requests on the waitqueue until they are actually
completed (or cancelled). To keep track of when aio_poll_complete_work
needs to be scheduled, use new fields in struct poll_iocb. Remove the
'done' field which is now redundant.
Note that this is consistent with how sys_poll() and eventpoll work;
their wakeup functions do *not* remove the waitqueue entries.
Eric Biggers [Thu, 9 Dec 2021 01:04:53 +0000 (17:04 -0800)]
signalfd: use wake_up_pollfree()
wake_up_poll() uses nr_exclusive=1, so it's not guaranteed to wake up
all exclusive waiters. Yet, POLLFREE *must* wake up all waiters. epoll
and aio poll are fortunately not affected by this, but it's very
fragile. Thus, the new function wake_up_pollfree() has been introduced.
Eric Biggers [Thu, 9 Dec 2021 01:04:52 +0000 (17:04 -0800)]
binder: use wake_up_pollfree()
wake_up_poll() uses nr_exclusive=1, so it's not guaranteed to wake up
all exclusive waiters. Yet, POLLFREE *must* wake up all waiters. epoll
and aio poll are fortunately not affected by this, but it's very
fragile. Thus, the new function wake_up_pollfree() has been introduced.
Eric Biggers [Thu, 9 Dec 2021 01:04:51 +0000 (17:04 -0800)]
wait: add wake_up_pollfree()
Several ->poll() implementations are special in that they use a
waitqueue whose lifetime is the current task, rather than the struct
file as is normally the case. This is okay for blocking polls, since a
blocking poll occurs within one task; however, non-blocking polls
require another solution. This solution is for the queue to be cleared
before it is freed, using 'wake_up_poll(wq, EPOLLHUP | POLLFREE);'.
However, that has a bug: wake_up_poll() calls __wake_up() with
nr_exclusive=1. Therefore, if there are multiple "exclusive" waiters,
and the wakeup function for the first one returns a positive value, only
that one will be called. That's *not* what's needed for POLLFREE;
POLLFREE is special in that it really needs to wake up everyone.
Considering the three non-blocking poll systems:
- io_uring poll doesn't handle POLLFREE at all, so it is broken anyway.
- aio poll is unaffected, since it doesn't support exclusive waits.
However, that's fragile, as someone could add this feature later.
- epoll doesn't appear to be broken by this, since its wakeup function
returns 0 when it sees POLLFREE. But this is fragile.
Although there is a workaround (see epoll), it's better to define a
function which always sends POLLFREE to all waiters. Add such a
function. Also make it verify that the queue really becomes empty after
all waiters have been woken up.
Linus Torvalds [Thu, 9 Dec 2021 18:49:36 +0000 (10:49 -0800)]
Merge tag 'netfs-fixes-20211207' of git://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-fs
Pull netfslib fixes from David Howells:
- Fix a lockdep warning and potential deadlock. This is takes the
simple approach of offloading the write-to-cache done from within a
network filesystem read to a worker thread to avoid taking the
sb_writer lock from the cache backing filesystem whilst holding the
mmap lock on an inode from the network filesystem.
Jan Kara posits a scenario whereby this can cause deadlock[1], though
it's quite complex and I think requires someone in userspace to
actually do I/O on the cache files. Matthew Wilcox isn't so certain,
though[2].
An alternative way to fix this, suggested by Darrick Wong, might be
to allow cachefiles to prevent userspace from performing I/O upon the
file - something like an exclusive open - but that's beyond the scope
of a fix here if we do want to make such a facility in the future.
- In some of the error handling paths where netfs_ops->cleanup() is
called, the arguments are transposed[3]. gcc doesn't complain because
one of the parameters is void* and one of the values is void*.
Jiri Olsa [Mon, 6 Dec 2021 18:20:32 +0000 (19:20 +0100)]
ftrace/samples: Add module to test multi direct modify interface
Adding ftrace-direct-multi-modify.ko kernel module that uses
modify_ftrace_direct_multi API. The core functionality is taken
from ftrace-direct-modify.ko kernel module and changed to fit
multi direct interface.
The init function creates kthread that periodically calls
modify_ftrace_direct_multi to change the trampoline address
for the direct ftrace_ops. The ftrace trace_pipe then shows
trace from both trampolines.
Loic Poulain [Thu, 9 Dec 2021 13:16:33 +0000 (18:46 +0530)]
bus: mhi: core: Add support for forced PM resume
For whatever reason, some devices like QCA6390, WCN6855 using ath11k
are not in M3 state during PM resume, but still functional. The
mhi_pm_resume should then not fail in those cases, and let the higher
level device specific stack continue resuming process.
Add an API mhi_pm_resume_force(), to force resuming irrespective of the
current MHI state. This fixes a regression with non functional ath11k WiFi
after suspend/resume cycle on some machines.
INTERCEPT_x are bit positions, but the code was using the raw value of
INTERCEPT_VINTR (4) instead of BIT(INTERCEPT_VINTR).
This resulted in masking of bit 2 - that is, SMI instead of VINTR.
Sasha Levin [Thu, 9 Dec 2021 16:51:13 +0000 (11:51 -0500)]
tools/lib/lockdep: drop leftover liblockdep headers
Clean up remaining headers that are specific to liblockdep but lived in
the shared header directory. These are all unused after the liblockdep
code was removed in commit 7246f4dcaccc ("tools/lib/lockdep: drop
liblockdep").
