Bluetooth: hci_sync: Fix attempting to suspend with unfiltered passive scan
When suspending the passive scanning _must_ have its filter_policy set
to 0x01 to use the accept list otherwise _any_ advertise report would
end up waking up the system.
In order to fix the filter_policy the code now checks for
hdev->suspended && HCI_CONN_FLAG_REMOTE_WAKEUP
first, since the MGMT_OP_SET_DEVICE_FLAGS will reject any attempt to
set HCI_CONN_FLAG_REMOTE_WAKEUP when it cannot be programmed in the
acceptlist, so it can return success causing the proper filter_policy
to be used.
Bluetooth: MGMT: Add conditions for setting HCI_CONN_FLAG_REMOTE_WAKEUP
HCI_CONN_FLAG_REMOTE_WAKEUP can only be set if device can be programmed
in the allowlist which in case of device using RPA requires LL Privacy
support to be enabled.
Sean Wang [Thu, 12 May 2022 21:38:11 +0000 (05:38 +0800)]
Bluetooth: btmtksdio: fix the reset takes too long
Sending WMT command during the reset in progress is invalid and would get
no response from firmware until the reset is complete, so we ignore the WMT
command here to resolve the issue which causes the whole reset process
taking too long.
Sean Wang [Thu, 12 May 2022 21:38:12 +0000 (05:38 +0800)]
Bluetooth: btmtksdio: fix possible FW initialization failure
According to FW advised sequence, mt7921s need to re-acquire privilege
immediately after the firmware download is complete before normal running.
Otherwise, it is still possible the bus may be stuck in an abnormal status
that causes FW initialization failure in the current driver.
Sean Wang [Wed, 11 May 2022 22:22:15 +0000 (06:22 +0800)]
Bluetooth: btmtksdio: fix use-after-free at btmtksdio_recv_event
We should not access skb buffer data anymore after hci_recv_frame was
called.
[ 39.634809] BUG: KASAN: use-after-free in btmtksdio_recv_event+0x1b0
[ 39.634855] Read of size 1 at addr ffffff80cf28a60d by task kworker
[ 39.634962] Call trace:
[ 39.634974] dump_backtrace+0x0/0x3b8
[ 39.634999] show_stack+0x20/0x2c
[ 39.635016] dump_stack_lvl+0x60/0x78
[ 39.635040] print_address_description+0x70/0x2f0
[ 39.635062] kasan_report+0x154/0x194
[ 39.635079] __asan_report_load1_noabort+0x44/0x50
[ 39.635099] btmtksdio_recv_event+0x1b0/0x1c4
[ 39.635129] btmtksdio_txrx_work+0x6cc/0xac4
[ 39.635157] process_one_work+0x560/0xc5c
[ 39.635177] worker_thread+0x7ec/0xcc0
[ 39.635195] kthread+0x2d0/0x3d0
[ 39.635215] ret_from_fork+0x10/0x20
[ 39.635247] Allocated by task 0:
[ 39.635260] (stack is not available)
[ 39.635281] Freed by task 2392:
[ 39.635295] kasan_save_stack+0x38/0x68
[ 39.635319] kasan_set_track+0x28/0x3c
[ 39.635338] kasan_set_free_info+0x28/0x4c
[ 39.635357] ____kasan_slab_free+0x104/0x150
[ 39.635374] __kasan_slab_free+0x18/0x28
[ 39.635391] slab_free_freelist_hook+0x114/0x248
[ 39.635410] kfree+0xf8/0x2b4
[ 39.635427] skb_free_head+0x58/0x98
[ 39.635447] skb_release_data+0x2f4/0x410
[ 39.635464] skb_release_all+0x50/0x60
[ 39.635481] kfree_skb+0xc8/0x25c
[ 39.635498] hci_event_packet+0x894/0xca4 [bluetooth]
[ 39.635721] hci_rx_work+0x1c8/0x68c [bluetooth]
[ 39.635925] process_one_work+0x560/0xc5c
[ 39.635951] worker_thread+0x7ec/0xcc0
[ 39.635970] kthread+0x2d0/0x3d0
[ 39.635990] ret_from_fork+0x10/0x20
[ 39.636021] The buggy address belongs to the object at ffffff80cf28a600
which belongs to the cache kmalloc-512 of size 512
[ 39.636039] The buggy address is located 13 bytes inside of
512-byte region [ffffff80cf28a600, ffffff80cf28a800)
Fixes: 9aebfd4a2200 ("Bluetooth: mediatek: add support for MediaTek MT7663S and MT7668S SDIO devices") Co-developed-by: Yake Yang <[email protected]> Signed-off-by: Yake Yang <[email protected]> Signed-off-by: Sean Wang <[email protected]> Signed-off-by: Marcel Holtmann <[email protected]>
Tim Harvey [Thu, 5 May 2022 19:50:10 +0000 (12:50 -0700)]
Bluetooth: btbcm: Add entry for BCM4373A0 UART Bluetooth
This patch adds the device ID for the BCM4373A0 module, found e.g. in
the Infineon (Cypress) CYW4373E chip. The required firmware file is
named 'BCM4373A0.hcd'.
Zijun Hu [Wed, 27 Apr 2022 02:16:51 +0000 (10:16 +0800)]
Bluetooth: btusb: Set HCI_QUIRK_BROKEN_ERR_DATA_REPORTING for QCA
Set HCI_QUIRK_BROKEN_ERR_DATA_REPORTING for QCA controllers since
they answer HCI_OP_READ_DEF_ERR_DATA_REPORTING with error code
"UNKNOWN HCI COMMAND" as shown below:
Bluetooth: core: Fix missing power_on work cancel on HCI close
Move power_on work cancel to hci_dev_close_sync to ensure that power_on
work is canceled after HCI interface down, power off, rfkill, etc.
For example, if
hciconfig hci0 down
is done early enough during boot, it may run before power_on work.
Then, power_on work will actually bring up interface despite above
hciconfig command.
Bluetooth: protect le accept and resolv lists with hdev->lock
Concurrent operations from events on le_{accept,resolv}_list are
currently unprotected by hdev->lock.
Most existing code do already protect the lists with that lock.
This can be observed in hci_debugfs and hci_sync.
Add the protection for these events too.
Fixes: b950aa88638c ("Bluetooth: Add definitions and track LE resolve list modification") Fixes: 0f36b589e4ee ("Bluetooth: Track LE white list modification via HCI commands") Signed-off-by: Niels Dossche <[email protected]> Signed-off-by: Marcel Holtmann <[email protected]>
Bluetooth: use hdev lock for accept_list and reject_list in conn req
All accesses (both reads and modifications) to
hdev->{accept,reject}_list are protected by hdev lock,
except the ones in hci_conn_request_evt. This can cause a race
condition in the form of a list corruption.
The solution is to protect these lists in hci_conn_request_evt as well.
I was unable to find the exact commit that introduced the issue for the
reject list, I was only able to find it for the accept list.