Note that there are still headers that were originally created for
liblockdep, that still have liblockdep references, but they are used by
other tools/ code at this point.
net: dsa: mv88e6xxx: allow use of PHYs on CPU and DSA ports
Martyn Welch reports that his CPU port is unable to link where it has
been necessary to use one of the switch ports with an internal PHY for
the CPU port. The reason behind this is the port control register is
left forcing the link down, preventing traffic flow.
This occurs because during initialisation, phylink expects the link to
be down, and DSA forces the link down by synthesising a call to the
DSA drivers phylink_mac_link_down() method, but we don't touch the
forced-link state when we later reconfigure the port.
Resolve this by also unforcing the link state when we are operating in
PHY mode and the PPU is set to poll the PHY to retrieve link status
information.
Jakub Kicinski [Thu, 9 Dec 2021 16:10:38 +0000 (08:10 -0800)]
Merge branch 'net-wwan-iosm-bug-fixes'
M Chetan Kumar says:
====================
net: wwan: iosm: bug fixes
This patch series brings in IOSM driver bug fixes. Patch details are
explained below.
PATCH1: stop sending unnecessary doorbell in IP tx flow.
PATCH2: Restore the IP channel configuration after fw flash.
PATCH3: Removed the unnecessary check around control port TX transfer.
====================
M Chetan Kumar [Thu, 9 Dec 2021 10:16:28 +0000 (15:46 +0530)]
net: wwan: iosm: fixes net interface nonfunctional after fw flash
Devlink initialization flow was overwriting the IP traffic
channel configuration. This was causing wwan0 network interface
to be unusable after fw flash.
When device boots to fully functional mode restore the IP channel
configuration.
M Chetan Kumar [Thu, 9 Dec 2021 10:16:27 +0000 (15:46 +0530)]
net: wwan: iosm: fixes unnecessary doorbell send
In TX packet accumulation flow transport layer is
giving a doorbell to device even though there is
no pending control TX transfer that needs immediate
attention.
Introduced a new hpda_ctrl_pending variable to keep
track of pending control TX transfer. If there is a
pending control TX transfer which needs an immediate
attention only then give a doorbell to device.
Andrea Mayer [Wed, 8 Dec 2021 19:54:09 +0000 (20:54 +0100)]
seg6: fix the iif in the IPv6 socket control block
When an IPv4 packet is received, the ip_rcv_core(...) sets the receiving
interface index into the IPv4 socket control block (v5.16-rc4,
net/ipv4/ip_input.c line 510):
IPCB(skb)->iif = skb->skb_iif;
If that IPv4 packet is meant to be encapsulated in an outer IPv6+SRH
header, the seg6_do_srh_encap(...) performs the required encapsulation.
In this case, the seg6_do_srh_encap function clears the IPv6 socket control
block (v5.16-rc4 net/ipv6/seg6_iptunnel.c line 163):
memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
The memset(...) was introduced in commit ef489749aae5 ("ipv6: sr: clear
IP6CB(skb) on SRH ip4ip6 encapsulation") a long time ago (2019-01-29).
Since the IPv6 socket control block and the IPv4 socket control block share
the same memory area (skb->cb), the receiving interface index info is lost
(IP6CB(skb)->iif is set to zero).
As a side effect, that condition triggers a NULL pointer dereference if
commit 0857d6f8c759 ("ipv6: When forwarding count rx stats on the orig
netdev") is applied.
To fix that issue, we set the IP6CB(skb)->iif with the index of the
receiving interface once again.
Jianglei Nie [Thu, 9 Dec 2021 06:15:11 +0000 (14:15 +0800)]
nfp: Fix memory leak in nfp_cpp_area_cache_add()
In line 800 (#1), nfp_cpp_area_alloc() allocates and initializes a
CPP area structure. But in line 807 (#2), when the cache is allocated
failed, this CPP area structure is not freed, which will result in
memory leak.
We can fix it by freeing the CPP area when the cache is allocated
failed (#2).
nfc: fix potential NULL pointer deref in nfc_genl_dump_ses_done
The done() netlink callback nfc_genl_dump_ses_done() should check if
received argument is non-NULL, because its allocation could fail earlier
in dumpit() (nfc_genl_dump_ses()).
Jakub Kicinski [Thu, 9 Dec 2021 15:43:22 +0000 (07:43 -0800)]
Merge tag 'linux-can-fixes-for-5.16-20211209' of git://git.kernel.org/pub/scm/linux/kernel/git/mkl/linux-can
Marc Kleine-Budde says:
====================
can 2021-12-09
Both patches are by Jimmy Assarsson. The first one fixes the
incrementing of the rx/tx error counters in the Kvaser PCIe FD driver.
The second one fixes the Kvaser USB driver by using the CAN clock
frequency provided by the device instead of using a hard coded value.
* tag 'linux-can-fixes-for-5.16-20211209' of git://git.kernel.org/pub/scm/linux/kernel/git/mkl/linux-can:
can: kvaser_usb: get CAN clock frequency from device
can: kvaser_pciefd: kvaser_pciefd_rx_error_frame(): increase correct stats->{rx,tx}_errors counter
====================
drm/i915/gen11: Moving WAs to icl_gt_workarounds_init()
Bspec page says "Reset: BUS", Accordingly moving w/a's:
Wa_1407352427,Wa_1406680159 to proper function icl_gt_workarounds_init()
Which will resolve guc enabling error
v2:
- Previous patch rev2 was created by email client which caused the
Build failure, This v2 is to resolve the previous broken series
The driver currently assumes that the notify callback is only received
when the device is done with all the queued buffers.