Fixes: a55bd29d5227 ("Bluetooth: Add white list lookup for incoming connection requests") Signed-off-by: Niels Dossche <[email protected]> Signed-off-by: Marcel Holtmann <[email protected]>
Bluetooth: use hdev lock in activate_scan for hci_is_adv_monitoring
hci_is_adv_monitoring's function documentation states that it must be
called under the hdev lock. Paths that leads to an unlocked call are:
discov_update => start_discovery => interleaved_discov => active_scan
and: discov_update => start_discovery => active_scan
The solution is to take the lock in active_scan during the duration of
the call to hci_is_adv_monitoring.
This prints warnings for controllers setting broken quirks to increase
their visibility and warn about broken controllers firmware that
probably needs updates to behave properly.
This adds HCI_QUIRK_BROKEN_ENHANCED_SETUP_SYNC_CONN quirk which can be
used to mark HCI_Enhanced_Setup_Synchronous_Connection as broken even
if its support command bit are set since some controller report it as
supported but the command don't work properly with some configurations
(e.g. BT_VOICE_TRANSPARENT/mSBC).
Steven Rostedt [Tue, 5 Apr 2022 14:02:00 +0000 (10:02 -0400)]
Bluetooth: hci_qca: Use del_timer_sync() before freeing
While looking at a crash report on a timer list being corrupted, which
usually happens when a timer is freed while still active. This is
commonly triggered by code calling del_timer() instead of
del_timer_sync() just before freeing.
One possible culprit is the hci_qca driver, which does exactly that.
Eric mentioned that wake_retrans_timer could be rearmed via the work
queue, so also move the destruction of the work queue before
del_timer_sync().
The only usage of regmap_ibt is to (after the regmap_init() macro is
expanded), pass its address to __regmap_init(), which takes a pointer to
const struct regmap_bus as input. Make it const to allow the compiler to
put it in read-only memory.
Ying Hsu [Sat, 26 Mar 2022 07:09:28 +0000 (07:09 +0000)]
Bluetooth: fix dangling sco_conn and use-after-free in sco_sock_timeout
Connecting the same socket twice consecutively in sco_sock_connect()
could lead to a race condition where two sco_conn objects are created
but only one is associated with the socket. If the socket is closed
before the SCO connection is established, the timer associated with the
dangling sco_conn object won't be canceled. As the sock object is being
freed, the use-after-free problem happens when the timer callback
function sco_sock_timeout() accesses the socket. Here's the call trace:
This patch supports the above naming schedule with inserting
[.board_name] between the firmware name and ".hcd". This scheme
is the same as used by the companion BRCM wireless chips
as can be seen in
drivers/net/wireless/broadcom/brcm80211/brcmfmac/firmware.c
or just by looking at the firmwares in linux-firmware/brcm.
Currently we only support board variants using the device
tree compatible string as board type, but other schemes are
possible.
This makes it possible to successfully load a few unique
firmware variants for some Samsung phones.
David S. Miller [Fri, 13 May 2022 10:35:46 +0000 (11:35 +0100)]
Merge branch 'lan95xx-no-polling'
Lukas Wunner says:
====================
Polling be gone on LAN95xx
Do away with link status polling on LAN95xx USB Ethernet
and rely on interrupts instead, thereby reducing bus traffic,
CPU overhead and improving interface bringup latency.
* Patch [5/7]:
* Drop call to __irq_enter_raw() which worked around a warning in
generic_handle_domain_irq(). That warning is gone since 792ea6a074ae (queued on tip.git/irq/urgent).
(Marc Zyngier, Thomas Gleixner)
====================
Lukas Wunner [Thu, 12 May 2022 08:42:07 +0000 (10:42 +0200)]
net: phy: smsc: Cope with hot-removal in interrupt handler
If reading the Interrupt Source Flag register fails with -ENODEV, then
the PHY has been hot-removed and the correct response is to bail out
instead of throwing a WARN splat and attempting to suspend the PHY.
The PHY should be stopped in due course anyway as the kernel
asynchronously tears down the device.
Lukas Wunner [Thu, 12 May 2022 08:42:05 +0000 (10:42 +0200)]
usbnet: smsc95xx: Forward PHY interrupts to PHY driver to avoid polling
Link status of SMSC LAN95xx chips is polled once per second, even though
they're capable of signaling PHY interrupts through the MAC layer.
Forward those interrupts to the PHY driver to avoid polling. Benefits
are reduced bus traffic, reduced CPU overhead and quicker interface
bringup.
Polling was introduced in 2016 by commit d69d16949346 ("usbnet:
smsc95xx: fix link detection for disabled autonegotiation").
Back then, the LAN95xx driver neglected to enable the ENERGYON interrupt,
hence couldn't detect link-up events when auto-negotiation was disabled.
The proper solution would have been to enable the ENERGYON interrupt
instead of polling.
Since then, PHY handling was moved from the LAN95xx driver to the SMSC
PHY driver with commit 05b35e7eb9a1 ("smsc95xx: add phylib support").
That PHY driver is capable of link detection with auto-negotiation
disabled because it enables the ENERGYON interrupt.
Note that signaling interrupts through the MAC layer not only works with
the integrated PHY, but also with an external PHY, provided its
interrupt pin is attached to LAN95xx's nPHY_INT pin.
In the unlikely event that the interrupt pin of an external PHY is
attached to a GPIO of the SoC (or not connected at all), the driver can
be amended to retrieve the irq from the PHY's of_node.
To forward PHY interrupts to phylib, it is not sufficient to call
phy_mac_interrupt(). Instead, the PHY's interrupt handler needs to run
so that PHY interrupts are cleared. That's because according to page
119 of the LAN950x datasheet, "The source of this interrupt is a level.
The interrupt persists until it is cleared in the PHY."
Therefore, create an IRQ domain with a single IRQ for the PHY. In the
future, the IRQ domain may be extended to support the 11 GPIOs on the
LAN95xx.
Normally the PHY interrupt should be masked until the PHY driver has
cleared it. However masking requires a (sleeping) USB transaction and
interrupts are received in (non-sleepable) softirq context. I decided
not to mask the interrupt at all (by using the dummy_irq_chip's noop
->irq_mask() callback): The USB interrupt endpoint is polled in 1 msec
intervals and normally that's sufficient to wake the PHY driver's IRQ
thread and have it clear the interrupt. If it does take longer, worst
thing that can happen is the IRQ thread is woken again. No big deal.
Because PHY interrupts are now perpetually enabled, there's no need to
selectively enable them on suspend. So remove all invocations of
smsc95xx_enable_phy_wakeup_interrupts().
In smsc95xx_resume(), move the call of phy_init_hw() before
usbnet_resume() (which restarts the status URB) to ensure that the PHY
is fully initialized when an interrupt is handled.
Lukas Wunner [Thu, 12 May 2022 08:42:04 +0000 (10:42 +0200)]
usbnet: smsc95xx: Avoid link settings race on interrupt reception
When a PHY interrupt is signaled, the SMSC LAN95xx driver updates the
MAC full duplex mode and PHY flow control registers based on cached data
in struct phy_device:
smsc95xx_status() # raises EVENT_LINK_RESET
usbnet_deferred_kevent()
smsc95xx_link_reset() # uses cached data in phydev
Simultaneously, phylib polls link status once per second and updates
that cached data:
phy_state_machine()
phy_check_link_status()
phy_read_status()
lan87xx_read_status()
genphy_read_status() # updates cached data in phydev
If smsc95xx_link_reset() wins the race against genphy_read_status(),
the registers may be updated based on stale data.