However, this is not true, since the notify callback could be called
without any of the queued buffers being completed (for example, with
virtio-pci and shared interrupts) or with only some of the buffers being
completed (since the driver makes them available to the device in
multiple separate virtqueue_add_sgs() calls).
This can lead to incorrect data on the I2C bus or memory corruption in
the guest if the device operates on buffers which are have been freed by
the driver. (The WARN_ON in the driver is also triggered.)
BUG kmalloc-128 (Tainted: G W ): Poison overwritten
First byte 0x0 instead of 0x6b
Allocated in i2cdev_ioctl_rdwr+0x9d/0x1de age=243 cpu=0 pid=28
memdup_user+0x2e/0xbd
i2cdev_ioctl_rdwr+0x9d/0x1de
i2cdev_ioctl+0x247/0x2ed
vfs_ioctl+0x21/0x30
sys_ioctl+0xb18/0xb41
Freed in i2cdev_ioctl_rdwr+0x1bb/0x1de age=68 cpu=0 pid=28
kfree+0x1bd/0x1cc
i2cdev_ioctl_rdwr+0x1bb/0x1de
i2cdev_ioctl+0x247/0x2ed
vfs_ioctl+0x21/0x30
sys_ioctl+0xb18/0xb41
Fix this by calling virtio_get_buf() from the notify handler like other
virtio drivers and by actually waiting for all the buffers to be
completed.
Jimmy Assarsson [Wed, 8 Dec 2021 15:21:22 +0000 (16:21 +0100)]
can: kvaser_usb: get CAN clock frequency from device
The CAN clock frequency is used when calculating the CAN bittiming
parameters. When wrong clock frequency is used, the device may end up
with wrong bittiming parameters, depending on user requested bittiming
parameters.
To avoid this, get the CAN clock frequency from the device. Various
existing Kvaser Leaf products use different CAN clocks.
drm/amd/display: Fix DPIA outbox timeout after S3/S4/reset
[Why]
The HW interrupt gets disabled after S3/S4/reset so we don't receive
notifications for HPD or AUX from DMUB - leading to timeout and
black screen with (or without) DPIA links connected.
[How]
Re-enable the interrupt after S3/S4/reset like we do for the other
DC interrupts.
Guard both instances of the outbox interrupt enable or we'll hang
during restore on ASIC that don't support it.
Fixes: 6eff272dbee7ad ("drm/amd/display: Fix DPIA outbox timeout after GPU reset") Reviewed-by: Jude Shih <[email protected]> Acked-by: Pavle Kotarac <[email protected]> Signed-off-by: Nicholas Kazlauskas <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
Louis Amas [Tue, 7 Dec 2021 14:34:22 +0000 (15:34 +0100)]
net: mvpp2: fix XDP rx queues registering
The registration of XDP queue information is incorrect because the
RX queue id we use is invalid. When port->id == 0 it appears to works
as expected yet it's no longer the case when port->id != 0.
The problem arised while using a recent kernel version on the
MACCHIATOBin. This board has several ports:
* eth0 and eth1 are 10Gbps interfaces ; both ports has port->id == 0;
* eth2 is a 1Gbps interface with port->id != 0.
Code from xdp-tutorial (more specifically advanced03-AF_XDP) was used
to test packet capture and injection on all these interfaces. The XDP
kernel was simplified to:
SEC("xdp_sock")
int xdp_sock_prog(struct xdp_md *ctx)
{
int index = ctx->rx_queue_index;
/* A set entry here means that the correspnding queue_id
* has an active AF_XDP socket bound to it. */
if (bpf_map_lookup_elem(&xsks_map, &index))
return bpf_redirect_map(&xsks_map, index, 0);
return XDP_PASS;
}
Starting the program using:
./af_xdp_user -d DEV
Gives the following result:
* eth0 : ok
* eth1 : ok
* eth2 : no capture, no injection
Investigating the issue shows that XDP rx queues for eth2 are wrong:
XDP expects their id to be in the range [0..3] but we found them to be
in the range [32..35].
Trying to force rx queue ids using:
./af_xdp_user -d eth2 -Q 32
fails as expected (we shall not have more than 4 queues).
When we register the XDP rx queue information (using
xdp_rxq_info_reg() in function mvpp2_rxq_init()) we tell it to use
rxq->id as the queue id. This value is computed as:
rxq->id = port->id * max_rxq_count + queue_id
where max_rxq_count depends on the device version. In the MACCHIATOBin
case, this value is 32, meaning that rx queues on eth2 are numbered
from 32 to 35 - there are four of them.
Clearly, this is not the per-port queue id that XDP is expecting:
it wants a value in the range [0..3]. It shall directly use queue_id
which is stored in rxq->logic_rxq -- so let's use that value instead.
rxq->id is left untouched ; its value is indeed valid but it should
not be used in this context.
This is consistent with the remaining part of the code in
mvpp2_rxq_init().
With this change, packet capture is working as expected on all the
MACCHIATOBin ports.
Hannes Reinecke [Wed, 8 Dec 2021 06:58:53 +0000 (07:58 +0100)]
libata: add horkage for ASMedia 1092
The ASMedia 1092 has a configuration mode which will present a
dummy device; sadly the implementation falsely claims to provide
a device with 100M which doesn't actually exist.