E.g. if the link was previously down, phydev->duplex is set to
DUPLEX_UNKNOWN and that's what smsc95xx_link_reset() will use, even
though genphy_read_status() may update it to DUPLEX_FULL afterwards.
PHY interrupts are currently only enabled on suspend to trigger wakeup,
so the impact of the race is limited, but we're about to enable them
perpetually.
Avoid the race by delaying execution of smsc95xx_link_reset() until
phy_state_machine() has done its job and calls back via
smsc95xx_handle_link_change().
Signaling EVENT_LINK_RESET on wakeup is not necessary because phylib
picks up link status changes through polling. So drop the declaration
of a ->link_reset() callback.
Note that the semicolon on a line by itself added in smsc95xx_status()
is a placeholder for a function call which will be added in a subsequent
commit. That function call will actually handle the INT_ENP_PHY_INT_
interrupt.
Lukas Wunner [Thu, 12 May 2022 08:42:03 +0000 (10:42 +0200)]
usbnet: smsc95xx: Don't reset PHY behind PHY driver's back
smsc95xx_reset() resets the PHY behind the PHY driver's back, which
seems like a bad idea generally. Remove that portion of the function.
We're about to use PHY interrupts instead of polling to detect link
changes on SMSC LAN95xx chips. Because smsc95xx_reset() is called from
usbnet_open(), PHY interrupt settings are lost whenever the net_device
is brought up.
There are two other callers of smsc95xx_reset(), namely smsc95xx_bind()
and smsc95xx_reset_resume(), and both may indeed benefit from a PHY
reset. However they already perform one through their calls to
phy_connect_direct() and phy_init_hw().
Upon receiving data from the Interrupt Endpoint, the SMSC LAN95xx driver
attempts to clear the signaled interrupts by writing "all ones" to the
Interrupt Status Register.
However the driver only ever enables a single type of interrupt, namely
the PHY Interrupt. And according to page 119 of the LAN950x datasheet,
its bit in the Interrupt Status Register is read-only. There's no other
way to clear it than in a separate PHY register:
The commit made binding and unbinding of USB Ethernet asymmetrical:
Before, usbnet_probe() first invoked the ->bind() callback and then
register_netdev(). usbnet_disconnect() mirrored that by first invoking
unregister_netdev() and then ->unbind().
Since the commit, the order in usbnet_disconnect() is reversed and no
longer mirrors usbnet_probe().
One consequence is that a PHY disconnected (and stopped) in ->unbind()
is afterwards stopped once more by unregister_netdev() as it closes the
netdev before unregistering. That necessitates a contortion in ->stop()
because the PHY may only be stopped if it hasn't already been
disconnected.
Reverting the commit allows making the call to phy_stop() unconditional
in ->stop().
Remove .owner field if calls are used which set it automatically.
./drivers/net/ethernet/sunplus/spl2sw_driver.c:569:3-8: No need to set
.owner here. The core will do it.
Jie Wang [Thu, 12 May 2022 06:56:31 +0000 (14:56 +0800)]
net: page_pool: add page allocation stats for two fast page allocate path
Currently If use page pool allocation stats to analysis a RX performance
degradation problem. These stats only count for pages allocate from
page_pool_alloc_pages. But nic drivers such as hns3 use
page_pool_dev_alloc_frag to allocate pages, so page stats in this API
should also be counted.
====================
net: inet: Retire port only listening_hash
This series is to retire the port only listening_hash.
The listen sk is currently stored in two hash tables,
listening_hash (hashed by port) and lhash2 (hashed by port and address).
After commit 0ee58dad5b06 ("net: tcp6: prefer listeners bound to an address")
and commit d9fbc7f6431f ("net: tcp: prefer listeners bound to an address"),
the TCP-SYN lookup fast path does not use listening_hash.
The commit 05c0b35709c5 ("tcp: seq_file: Replace listening_hash with lhash2")
also moved the seq_file (/proc/net/tcp) iteration usage from
listening_hash to lhash2.
There are still a few listening_hash usages left.
One of them is inet_reuseport_add_sock() which uses the listening_hash
to search a listen sk during the listen() system call. This turns
out to be very slow on use cases that listen on many different
VIPs at a popular port (e.g. 443). [ On top of the slowness in
adding to the tail in the IPv6 case ]. A latter patch has a
selftest to demonstrate this case.
This series takes this chance to move all remaining listening_hash
usages to lhash2 and then retire listening_hash.
====================
Martin KaFai Lau [Thu, 12 May 2022 00:06:11 +0000 (17:06 -0700)]
net: selftests: Stress reuseport listen
This patch adds a test that has 300 VIPs listening on port 443.
Each VIP:443 will have 80 listening socks by using SO_REUSEPORT.
Thus, it will have 24000 listening socks.
Before removing the port only listening_hash, all socks will be in the
same port 443 bucket and inet_reuseport_add_sock() spends much time to
walk through the bucket. After removing the port only listening_hash
and move all usage to the port+addr lhash2, each bucket in the
ideal case has 80 sk which is much smaller than before.
Here is the test result from a qemu:
Before: listen 24000 socks took 210.210485362 (~210s)
After: listen 24000 socks took 0.207173 (~210ms)
Martin KaFai Lau [Thu, 12 May 2022 00:06:05 +0000 (17:06 -0700)]
net: inet: Retire port only listening_hash
The listen sk is currently stored in two hash tables,
listening_hash (hashed by port) and lhash2 (hashed by port and address).
After commit 0ee58dad5b06 ("net: tcp6: prefer listeners bound to an address")
and commit d9fbc7f6431f ("net: tcp: prefer listeners bound to an address"),
the TCP-SYN lookup fast path does not use listening_hash.
The commit 05c0b35709c5 ("tcp: seq_file: Replace listening_hash with lhash2")
also moved the seq_file (/proc/net/tcp) iteration usage from
listening_hash to lhash2.
There are still a few listening_hash usages left.
One of them is inet_reuseport_add_sock() which uses the listening_hash
to search a listen sk during the listen() system call. This turns
out to be very slow on use cases that listen on many different
VIPs at a popular port (e.g. 443). [ On top of the slowness in
adding to the tail in the IPv6 case ]. The latter patch has a
selftest to demonstrate this case.
This patch takes this chance to move all remaining listening_hash
usages to lhash2 and then retire listening_hash.
Since most changes need to be done together, it is hard to cut
the listening_hash to lhash2 switch into small patches. The
changes in this patch is highlighted here for the review
purpose.
1. Because of the listening_hash removal, lhash2 can use the
sk->sk_nulls_node instead of the icsk->icsk_listen_portaddr_node.
This will also keep the sk_unhashed() check to work as is
after stop adding sk to listening_hash.
The union is removed from inet_listen_hashbucket because
only nulls_head is needed.
2. icsk->icsk_listen_portaddr_node and its helpers are removed.
3. The current lhash2 users needs to iterate with sk_nulls_node
instead of icsk_listen_portaddr_node.
One case is in the inet[6]_lhash2_lookup().
Another case is the seq_file iterator in tcp_ipv4.c.
One thing to note is sk_nulls_next() is needed
because the old inet_lhash2_for_each_icsk_continue()
does a "next" first before iterating.
4. Move the remaining listening_hash usage to lhash2
inet_reuseport_add_sock() which this series is
trying to improve.
inet_diag.c and mptcp_diag.c are the final two
remaining use cases and is moved to lhash2 now also.
Martin KaFai Lau [Thu, 12 May 2022 00:05:58 +0000 (17:05 -0700)]
net: inet: Open code inet_hash2 and inet_unhash2
This patch folds lhash2 related functions into __inet_hash and
inet_unhash. This will make the removal of the listening_hash
in a latter patch easier to review.
First, this patch folds inet_hash2 into __inet_hash.
For unhash, the current call sequence is like
inet_unhash() => __inet_unhash() => inet_unhash2().
The specific testing cases in __inet_unhash() are mostly related
to TCP_LISTEN sk and its caller inet_unhash() already has
the TCP_LISTEN test, so this patch folds both __inet_unhash() and
inet_unhash2() into inet_unhash().
Note that all listening_hash users also have lhash2 initialized,
so the !h->lhash2 check is no longer needed.
Martin KaFai Lau [Thu, 12 May 2022 00:05:52 +0000 (17:05 -0700)]
net: inet: Remove count from inet_listen_hashbucket
After commit 0ee58dad5b06 ("net: tcp6: prefer listeners bound to an address")
and commit d9fbc7f6431f ("net: tcp: prefer listeners bound to an address"),
the count is no longer used. This patch removes it.
====================
Make sfc-siena.ko specific to Siena
This series is a follow-up to the one titled "Move Siena into
a separate subdirectory".
It enhances the new sfc-siena.ko module to differentiate it from sfc.ko.
Patches
Patches 1-5 create separate Kconfig options for Siena, and adjusts the
various names used for work items and directories.
Patch 6 reinstates SRIOV functionality in sfc-siena.ko.
Testing
Various build tests were done such as allyesconfig, W=1 and sparse.
The new sfc-siena.ko and sfc.ko modules were tested on a machine with NICs
for both modules in them.
Inserting the updated sfc.ko and the new sfc-siena.ko modules at the same
time works, and no work items and directories exist with the same name.
====================
Martin Habets [Wed, 11 May 2022 16:20:25 +0000 (17:20 +0100)]
sfc/siena: Reinstate SRIOV init/fini function calls
They were removed in the first series since they were not used for EF10.
Put that code back for Siena, with the prototypes in siena_sriov.h
since that file is a more applicable place for it.
Martin Habets [Wed, 11 May 2022 16:20:01 +0000 (17:20 +0100)]
sfc/siena: Make MCDI logging support specific for Siena
Add a Siena Kconfig option and use it in stead of the sfc one.
Rename the internal variable for the 'mcdi_logging_default' module
parameter to avoid a naming conflict with the one in sfc.ko.
This patch set represents preparation for further work. It adds an
"index" field to struct ocelot_port, and populates it from the Felix DSA
driver and Ocelot switchdev driver.
The users of struct ocelot_port :: index are the same users as those of
struct ocelot_port_private :: chip_port.
====================
Vladimir Oltean [Wed, 11 May 2022 10:06:37 +0000 (13:06 +0300)]
net: mscc: ocelot: move ocelot_port_private :: chip_port to ocelot_port :: index
Currently the ocelot switch lib is unaware of the index of a struct
ocelot_port, since that is kept in the encapsulating structures of outer
drivers (struct dsa_port :: index, struct ocelot_port_private :: chip_port).
With the upcoming increase in complexity associated with assigning DSA
tag_8021q CPU ports to certain user ports, it becomes necessary for the
switch lib to be able to retrieve the index of a certain ocelot_port.
Therefore, introduce a new u8 to ocelot_port (same size as the chip_port
used by the ocelot switchdev driver) and rework the existing code to
populate and use it.
====================
DSA changes for multiple CPU ports (part 1)
I am trying to enable the second internal port pair from the NXP LS1028A
Felix switch for DSA-tagged traffic via "ocelot-8021q". This series
represents part 1 (of an unknown number) of that effort.
It does some preparation work, like managing host flooding in DSA via a
dedicated method, and removing the CPU port as argument from the tagging
protocol change procedure.
In terms of driver-specific changes, it reworks the 2 tag protocol
implementations in the Felix driver to have a structured data format.
It enables host flooding towards all tag_8021q CPU ports. It dynamically
updates the tag_8021q CPU port used for traps. It also fixes a bug
introduced by a previous refactoring/oversimplification commit in
net-next.
====================
Vladimir Oltean [Wed, 11 May 2022 09:50:20 +0000 (12:50 +0300)]
net: dsa: felix: reimplement tagging protocol change with function pointers
The error handling for the current tagging protocol change procedure is
a bit brittle (we dismantle the previous tagging protocol entirely
before setting up the new one). By identifying which parts of a tagging
protocol are unique to itself and which parts are shared with the other,
we can implement a protocol change procedure where error handling is a
bit more robust, because we start setting up the new protocol first, and
tear down the old one only after the setup of the specific and shared
parts succeeded.
The protocol change is a bit too open-coded too, in the area of
migrating host flood settings and MDBs. By identifying what differs
between tagging protocols (the forwarding masks for host flooding) we
can implement a more straightforward migration procedure which is
handled in the shared portion of the protocol change, rather than
individually by each protocol.
Therefore, a more structured approach calls for the introduction of a
structure of function pointers per tagging protocol. This covers setup,
teardown and the host forwarding mask. In the future it will also cover
how to prepare for a new DSA master.
The initial tagging protocol setup (at driver probe time) and the final
teardown (at driver removal time) are also adapted to call into the
structured methods of the specific protocol in current use. This is
especially relevant for teardown, where we previously called
felix_del_tag_protocol() only for the first CPU port. But by not
specifying which CPU port this is for, we gain more flexibility to
support multiple CPU ports in the future.
Vladimir Oltean [Wed, 11 May 2022 09:50:19 +0000 (12:50 +0300)]
net: dsa: felix: dynamically determine tag_8021q CPU port for traps
Ocelot switches support a single active CPU port at a time (at least as
a trapping destination, i.e. for control traffic). This is true
regardless of whether we are using the native copy-to-CPU-port-module
functionality, or a redirect action towards the software-defined
tag_8021q CPU port.
Currently we assume that the trapping destination in tag_8021q mode is
the first CPU port, yet in the future we may want to migrate the user
ports to the second CPU port.
For that to work, we need to make sure that the tag_8021q trapping
destination is a CPU port that is active, i.e. is used by at least some
user port on which the trap was added. Otherwise, we may end up
redirecting the traffic to a CPU port which isn't even up.
Note that due to the current design where we simply choose the CPU port
of the first port from the trap's ingress port mask, it may be that a
CPU port absorbes control traffic from user ports which aren't affine to
it as per user space's request. This isn't ideal, but is the lesser of
two evils. Following the user-configured affinity for traps would mean
that we can no longer reuse a single TCAM entry for multiple traps,
which is what we actually do for e.g. PTP. Either we duplicate and
deduplicate TCAM entries on the fly when user-to-CPU-port mappings
change (which is unnecessarily complicated), or we redirect trapped
traffic to all tag_8021q CPU ports if multiple such ports are in use.