So disable this device to avoid errors during boot.
Ronak Doshi [Tue, 7 Dec 2021 08:17:37 +0000 (00:17 -0800)]
vmxnet3: fix minimum vectors alloc issue
'Commit 39f9895a00f4 ("vmxnet3: add support for 32 Tx/Rx queues")'
added support for 32Tx/Rx queues. Within that patch, value of
VMXNET3_LINUX_MIN_MSIX_VECT was updated.
However, there is a case (numvcpus = 2) which actually requires 3
intrs which matches VMXNET3_LINUX_MIN_MSIX_VECT which then is
treated as failure by stack to allocate more vectors. This patch
fixes this issue.
1) Fix bogus compilter warning in nfnetlink_queue, from Florian Westphal.
2) Don't run conntrack on vrf with !dflt qdisc, from Nicolas Dichtel.
3) Fix nft_pipapo bucket load in AVX2 lookup routine for six 8-bit
groups, from Stefano Brivio.
4) Break rule evaluation on malformed TCP options.
5) Use socat instead of nc in selftests/netfilter/nft_zones_many.sh,
also from Florian
6) Fix KCSAN data-race in conntrack timeout updates, from Eric Dumazet.
* git://git.kernel.org/pub/scm/linux/kernel/git/pablo/nf:
netfilter: conntrack: annotate data-races around ct->timeout
selftests: netfilter: switch zone stress to socat
netfilter: nft_exthdr: break evaluation if setting TCP option fails
selftests: netfilter: Add correctness test for mac,net set type
nft_set_pipapo: Fix bucket load in AVX2 lookup routine for six 8-bit groups
vrf: don't run conntrack on vrf with !dflt qdisc
netfilter: nfnetlink_queue: silence bogus compiler warning
====================
We've added 12 non-merge commits during the last 22 day(s) which contain
a total of 29 files changed, 659 insertions(+), 80 deletions(-).
The main changes are:
1) Fix an off-by-two error in packet range markings and also add a batch of
new tests for coverage of these corner cases, from Maxim Mikityanskiy.
2) Fix a compilation issue on MIPS JIT for R10000 CPUs, from Johan Almbladh.
3) Fix two functional regressions and a build warning related to BTF kfunc
for modules, from Kumar Kartikeya Dwivedi.
4) Fix outdated code and docs regarding BPF's migrate_disable() use on non-
PREEMPT_RT kernels, from Sebastian Andrzej Siewior.
5) Add missing includes in order to be able to detangle cgroup vs bpf header
dependencies, from Jakub Kicinski.
6) Fix regression in BPF sockmap tests caused by missing detachment of progs
from sockets when they are removed from the map, from John Fastabend.
7) Fix a missing "no previous prototype" warning in x86 JIT caused by BPF
dispatcher, from Björn Töpel.
* https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf:
bpf: Add selftests to cover packet access corner cases
bpf: Fix the off-by-two error in range markings
treewide: Add missing includes masked by cgroup -> bpf dependency
tools/resolve_btfids: Skip unresolved symbol warning for empty BTF sets
bpf: Fix bpf_check_mod_kfunc_call for built-in modules
bpf: Make CONFIG_DEBUG_INFO_BTF depend upon CONFIG_BPF_SYSCALL
mips, bpf: Fix reference to non-existing Kconfig symbol
bpf: Make sure bpf_disable_instrumentation() is safe vs preemption.
Documentation/locking/locktypes: Update migrate_disable() bits.
bpf, sockmap: Re-evaluate proto ops when psock is removed from sockmap
bpf, sockmap: Attach map progs to psock early for feature probes
bpf, x86: Fix "no previous prototype" warning
====================
Paulo Alcantara [Wed, 8 Dec 2021 01:51:04 +0000 (22:51 -0300)]
cifs: fix ntlmssp auth when there is no key exchange
Warn on the lack of key exchange during NTLMSSP authentication rather
than aborting it as there are some servers that do not set it in
CHALLENGE message.
net: dsa: mv88e6xxx: fix "don't use PHY_DETECT on internal PHY's"
This commit fixes a misunderstanding in commit 4a3e0aeddf09 ("net: dsa:
mv88e6xxx: don't use PHY_DETECT on internal PHY's").
For Marvell DSA switches with the PHY_DETECT bit (for non-6250 family
devices), controls whether the PPU polls the PHY to retrieve the link,
speed, duplex and pause status to update the port configuration. This
applies for both internal and external PHYs.
For some switches such as 88E6352 and 88E6390X, PHY_DETECT has an
additional function of enabling auto-media mode between the internal
PHY and SERDES blocks depending on which first gains link.
The original intention of commit 5d5b231da7ac (net: dsa: mv88e6xxx: use
PHY_DETECT in mac_link_up/mac_link_down) was to allow this bit to be
used to detect when this propagation is enabled, and allow software to
update the port configuration. This has found to be necessary for some
switches which do not automatically propagate status from the SERDES to
the port, which includes the 88E6390. However, commit 4a3e0aeddf09
("net: dsa: mv88e6xxx: don't use PHY_DETECT on internal PHY's") breaks
this assumption.
Maarten Zanders has confirmed that the issue he was addressing was for
an 88E6250 switch, which does not have a PHY_DETECT bit in bit 12, but
instead a link status bit. Therefore, mv88e6xxx_port_ppu_updates() does
not report correctly.
This patch resolves the above issues by reverting Maarten's change and
instead making mv88e6xxx_port_ppu_updates() indicate whether the port
is internal for the 88E6250 family of switches.
Yes, you're right, I'm targeting the 6250 family. And yes, your
suggestion would solve my case and is a better implementation for
the other devices (as far as I can see).
This added tool dependencies on various build systems using %.dtb
targets. Validation will need to be controlled by a kconfig or make
variable instead, but for now let's just revert it.
Peter Zijlstra [Sat, 4 Dec 2021 09:14:02 +0000 (10:14 +0100)]
sched,x86: Don't use cluster topology for x86 hybrid CPUs
For x86 hybrid CPUs like Alder Lake, the order of CPU selection should
be based strictly on CPU priority. Don't include cluster topology for
hybrid CPUs to avoid interference with such CPU selection order.
On Alder Lake, the Atom CPU cluster has more capacity (4 Atom CPUs) vs
Big core cluster (2 hyperthread CPUs). This could potentially bias CPU
selection towards Atom over Big Core, when Big core CPU has higher
priority.
Jesse Brandeburg [Sat, 13 Nov 2021 01:06:02 +0000 (17:06 -0800)]
ice: safer stats processing
The driver was zeroing live stats that could be fetched by
ndo_get_stats64 at any time. This could result in inconsistent
statistics, and the telltale sign was when reading stats frequently from
/proc/net/dev, the stats would go backwards.
Fix by collecting stats into a local, and delaying when we write to the
structure so it's not incremental.
Dan Carpenter [Thu, 2 Dec 2021 04:48:31 +0000 (20:48 -0800)]
drm/i915: Fix error pointer dereference in i915_gem_do_execbuffer()
Originally "out_fence" was set using out_fence = sync_file_create() but
which returns NULL, but now it is set with out_fence = eb_requests_create()
which returns error pointers. The error path needs to be modified to
avoid an Oops in the "goto err_request;" path.
Hans de Goede [Tue, 7 Dec 2021 12:10:53 +0000 (13:10 +0100)]
HID: Ignore battery for Elan touchscreen on Asus UX550VE
Battery status is reported for the Asus UX550VE touchscreen even though
it does not have a battery. Prevent it from always reporting the
battery as low.
Jiri Olsa [Mon, 6 Dec 2021 18:20:30 +0000 (19:20 +0100)]
ftrace: Use direct_ops hash in unregister_ftrace_direct
Now when we have *direct_multi interface the direct_functions
hash is no longer owned just by direct_ops. It's also used by
any other ftrace_ops passed to *direct_multi interface.
Thus to find out that we are unregistering the last function
from direct_ops, we need to check directly direct_ops's hash.
drm/syncobj: Deal with signalled fences in drm_syncobj_find_fence.
dma_fence_chain_find_seqno only ever returns the top fence in the
chain or an unsignalled fence. Hence if we request a seqno that
is already signalled it returns a NULL fence. Some callers are
not prepared to handle this, like the syncobj transfer functions
for example.
This behavior is "new" with timeline syncobj and it looks like
not all callers were updated. To fix this behavior make sure
that a successful drm_sync_find_fence always returns a non-NULL
fence.
v2: Move the fix to drm_syncobj_find_fence from the transfer
functions.
Sagi Grimberg [Wed, 8 Dec 2021 13:35:06 +0000 (15:35 +0200)]
nvmet-tcp: fix possible list corruption for unexpected command failure
nvmet_tcp_handle_req_failure needs to understand weather to prepare
for incoming data or the next pdu. However if we misidentify this, we
will wait for 0-length data, and queue the response although nvmet_req_init
already did that.
The particular command was namespace management command with no data,
which was incorrectly categorized as a command with incapsule data.
Also, add a code comment of what we are trying to do here.
Qu Wenruo [Wed, 1 Dec 2021 11:56:17 +0000 (19:56 +0800)]
btrfs: replace the BUG_ON in btrfs_del_root_ref with proper error handling
I hit the BUG_ON() with generic/475 test case, and to my surprise, all
callers of btrfs_del_root_ref() are already aborting transaction, thus
there is not need for such BUG_ON(), just go to @out label and caller
will properly handle the error.
btrfs: zoned: clear data relocation bg on zone finish
When finishing a zone that is used by a dedicated data relocation
block group, also remove its reference from fs_info, so we're not trying
to use a full block group for allocations during data relocation, which
will always fail.
The result is we're not making any forward progress and end up in a
deadlock situation.
Fixes: c2707a255623 ("btrfs: zoned: add a dedicated data relocation block group") Reviewed-by: Naohiro Aota <[email protected]> Signed-off-by: Johannes Thumshirn <[email protected]> Signed-off-by: David Sterba <[email protected]>
In case brtfs_qgroup_reserve_data() or btrfs_delalloc_reserve_metadata()
fail the allocated extent_changeset will not be freed.
So in btrfs_check_data_free_space() and btrfs_delalloc_reserve_space()
free the allocated extent_changeset to get rid of the allocated memory.
The issue currently only happens in the direct IO write path, but only
after 65b3c08606e5 ("btrfs: fix ENOSPC failure when attempting direct IO
write into NOCOW range"), and also at defrag_one_locked_target(). Every
other place is always calling extent_changeset_free() even if its call
to btrfs_delalloc_reserve_space() or btrfs_check_data_free_space() has
failed.