The latter would have actually been nice, if it actually worked, but it
doesn't, since a OCELOT_MASK_MODE_REDIRECT action towards multiple ports
would not take PGID_SRC into consideration, and it would just duplicate
the packet towards each (CPU) port, leading to duplicates in software.
Vladimir Oltean [Wed, 11 May 2022 09:50:18 +0000 (12:50 +0300)]
net: dsa: remove port argument from ->change_tag_protocol()
DSA has not supported (and probably will not support in the future
either) independent tagging protocols per CPU port.
Different switch drivers have different requirements, some may need to
replicate some settings for each CPU port, some may need to apply some
settings on a single CPU port, while some may have to configure some
global settings and then some per-CPU-port settings.
In any case, the current model where DSA calls ->change_tag_protocol for
each CPU port turns out to be impractical for drivers where there are
global things to be done. For example, felix calls dsa_tag_8021q_register(),
which makes no sense per CPU port, so it suppresses the second call.
Let drivers deal with replication towards all CPU ports, and remove the
CPU port argument from the function prototype.
Vladimir Oltean [Wed, 11 May 2022 09:50:17 +0000 (12:50 +0300)]
net: dsa: felix: manage host flooding using a specific driver callback
At the time - commit 7569459a52c9 ("net: dsa: manage flooding on the CPU
ports") - not introducing a dedicated switch callback for host flooding
made sense, because for the only user, the felix driver, there was
nothing different to do for the CPU port than set the flood flags on the
CPU port just like on any other bridge port.
There are 2 reasons why this approach is not good enough, however.
(1) Other drivers, like sja1105, support configuring flooding as a
function of {ingress port, egress port}, whereas the DSA
->port_bridge_flags() function only operates on an egress port.
So with that driver we'd have useless host flooding from user ports
which don't need it.
(2) Even with the felix driver, support for multiple CPU ports makes it
difficult to piggyback on ->port_bridge_flags(). The way in which
the felix driver is going to support host-filtered addresses with
multiple CPU ports is that it will direct these addresses towards
both CPU ports (in a sort of multicast fashion), then restrict the
forwarding to only one of the two using the forwarding masks.
Consequently, flooding will also be enabled towards both CPU ports.
However, ->port_bridge_flags() gets passed the index of a single CPU
port, and that leaves the flood settings out of sync between the 2
CPU ports.
This is to say, it's better to have a specific driver method for host
flooding, which takes the user port as argument. This solves problem (1)
by allowing the driver to do different things for different user ports,
and problem (2) by abstracting the operation and letting the driver do
whatever, rather than explicitly making the DSA core point to the CPU
port it thinks needs to be touched.
This new method also creates a problem, which is that cross-chip setups
are not handled. However I don't have hardware right now where I can
test what is the proper thing to do, and there isn't hardware compatible
with multi-switch trees that supports host flooding. So it remains a
problem to be tackled in the future.
Vladimir Oltean [Wed, 11 May 2022 09:50:16 +0000 (12:50 +0300)]
net: dsa: introduce the dsa_cpu_ports() helper
Similar to dsa_user_ports() which retrieves a port mask of all user
ports, introduce dsa_cpu_ports() which retrieves the mask of all CPU
ports of a switch.
Vladimir Oltean [Wed, 11 May 2022 09:50:15 +0000 (12:50 +0300)]
net: dsa: felix: bring the NPI port indirection for host flooding to surface
For symmetry with host FDBs and MDBs where the indirection is now
handled outside the ocelot switch lib, do the same for bridge port
flags (unicast/multicast/broadcast flooding).
The only caller of the ocelot switch lib which uses the NPI port is the
Felix DSA driver.
Vladimir Oltean [Wed, 11 May 2022 09:50:14 +0000 (12:50 +0300)]
net: dsa: felix: bring the NPI port indirection for host MDBs to surface
For symmetry with host FDBs where the indirection is now handled outside
the ocelot switch lib, do the same for host MDB entries. The only caller
of the ocelot switch lib which uses the NPI port is the Felix DSA driver.
Vladimir Oltean [Wed, 11 May 2022 09:50:13 +0000 (12:50 +0300)]
net: dsa: felix: program host FDB entries towards PGID_CPU for tag_8021q too
I remembered why we had the host FDB migration procedure in place.
It is true that host FDB entry migration can be done by changing the
value of PGID_CPU, but the problem is that only host FDB entries learned
while operating in NPI mode go to PGID_CPU. When the CPU port operates
in tag_8021q mode, the FDB entries are learned towards the unicast PGID
equal to the physical port number of this CPU port, bypassing the
PGID_CPU indirection.
So host FDB entries learned in tag_8021q mode are not migrated any
longer towards the NPI port.
Fix this by extracting the NPI port -> PGID_CPU redirection from the
ocelot switch lib, moving it to the Felix DSA driver, and applying it
for any CPU port regardless of its kind (NPI or tag_8021q).
Fixes: a51c1c3f3218 ("net: dsa: felix: stop migrating FDBs back and forth on tag proto change") Signed-off-by: Vladimir Oltean <[email protected]> Signed-off-by: Jakub Kicinski <[email protected]>
Horatiu Vultur [Wed, 11 May 2022 20:40:59 +0000 (22:40 +0200)]
net: lan966x: Fix use of pointer after being freed
The smatch found the following warning:
drivers/net/ethernet/microchip/lan966x/lan966x_fdma.c:736 lan966x_fdma_reload()
warn: 'rx_dcbs' was already freed.
This issue can happen when changing the MTU on one of the ports and once
the RX buffers are allocated and then the TX buffer allocation fails.
In that case the RX buffers should not be restore. This fix this issue
such that the RX buffers will not be restored if the TX buffers failed
to be allocated.
Jakub Kicinski [Wed, 11 May 2022 17:23:05 +0000 (10:23 -0700)]
skbuff: replace a BUG_ON() with the new DEBUG_NET_WARN_ON_ONCE()
Very few drivers actually have Kconfig knobs for adding
-DDEBUG. 8 according to a quick grep, while there are
93 users of skb_checksum_none_assert(). Switch to the
new DEBUG_NET_WARN_ON_ONCE() to catch bad skbs.
The driver currently has three interrupt counters,
which are incremented every time each interrupt handler
executes. These driver-managed counters are not
necessary as the kernel already has logic that manages
interrupt counts and exposes them via /proc/interrupts.
This patch removes the driver-managed counters.