Naohiro Aota [Tue, 30 Nov 2021 03:40:21 +0000 (12:40 +0900)]
btrfs: fix re-dirty process of tree-log nodes
There is a report of a transaction abort of -EAGAIN with the following
script.
#!/bin/sh
for d in sda sdb; do
mkfs.btrfs -d single -m single -f /dev/\${d}
done
mount /dev/sda /mnt/test
mount /dev/sdb /mnt/scratch
for dir in test scratch; do
echo 3 >/proc/sys/vm/drop_caches
fio --directory=/mnt/\${dir} --name=fio.\${dir} --rw=read --size=50G --bs=64m \
--numjobs=$(nproc) --time_based --ramp_time=5 --runtime=480 \
--group_reporting |& tee /dev/shm/fio.\${dir}
echo 3 >/proc/sys/vm/drop_caches
done
The abort occurs because of a write hole while writing out freeing tree
nodes of a tree-log tree. For zoned btrfs, we re-dirty a freed tree
node to ensure btrfs can write the region and does not leave a hole on
write on a zoned device. The current code fails to re-dirty a node
when the tree-log tree's depth is greater or equal to 2. That leads to
a transaction abort with -EAGAIN.
Fix the issue by properly re-dirtying a node on walking up the tree.
Josef Bacik [Wed, 24 Nov 2021 19:14:25 +0000 (14:14 -0500)]
btrfs: call mapping_set_error() on btree inode with a write error
generic/484 fails sometimes with compression on because the write ends
up small enough that it goes into the btree. This means that we never
call mapping_set_error() on the inode itself, because the page gets
marked as fine when we inline it into the metadata. When the metadata
writeback happens we see it and abort the transaction properly and mark
the fs as readonly, however we don't do the mapping_set_error() on
anything. In syncfs() we will simply return 0 if the sb is marked
read-only, so we can't check for this in our syncfs callback. The only
way the error gets returned if we called mapping_set_error() on
something. Fix this by calling mapping_set_error() on the btree inode
mapping. This allows us to properly return an error on syncfs and pass
generic/484 with compression on.
This happens because we're trying to read from a extent buffer page that
is !PageUptodate. This happens because we will clear the page uptodate
when we have an IO error, but we don't clear the extent buffer uptodate.
If we do a read later and find this extent buffer we'll think its valid
and not return an error, and then trip over this warning.
Fix this by also clearing uptodate on the extent buffer when this
happens, so that we get an error when we do a btrfs_search_slot() and
find this block later.
bpf: Add selftests to cover packet access corner cases
This commit adds BPF verifier selftests that cover all corner cases by
packet boundary checks. Specifically, 8-byte packet reads are tested at
the beginning of data and at the beginning of data_meta, using all kinds
of boundary checks (all comparison operators: <, >, <=, >=; both
permutations of operands: data + length compared to end, end compared to
data + length). For each case there are three tests:
1. Length is just enough for an 8-byte read. Length is either 7 or 8,
depending on the comparison.
2. Length is increased by 1 - should still pass the verifier. These
cases are useful, because they failed before commit 2fa7d94afc1a
("bpf: Fix the off-by-two error in range markings").
3. Length is decreased by 1 - should be rejected by the verifier.
Some existing tests are just renamed to avoid duplication.
Josef Bacik [Mon, 22 Nov 2021 22:04:19 +0000 (17:04 -0500)]
btrfs: fail if fstrim_range->start == U64_MAX
We've always been failing generic/260 because it's testing things we
actually don't care about and thus won't fail for. However we probably
should fail for fstrim_range->start == U64_MAX since we clearly can't
trim anything past that. This in combination with an update to
generic/260 will allow us to pass this test properly.
Dan Carpenter [Tue, 16 Nov 2021 11:50:25 +0000 (14:50 +0300)]
btrfs: fix error pointer dereference in btrfs_ioctl_rm_dev_v2()
If memdup_user() fails the error handing will crash when it tries
to kfree() an error pointer. Just return directly because there is
no cleanup required.
Fixes: 1a15eb724aae ("btrfs: use btrfs_get_dev_args_from_path in dev removal ioctls") Reviewed-by: Josef Bacik <[email protected]> Signed-off-by: Dan Carpenter <[email protected]> Reviewed-by: David Sterba <[email protected]> Signed-off-by: David Sterba <[email protected]>
tracefs: Set all files to the same group ownership as the mount option
As people have been asking to allow non-root processes to have access to
the tracefs directory, it was considered best to only allow groups to have
access to the directory, where it is easier to just set the tracefs file
system to a specific group (as other would be too dangerous), and that way
the admins could pick which processes would have access to tracefs.
Unfortunately, this broke tooling on Android that expected the other bit
to be set. For some special cases, for non-root tools to trace the system,
tracefs would be mounted and change the permissions of the top level
directory which gave access to all running tasks permission to the
tracing directory. Even though this would be dangerous to do in a
production environment, for testing environments this can be useful.
Now with the new changes to not allow other (which is still the proper
thing to do), it breaks the testing tooling. Now more code needs to be
loaded on the system to change ownership of the tracing directory.
The real solution is to have tracefs honor the gid=xxx option when
mounting. That is,
(tracing group tracing has value 1003)
mount -t tracefs -o gid=1003 tracefs /sys/kernel/tracing
should have it that all files in the tracing directory should be of the
given group.
Copy the logic from d_walk() from dcache.c and simplify it for the mount
case of tracefs if gid is set. All the files in tracefs will be walked and
their group will be set to the value passed in.
tracefs: Have new files inherit the ownership of their parent
If directories in tracefs have their ownership changed, then any new files
and directories that are created under those directories should inherit
the ownership of the director they are created in.
Merge tag 'iio-fixes-for-5.16b' of https://git.kernel.org/pub/scm/linux/kernel/git/jic23/iio into char-misc-next
Jonathan writes:
2nd set of IIO fixes for 5.16
Note 1st set were before the merge window.
Biggest set in here fix what happens when things go wrong in
the interrupt handlers for an IIO trigger.
Otherwise normal mix of recent and ancient bugs.
trigger core
- Fix reference counting bug that was preventing the iio_trig
structures from being released.
adxrs290
- Correctly sign extend the rate and temperature data.
at91-sama5d2
- Fix sign extension from the wrong bit and use the scan_type
values to avoid it being open coded in two places (which were
out of sync)
axp20x_adc
- Fix current reporting bit depth.
dln2-adc
- Fix a lock ordering issue and lockdep complaint that results.
- Add error handling for failure to register the trigger.
imx8qxp
- Wrong config dependency
kxcjk-1013
- Potential leak due to wrong guard on cleanup.
ltr501, kxsd9, stk3310, itg3200, ad7768
- Don't return error codes from interrupt handler and call
iio_trigger_notify_done() on all paths to avoid leaving
trigger disabled on an intermittent fault.
mma8452
- Fix missing iio_trigger_get() that could lead to use after free.
stm32
- Fix a current leak.
- Avoid null pointer derefence on defer_probe error due to wrong
struct device being passed.
stm32-timer
- Drop space in MODULE_ALIAS.
* tag 'iio-fixes-for-5.16b' of https://git.kernel.org/pub/scm/linux/kernel/git/jic23/iio:
iio: trigger: stm32-timer: fix MODULE_ALIAS
iio: adc: stm32: fix null pointer on defer_probe error
iio: at91-sama5d2: Fix incorrect sign extension
iio: adc: axp20x_adc: fix charging current reporting on AXP22x
iio: gyro: adxrs290: fix data signedness
iio: ad7768-1: Call iio_trigger_notify_done() on error
iio: itg3200: Call iio_trigger_notify_done() on error
iio: imx8qxp-adc: fix dependency to the intended ARCH_MXC config
iio: dln2: Check return value of devm_iio_trigger_register()
iio: trigger: Fix reference counting
iio: dln2-adc: Fix lockdep complaint
iio: adc: stm32: fix a current leak by resetting pcsel before disabling vdda
iio: mma8452: Fix trigger reference couting
iio: stk3310: Don't return error code in interrupt handler
iio: kxsd9: Don't return error code in trigger handler
iio: ltr501: Don't return error code in trigger handler
iio: accel: kxcjk-1013: Fix possible memory leak in probe and remove
Wudi Wang [Wed, 8 Dec 2021 01:54:29 +0000 (09:54 +0800)]
irqchip/irq-gic-v3-its.c: Force synchronisation when issuing INVALL
INVALL CMD specifies that the ITS must ensure any caching associated with
the interrupt collection defined by ICID is consistent with the LPI
configuration tables held in memory for all Redistributors. SYNC is
required to ensure that INVALL is executed.
Currently, LPI configuration data may be inconsistent with that in the
memory within a short period of time after the INVALL command is executed.
Vitaly Kuznetsov [Mon, 29 Nov 2021 09:47:01 +0000 (10:47 +0100)]
KVM: nVMX: Don't use Enlightened MSR Bitmap for L3
When KVM runs as a nested hypervisor on top of Hyper-V it uses Enlightened
VMCS and enables Enlightened MSR Bitmap feature for its L1s and L2s (which
are actually L2s and L3s from Hyper-V's perspective). When MSR bitmap is
updated, KVM has to reset HV_VMX_ENLIGHTENED_CLEAN_FIELD_MSR_BITMAP from
clean fields to make Hyper-V aware of the change. For KVM's L1s, this is
done in vmx_disable_intercept_for_msr()/vmx_enable_intercept_for_msr().
MSR bitmap for L2 is build in nested_vmx_prepare_msr_bitmap() by blending
MSR bitmap for L1 and L1's idea of MSR bitmap for L2. KVM, however, doesn't
check if the resulting bitmap is different and never cleans
HV_VMX_ENLIGHTENED_CLEAN_FIELD_MSR_BITMAP in eVMCS02. This is incorrect and
may result in Hyper-V missing the update.
The issue could've been solved by calling evmcs_touch_msr_bitmap() for
eVMCS02 from nested_vmx_prepare_msr_bitmap() unconditionally but doing so
would not give any performance benefits (compared to not using Enlightened
MSR Bitmap at all). 3-level nesting is also not a very common setup
nowadays.
Don't enable 'Enlightened MSR Bitmap' feature for KVM's L2s (real L3s) for
now.
Joakim Zhang [Mon, 6 Dec 2021 13:54:57 +0000 (21:54 +0800)]
net: fec: only clear interrupt of handling queue in fec_enet_rx_queue()
Background:
We have a customer is running a Profinet stack on the 8MM which receives and
responds PNIO packets every 4ms and PNIO-CM packets every 40ms. However, from
time to time the received PNIO-CM package is "stock" and is only handled when
receiving a new PNIO-CM or DCERPC-Ping packet (tcpdump shows the PNIO-CM and
the DCERPC-Ping packet at the same time but the PNIO-CM HW timestamp is from
the expected 40 ms and not the 2s delay of the DCERPC-Ping).