Build issue in drivers/net/ethernet/sfc/ptp.c 54fccfdd7c66 ("sfc: efx_default_channel_type APIs can be static") 49e6123c65da ("net: sfc: fix memory leak due to ptp channel")
https://lore.kernel.org/all/20220510130556.52598fe2@canb.auug.org.au/
- wifi: iwlwifi: iwl-dbg: use del_timer_sync() before freeing
- wifi: ath11k: reduce the wait time of 11d scan and hw scan while
adding an interface
- mac80211: fix rx reordering with non explicit / psmp ack policy
- mac80211: reset MBSSID parameters upon connection
- nl80211: fix races in nl80211_set_tx_bitrate_mask()
- tls: fix context leak on tls_device_down
- sched: act_pedit: really ensure the skb is writable
- batman-adv: don't skb_split skbuffs with frag_list
- eth: ocelot: fix various issues with TC actions (null-deref; bad
stats; ineffective drops; ineffective filter removal)"
* tag 'net-5.18-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (61 commits)
tls: Fix context leak on tls_device_down
net: sfc: ef10: fix memory leak in efx_ef10_mtd_probe()
net/smc: non blocking recvmsg() return -EAGAIN when no data and signal_pending
net: dsa: bcm_sf2: Fix Wake-on-LAN with mac_link_down()
mlxsw: Avoid warning during ip6gre device removal
net: bcmgenet: Check for Wake-on-LAN interrupt probe deferral
net: ethernet: mediatek: ppe: fix wrong size passed to memset()
Bluetooth: Fix the creation of hdev->name
i40e: i40e_main: fix a missing check on list iterator
net/sched: act_pedit: really ensure the skb is writable
s390/lcs: fix variable dereferenced before check
s390/ctcm: fix potential memory leak
s390/ctcm: fix variable dereferenced before check
net: atlantic: verify hw_head_ lies within TX buffer ring
net: atlantic: add check for MAX_SKB_FRAGS
net: atlantic: reduce scope of is_rsc_complete
net: atlantic: fix "frag[0] not initialized"
net: stmmac: fix missing pci_disable_device() on error in stmmac_pci_probe()
net: phy: micrel: Fix incorrect variable type in micrel
decnet: Use container_of() for struct dn_neigh casts
...
Linus Torvalds [Thu, 12 May 2022 17:42:56 +0000 (10:42 -0700)]
Merge branch 'for-5.18-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup
Pull cgroup fix from Tejun Heo:
"Waiman's fix for a cgroup2 cpuset bug where it could miss nodes which
were hot-added"
* 'for-5.18-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup:
cgroup/cpuset: Remove cpus_allowed/mems_allowed setup in cpuset_init_smp()
Linus Torvalds [Thu, 12 May 2022 17:21:44 +0000 (10:21 -0700)]
Merge tag 'fixes_for_v5.18-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs
Pull fs fixes from Jan Kara:
"Three fixes that I'd still like to get to 5.18:
- add a missing sanity check in the fanotify FAN_RENAME feature
(added in 5.17, let's fix it before it gets wider usage in
userspace)
- udf fix for recently introduced filesystem corruption issue
- writeback fix for a race in inode list handling that can lead to
delayed writeback and possible dirty throttling stalls"
* tag 'fixes_for_v5.18-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs:
udf: Avoid using stale lengthOfImpUse
writeback: Avoid skipping inode writeback
fanotify: do not allow setting dirent events in mask of non-dir
The commit cited below claims to fix a use-after-free condition after
tls_device_down. Apparently, the description wasn't fully accurate. The
context stayed alive, but ctx->netdev became NULL, and the offload was
torn down without a proper fallback, so a bug was present, but a
different kind of bug.
Due to misunderstanding of the issue, the original patch dropped the
refcount_dec_and_test line for the context to avoid the alleged
premature deallocation. That line has to be restored, because it matches
the refcount_inc_not_zero from the same function, otherwise the contexts
that survived tls_device_down are leaked.
This patch fixes the described issue by restoring refcount_dec_and_test.
After this change, there is no leak anymore, and the fallback to
software kTLS still works.
Taehee Yoo [Thu, 12 May 2022 05:47:09 +0000 (05:47 +0000)]
net: sfc: ef10: fix memory leak in efx_ef10_mtd_probe()
In the NIC ->probe() callback, ->mtd_probe() callback is called.
If NIC has 2 ports, ->probe() is called twice and ->mtd_probe() too.
In the ->mtd_probe(), which is efx_ef10_mtd_probe() it allocates and
initializes mtd partiion.
But mtd partition for sfc is shared data.
So that allocated mtd partition data from last called
efx_ef10_mtd_probe() will not be used.
Therefore it must be freed.
But it doesn't free a not used mtd partition data in efx_ef10_mtd_probe().
Guangguan Wang [Thu, 12 May 2022 03:08:20 +0000 (11:08 +0800)]
net/smc: non blocking recvmsg() return -EAGAIN when no data and signal_pending
Non blocking sendmsg will return -EAGAIN when any signal pending
and no send space left, while non blocking recvmsg return -EINTR
when signal pending and no data received. This may makes confused.
As TCP returns -EAGAIN in the conditions described above. Align the
behavior of smc with TCP.
Florian Fainelli [Thu, 12 May 2022 02:17:31 +0000 (19:17 -0700)]
net: dsa: bcm_sf2: Fix Wake-on-LAN with mac_link_down()
After commit 2d1f90f9ba83 ("net: dsa/bcm_sf2: fix incorrect usage of
state->link") the interface suspend path would call our mac_link_down()
call back which would forcibly set the link down, thus preventing
Wake-on-LAN packets from reaching our management port.
Fix this by looking at whether the port is enabled for Wake-on-LAN and
not clearing the link status in that case to let packets go through.
Amit Cohen [Wed, 11 May 2022 11:57:47 +0000 (14:57 +0300)]
mlxsw: Avoid warning during ip6gre device removal
IPv6 addresses which are used for tunnels are stored in a hash table
with reference counting. When a new GRE tunnel is configured, the driver
is notified and configures it in hardware.
Currently, any change in the tunnel is not applied in the driver. It
means that if the remote address is changed, the driver is not aware of
this change and the first address will be used.
This behavior results in a warning [1] in scenarios such as the
following:
# ip link add name gre1 type ip6gre local 2000::3 remote 2000::fffe tos inherit ttl inherit
# ip link set name gre1 type ip6gre local 2000::3 remote 2000::ffff ttl inherit
# ip link delete gre1
The change of the address is not applied in the driver. Currently, the
driver uses the remote address which is stored in the 'parms' of the
overlay device. When the tunnel is removed, the new IPv6 address is
used, the driver tries to release it, but as it is not aware of the
change, this address is not configured and it warns about releasing non
existing IPv6 address.
Fix it by using the IPv6 address which is cached in the IPIP entry, this
address is the last one that the driver used, so even in cases such the
above, the first address will be released, without any warning.
Paolo Abeni [Thu, 12 May 2022 11:03:12 +0000 (13:03 +0200)]
Merge branch 'nfp-vf-rate-limit-support'
Simon Horman says:
====================
*nfp: VF rate limit support
this short series adds VF rate limiting to the NFP driver.
The first patch, as suggested by Jakub Kicinski, adds a helper
to check that ndo_set_vf_rate() rate parameters are sane.
It also provides a place for further parameter checking to live,
if needed in future.
The second patch adds VF rate limit support to the NFP driver.
It addresses several comments made on v1, including removing
the parameter check that is now provided by the helper added
in the first patch.
====================
Bin Chen [Wed, 11 May 2022 11:39:32 +0000 (13:39 +0200)]
nfp: VF rate limit support
Add VF rate limit feature
This patch enhances the NFP driver to supports assignment of
both max_tx_rate and min_tx_rate to VFs
The template of configurations below is all supported.
e.g.