After debugging, we noticed PNIO, PNIO-CM and DCERPC-Ping packets would
be handled by different RX queues.
The root cause should be driver ack all queues' interrupt when handle a
specific queue in fec_enet_rx_queue(). The blamed patch is introduced to
receive as much packets as possible once to avoid interrupt flooding.
But it's unreasonable to clear other queues'interrupt when handling one
queue, this patch tries to fix it.
The exit function is wrongly placed in the __init section and this leads
to a crash when the module is unloaded. Just remove both the init and
exit functions since this module does not need them.
Reported by Kernel Concurrency Sanitizer on:
CPU: 1 PID: 17446 Comm: kworker/u4:5 Tainted: G W 5.16.0-rc4-syzkaller #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
Workqueue: krdsd rds_send_worker
Note: I chose an arbitrary commit for the Fixes: tag,
because I do not think we need to backport this fix to very old kernels.
Fixes: e37542ba111f ("netfilter: conntrack: avoid possible false sharing") Signed-off-by: Eric Dumazet <[email protected]> Reported-by: syzbot <[email protected]> Signed-off-by: Pablo Neira Ayuso <[email protected]>
Stefano Brivio [Sat, 27 Nov 2021 10:33:38 +0000 (11:33 +0100)]
selftests: netfilter: Add correctness test for mac,net set type
The existing net,mac test didn't cover the issue recently reported
by Nikita Yushchenko, where MAC addresses wouldn't match if given
as first field of a concatenated set with AVX2 and 8-bit groups,
because there's a different code path covering the lookup of six
8-bit groups (MAC addresses) if that's the first field.
Add a similar mac,net test, with MAC address and IPv4 address
swapped in the set specification.
Stefano Brivio [Sat, 27 Nov 2021 10:33:37 +0000 (11:33 +0100)]
nft_set_pipapo: Fix bucket load in AVX2 lookup routine for six 8-bit groups
The sixth byte of packet data has to be looked up in the sixth group,
not in the seventh one, even if we load the bucket data into ymm6
(and not ymm5, for convenience of tracking stalls).
Without this fix, matching on a MAC address as first field of a set,
if 8-bit groups are selected (due to a small set size) would fail,
that is, the given MAC address would never match.
Nicolas Dichtel [Fri, 26 Nov 2021 14:36:12 +0000 (15:36 +0100)]
vrf: don't run conntrack on vrf with !dflt qdisc
After the below patch, the conntrack attached to skb is set to "notrack" in
the context of vrf device, for locally generated packets.
But this is true only when the default qdisc is set to the vrf device. When
changing the qdisc, notrack is not set anymore.
In fact, there is a shortcut in the vrf driver, when the default qdisc is
set, see commit dcdd43c41e60 ("net: vrf: performance improvements for
IPv4") for more details.
This patch ensures that the behavior is always the same, whatever the qdisc
is.
To demonstrate the difference, a new test is added in conntrack_vrf.sh.
Fixes: 8c9c296adfae ("vrf: run conntrack only in context of lower/physdev for locally generated packets") Signed-off-by: Nicolas Dichtel <[email protected]> Acked-by: Florian Westphal <[email protected]> Reviewed-by: David Ahern <[email protected]> Signed-off-by: Pablo Neira Ayuso <[email protected]>
Linus Torvalds [Tue, 7 Dec 2021 23:36:45 +0000 (15:36 -0800)]
Merge tag 'perf-tools-fixes-for-v5.16-2021-12-07' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux
Pull perf tools fixes from Arnaldo Carvalho de Melo:
- Fix SMT detection fast read path on sysfs.
- Fix memory leaks when processing feature headers in perf.data files.
- Fix 'Simple expression parser' 'perf test' on arch without CPU die
topology info, such as s/390.
- Fix building perf with BUILD_BPF_SKEL=1.
- Fix 'perf bench' by reverting "perf bench: Fix two memory leaks
detected with ASan".
- Fix itrace space allowed for new attributes in 'perf script'.
- Fix the build feature detection fast path, that was always failing on
systems with python3 development packages, speeding up the build.
- Reset shadow counts before loading, fixing metrics using
duration_time.
- Sync more kernel headers changed by the new futex_waitv syscall: s390
and powerpc.
* tag 'perf-tools-fixes-for-v5.16-2021-12-07' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux:
perf bpf_skel: Do not use typedef to avoid error on old clang
perf bpf: Fix building perf with BUILD_BPF_SKEL=1 by default in more distros
perf header: Fix memory leaks when processing feature headers
perf test: Reset shadow counts before loading
perf test: Fix 'Simple expression parser' test on arch without CPU die topology info
tools build: Remove needless libpython-version feature check that breaks test-all fast path
perf tools: Fix SMT detection fast read path
tools headers UAPI: Sync powerpc syscall table file changed by new futex_waitv syscall
perf inject: Fix itrace space allowed for new attributes
tools headers UAPI: Sync s390 syscall table file changed by new futex_waitv syscall
Revert "perf bench: Fix two memory leaks detected with ASan"
__blkdev_direct_IO_async() returns errors from bio_iov_iter_get_pages()
directly, in which case upper layers won't be expecting ->ki_complete
to be called by the block layer and will terminate the request. However,
there is also bio_endio() leading to a second ->ki_complete and a double
free.
projects / linux.git / log