# ip link set $DEV vf $VF_NUM max_tx_rate $RATE_VALUE
# ip link set $DEV vf $VF_NUM min_tx_rate $RATE_VALUE
# ip link set $DEV vf $VF_NUM max_tx_rate $RATE_VALUE \
min_tx_rate $RATE_VALUE
# ip link set $DEV vf $VF_NUM min_tx_rate $RATE_VALUE \
max_tx_rate $RATE_VALUE
The max RATE_VALUE is limited to 0xFFFF which is about
63Gbps (using 1024 for 1G)
Bin Chen [Wed, 11 May 2022 11:39:31 +0000 (13:39 +0200)]
rtnetlink: verify rate parameters for calls to ndo_set_vf_rate
When calling ndo_set_vf_rate() the max_tx_rate parameter may be zero,
in which case the setting is cleared, or it must be greater or equal to
min_tx_rate.
Enforce this requirement on all calls to ndo_set_vf_rate via a wrapper
which also only calls ndo_set_vf_rate() if defined by the driver.
Vladimir Oltean [Wed, 11 May 2022 09:42:00 +0000 (12:42 +0300)]
net: enetc: kill PHY-less mode for PFs
Right now, a PHY-less port (no phy-mode, no fixed-link, no phy-handle)
doesn't register with phylink, but calls netif_carrier_on() from
enetc_start().
This makes sense for a VF, but for a PF, this is braindead, because we
never call enetc_mac_enable() so the MAC is left inoperational.
Furthermore, commit 71b77a7a27a3 ("enetc: Migrate to PHYLINK and
PCS_LYNX") put the nail in the coffin because it removed the initial
netif_carrier_off() call done right after register_netdev().
Without that call, netif_carrier_on() does not call
linkwatch_fire_event(), so the operstate remains IF_OPER_UNKNOWN.
Just deny the broken configuration by requiring that a phy-mode is
present, and always register a PF with phylink.
Kees Cook [Wed, 11 May 2022 02:53:01 +0000 (19:53 -0700)]
fortify: Provide a memcpy trap door for sharp corners
As we continue to narrow the scope of what the FORTIFY memcpy() will
accept and build alternative APIs that give the compiler appropriate
visibility into more complex memcpy scenarios, there is a need for
"unfortified" memcpy use in rare cases where combinations of compiler
behaviors, source code layout, etc, result in cases where the stricter
memcpy checks need to be bypassed until appropriate solutions can be
developed (i.e. fix compiler bugs, code refactoring, new API, etc). The
intention is for this to be used only if there's no other reasonable
solution, for its use to include a justification that can be used
to assess future solutions, and for it to be temporary.
Example usage included, based on analysis and discussion from:
https://lore.kernel.org/netdev/CANn89iLS_2cshtuXPyNUGDPaic=sJiYfvTb_wNLgWrZRyBxZ_g@mail.gmail.com
Florian Fainelli [Wed, 11 May 2022 03:17:51 +0000 (20:17 -0700)]
net: bcmgenet: Check for Wake-on-LAN interrupt probe deferral
The interrupt controller supplying the Wake-on-LAN interrupt line maybe
modular on some platforms (irq-bcm7038-l1.c) and might be probed at a
later time than the GENET driver. We need to specifically check for
-EPROBE_DEFER and propagate that error to ensure that we eventually
fetch the interrupt descriptor.
Jakub Kicinski [Thu, 12 May 2022 00:40:39 +0000 (17:40 -0700)]
Merge tag 'for-net-2022-05-11' of git://git.kernel.org/pub/scm/linux/kernel/git/bluetooth/bluetooth
Luiz Augusto von Dentz says:
====================
bluetooth pull request for net:
- Fix the creation of hdev->name when index is greater than 9999
* tag 'for-net-2022-05-11' of git://git.kernel.org/pub/scm/linux/kernel/git/bluetooth/bluetooth:
Bluetooth: Fix the creation of hdev->name
====================
Jakub Kicinski [Thu, 12 May 2022 00:33:01 +0000 (17:33 -0700)]
Merge tag 'wireless-2022-05-11' of git://git.kernel.org/pub/scm/linux/kernel/git/wireless/wireless
Kalle Valo says:
====================
wireless fixes for v5.18
Second set of fixes for v5.18 and hopefully the last one. We have a
new iwlwifi maintainer, a fix to rfkill ioctl interface and important
fixes to both stack and two drivers.
* tag 'wireless-2022-05-11' of git://git.kernel.org/pub/scm/linux/kernel/git/wireless/wireless:
rfkill: uapi: fix RFKILL_IOCTL_MAX_SIZE ioctl request definition
nl80211: fix locking in nl80211_set_tx_bitrate_mask()
mac80211_hwsim: call ieee80211_tx_prepare_skb under RCU protection
mac80211_hwsim: fix RCU protected chanctx access
mailmap: update Kalle Valo's email
mac80211: Reset MBSSID parameters upon connection
cfg80211: retrieve S1G operating channel number
nl80211: validate S1G channel width
mac80211: fix rx reordering with non explicit / psmp ack policy
ath11k: reduce the wait time of 11d scan and hw scan while add interface
MAINTAINERS: update iwlwifi driver maintainer
iwlwifi: iwl-dbg: Use del_timer_sync() before freeing
====================
Itay Iellin [Sat, 7 May 2022 12:32:48 +0000 (08:32 -0400)]
Bluetooth: Fix the creation of hdev->name
Set a size limit of 8 bytes of the written buffer to "hdev->name"
including the terminating null byte, as the size of "hdev->name" is 8
bytes. If an id value which is greater than 9999 is allocated,
then the "snprintf(hdev->name, sizeof(hdev->name), "hci%d", id)"
function call would lead to a truncation of the id value in decimal
notation.
Set an explicit maximum id parameter in the id allocation function call.
The id allocation function defines the maximum allocated id value as the
maximum id parameter value minus one. Therefore, HCI_MAX_ID is defined
as 10000.
====================
Count tc-taprio window drops in enetc driver
This series includes a patch from Po Liu (no longer with NXP) which
counts frames dropped by the tc-taprio offload in ethtool -S and in
ndo_get_stats64. It also contains a preparation patch from myself.
====================
Po Liu [Tue, 10 May 2022 16:36:15 +0000 (19:36 +0300)]
net: enetc: count the tc-taprio window drops
The enetc scheduler for IEEE 802.1Qbv has 2 options (depending on
PTGCR[TG_DROP_DISABLE]) when we attempt to send an oversized packet
which will never fit in its allotted time slot for its traffic class:
either block the entire port due to head-of-line blocking, or drop the
packet and set a bit in the writeback format of the transmit buffer
descriptor, allowing other packets to be sent.
We obviously choose the second option in the driver, but we do not
detect the drop condition, so from the perspective of the network stack,
the packet is sent and no error counter is incremented.
This change checks the writeback of the TX BD when tc-taprio is enabled,
and increments a specific ethtool statistics counter and a generic
"tx_dropped" counter in ndo_get_stats64.
Vladimir Oltean [Tue, 10 May 2022 16:36:14 +0000 (19:36 +0300)]
net: enetc: manage ENETC_F_QBV in priv->active_offloads only when enabled
Future work in this driver would like to look at priv->active_offloads &
ENETC_F_QBV to determine whether a tc-taprio qdisc offload was
installed, but this does not produce the intended effect.
All the other flags in priv->active_offloads are managed dynamically,
except ENETC_F_QBV which is set statically based on the probed SI capability.
This change makes priv->active_offloads & ENETC_F_QBV really track the
presence of a tc-taprio schedule on the port.
Some existing users, like the enetc_sched_speed_set() call from
phylink_mac_link_up(), are best kept using the old logic: the tc-taprio
offload does not re-trigger another link mode resolve, so the scheduler
needs to be functional from the get go, as long as Qbv is supported at
all on the port. So to preserve functionality there, look at the static
station interface capability from pf->si->hw_features instead.
Jakub Kicinski [Wed, 11 May 2022 23:14:15 +0000 (16:14 -0700)]
Merge branch 'macb-napi-improvements'
Robert Hancock says:
====================
MACB NAPI improvements
Simplify the logic in the Cadence MACB/GEM driver for determining
when to reschedule NAPI processing, and update it to use NAPI for the
TX path as well as the RX path.
Changes since v1: Changed to use separate TX and RX NAPI instances and
poll functions to avoid unnecessary checks of the other ring (TX/RX)
states during polling and to use budget handling for both RX and TX.
Fixed locking to protect against concurrent access to TX ring on
TX transmit and TX poll paths.
====================
Robert Hancock [Mon, 9 May 2022 19:46:35 +0000 (13:46 -0600)]
net: macb: use NAPI for TX completion path
This driver was using the TX IRQ handler to perform all TX completion
tasks. Under heavy TX network load, this can cause significant irqs-off
latencies (found to be in the hundreds of microseconds using ftrace).
This can cause other issues, such as overrunning serial UART FIFOs when
using high baud rates with limited UART FIFO sizes.
Switch to using a NAPI poll handler to perform the TX completion work
to get this out of hard IRQ context and avoid the IRQ latency impact. A
separate NAPI instance is used for TX and RX to avoid checking the other
ring's state unnecessarily when doing the poll, and so that the NAPI
budget handling can work for both TX and RX packets.
A new per-queue tx_ptr_lock spinlock has been added to avoid using the
main device lock (with IRQs needing to be disabled) across the entire TX
mapping operation, and also to protect the TX queue pointers from
concurrent access between the TX start and TX poll operations.
The TX Used Bit Read interrupt (TXUBR) handling also needs to be moved into
the TX NAPI poll handler to maintain the proper order of operations. A flag
is used to notify the poll handler that a UBR condition needs to be
handled. The macb_tx_restart handler has had some locking added for global
register access, since this could now potentially happen concurrently on
different queues.
Robert Hancock [Mon, 9 May 2022 19:46:34 +0000 (13:46 -0600)]
net: macb: simplify/cleanup NAPI reschedule checking
Previously the macb_poll method was checking the RSR register after
completing its RX receive work to see if additional packets had been
received since IRQs were disabled, since this controller does not
maintain the pending IRQ status across IRQ disable. It also had to
double-check the register after re-enabling IRQs to detect if packets
were received after the first check but before IRQs were enabled.
Using the RSR register for this purpose is problematic since it reflects
the global device state rather than the per-queue state, so if packets
are being received on multiple queues it may end up retriggering receive
on a queue where the packets did not actually arrive and not on the one
where they did arrive. This will also cause problems with an upcoming
change to use NAPI for the TX path where use of multiple queues is more
likely.
Add a macb_rx_pending function to check the RX ring to see if more
packets have arrived in the queue, and use that to check if NAPI should
be rescheduled rather than the RSR register. By doing this, we can just
ignore the global RSR register entirely, and thus save some extra device
register accesses at the same time.
This also makes the previous first check for pending packets rather
redundant, since it would be checking the RX ring state which was just
checked in the receive work function. Therefore we can get rid of it and
just check after enabling interrupts whether packets are already
pending.
Xiaomeng Tong [Tue, 10 May 2022 20:48:46 +0000 (13:48 -0700)]
i40e: i40e_main: fix a missing check on list iterator
The bug is here:
ret = i40e_add_macvlan_filter(hw, ch->seid, vdev->dev_addr, &aq_err);
The list iterator 'ch' will point to a bogus position containing
HEAD if the list is empty or no element is found. This case must
be checked before any use of the iterator, otherwise it will
lead to a invalid memory access.
To fix this bug, use a new variable 'iter' as the list iterator,
while use the origin variable 'ch' as a dedicated pointer to
point to the found element.
Vladimir Oltean [Tue, 10 May 2022 22:09:04 +0000 (01:09 +0300)]
selftests: forwarding: tc_actions: allow mirred egress test to run on non-offloaded h2
The host interfaces $h1 and $h2 don't have to be switchdev interfaces,
but due to the fact that we pass $tcflags which may have the value of
"skip_sw", we force $h2 to offload a drop rule for dst_ip, something
which it may not be able to do.
The selftest only wants to verify the hit count of this rule as a means
of figuring out whether the packet was received, so remove the $tcflags
for it and let it be done in software.
Jakub Kicinski [Wed, 11 May 2022 22:11:32 +0000 (15:11 -0700)]
Merge branch '1GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next-queue
Tony Nguyen says:
====================
1GbE Intel Wired LAN Driver Updates 2022-05-10
This series contains updates to igc driver only.
Sasha cleans up the code by removing an unused function and removing an
enum for PHY type as there is only one PHY. The return type for
igc_check_downshift() is changed to void as it always returns success.
* '1GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next-queue:
igc: Change type of the 'igc_check_downshift' method
igc: Remove unused phy_type enum
igc: Remove igc_set_spd_dplx method
====================
Paolo Abeni [Tue, 10 May 2022 14:57:34 +0000 (16:57 +0200)]
net/sched: act_pedit: really ensure the skb is writable
Currently pedit tries to ensure that the accessed skb offset
is writable via skb_unclone(). The action potentially allows
touching any skb bytes, so it may end-up modifying shared data.
The above causes some sporadic MPTCP self-test failures, due to
this code:
The above modifies a data byte outside the skb head and the skb is
a cloned one, carrying a TCP output packet.
This change addresses the issue by keeping track of a rough
over-estimate highest skb offset accessed by the action and ensuring
such offset is really writable.
Note that this may cause performance regressions in some scenarios,
but hopefully pedit is not in the critical path.
Jakub Kicinski [Mon, 9 May 2022 15:05:32 +0000 (08:05 -0700)]
eth: amd: remove NI6510 support (ni65)
Looks like all the changes to this driver had been tree-wide
refactoring since git era begun. The driver is using virt_to_bus()
we should make it use more modern DMA APIs but since it's unlikely
to be getting any use these days delete it instead. We can always
revert to bring it back.
Jakub Kicinski [Mon, 9 May 2022 15:01:30 +0000 (08:01 -0700)]
net: appletalk: remove Apple/Farallon LocalTalk PC support
Looks like all the changes to this driver had been tree-wide
refactoring since git era begun. The driver is using virt_to_bus()
we should make it use more modern DMA APIs but since it's unlikely
to be getting any use these days delete it instead. We can always
revert to bring it back.
projects / linux.git / log