Stas Sergeev [Thu, 5 Dec 2024 07:36:14 +0000 (10:36 +0300)]
tun: fix group permission check
Currently tun checks the group permission even if the user have matched.
Besides going against the usual permission semantic, this has a
very interesting implication: if the tun group is not among the
supplementary groups of the tun user, then effectively no one can
access the tun device. CAP_SYS_ADMIN still can, but its the same as
not setting the tun ownership.
This patch relaxes the group checking so that either the user match
or the group match is enough. This avoids the situation when no one
can access the device even though the ownership is properly set.
Also I simplified the logic by removing the redundant inversions:
tun_not_capable() --> !tun_capable()
Signed-off-by: Stas Sergeev <stsp2@yandex.ru> Reviewed-by: Willem de Bruijn <willemb@google.com> Acked-by: Jason Wang <jasowang@redhat.com> Link: https://patch.msgid.link/20241205073614.294773-1-stsp2@yandex.ru Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Johannes Berg [Fri, 6 Dec 2024 10:30:57 +0000 (11:30 +0100)]
tools: ynl-gen-c: don't require -o argument
Without -o the tool currently crashes, but it's not marked
as required. The only thing we can't do without it is to
generate the correct #include for user source files, but
we can put a placeholder instead.
====================
net: Convert some UDP tunnel drivers to NETDEV_PCPU_STAT_DSTATS.
VXLAN, Geneve and Bareudp use various device counters for managing
RX and TX statistics:
* VXLAN uses the device core_stats for RX and TX drops, tstats for
regular RX/TX counters and DEV_STATS_INC() for various types of
RX/TX errors.
* Geneve uses tstats for regular RX/TX counters and DEV_STATS_INC()
for everything else, include RX/TX drops.
* Bareudp, was recently converted to follow VXLAN behaviour, that is,
device core_stats for RX and TX drops, tstats for regular RX/TX
counters and DEV_STATS_INC() for other counter types.
Let's consolidate statistics management around the dstats counters
instead. This avoids using core_stats in VXLAN and Bareudp, as
core_stats is supposed to be used by core networking code only (and not
in drivers). This also allows Geneve to avoid using atomic increments
when updating RX and TX drop counters, as dstats is per-cpu. Finally,
this also simplifies the code as all three modules now handle stats in
the same way and with only two different sets of counters (the per-cpu
dstats and the atomic DEV_STATS_INC()).
Patch 1 creates dstats helper functions that can be used outside of VRF
(until then, dstats was VRF-specific).
Then patches 2 to 4, convert VXLAN, Geneve and Bareudp, one by one.
====================
Guillaume Nault [Wed, 4 Dec 2024 12:11:32 +0000 (13:11 +0100)]
bareudp: Handle stats using NETDEV_PCPU_STAT_DSTATS.
Bareudp uses the TSTATS infrastructure (dev_sw_netstats_*()) for RX
packet counters. It was also recently converted to use the device core
stats (dev_core_stats_*()) for RX and TX drops (see commit 788d5d655bc9
("bareudp: Use pcpu stats to update rx_dropped counter.")).
Since core stats are to be avoided in drivers, and for consistency with
VXLAN and Geneve, let's convert packet stats handling to DSTATS, which
can handle RX/TX stats and packet drops. Statistics that don't fit
DSTATS are still updated atomically with DEV_STATS_INC().
Guillaume Nault [Wed, 4 Dec 2024 12:11:30 +0000 (13:11 +0100)]
geneve: Handle stats using NETDEV_PCPU_STAT_DSTATS.
Geneve uses the TSTATS infrastructure (dev_sw_netstats_*()) for RX
packet counters. All other counters are handled using atomic increments
with DEV_STATS_INC().
Let's convert packet stats handling to DSTATS, which has a per-cpu
counter for packet drops too, to avoid the cost of atomic increments
in these cases. Statistics that don't fit DSTATS are still updated
atomically with DEV_STATS_INC().
Guillaume Nault [Wed, 4 Dec 2024 12:11:27 +0000 (13:11 +0100)]
vxlan: Handle stats using NETDEV_PCPU_STAT_DSTATS.
VXLAN uses the TSTATS infrastructure (dev_sw_netstats_*()) for RX and
TX packet counters. It also uses the device core stats
(dev_core_stats_*()) for RX and TX drops.
Let's consolidate that using the DSTATS infrastructure, which can
handle both packet counters and packet drops. Statistics that don't
fit DSTATS are still updated atomically with DEV_STATS_INC().
While there, convert the "len" variable of vxlan_encap_bypass() to
unsigned int, to respect the types of skb->len and
dev_dstats_[rt]x_add().
Guillaume Nault [Wed, 4 Dec 2024 12:11:21 +0000 (13:11 +0100)]
vrf: Make pcpu_dstats update functions available to other modules.
Currently vrf is the only module that uses NETDEV_PCPU_STAT_DSTATS.
In order to make this kind of statistics available to other modules,
we need to define the update functions in netdevice.h.
Therefore, let's define dev_dstats_*() functions for RX and TX packet
updates (packets, bytes and drops). Use these new functions in vrf.c
instead of vrf_rx_stats() and the other manual counter updates.
While there, update the type of the "len" variables to "unsigned int",
so that there're aligned with both skb->len and the new dstats update
functions.
Jakub Kicinski [Sat, 7 Dec 2024 01:53:29 +0000 (17:53 -0800)]
Merge branch 'lan78xx-preparations-for-phylink'
Oleksij Rempel says:
====================
lan78xx: Preparations for PHYlink
This patch set is part of the preparatory work for migrating the lan78xx
USB Ethernet driver to the PHYlink framework. During extensive testing,
I observed that resetting the USB adapter can lead to various read/write
errors. While the errors themselves are acceptable, they generate
excessive log messages, resulting in significant log spam. This set
improves error handling to reduce logging noise by addressing errors
directly and returning early when necessary.
Key highlights of this series include:
- Enhanced error handling to reduce log spam while preserving the
original error values, avoiding unnecessary overwrites.
- Improved error reporting using the `%pe` specifier for better clarity
in log messages.
- Removal of redundant and problematic PHY fixups for LAN8835 and
KSZ9031, with detailed explanations in the respective patches.
- Cleanup of code structure, including unified `goto` labels for better
readability and maintainability, even in simple editors.
====================
Oleksij Rempel [Wed, 4 Dec 2024 08:41:42 +0000 (09:41 +0100)]
net: usb: lan78xx: Improve error handling in dataport and multicast writes
Update `lan78xx_dataport_write` and `lan78xx_deferred_multicast_write`
to:
- Handle errors during register read/write operations.
- Exit immediately on errors and log them using `%pe` for clarity.
- Avoid silent failures by propagating error codes properly.
Oleksij Rempel [Wed, 4 Dec 2024 08:41:41 +0000 (09:41 +0100)]
net: usb: lan78xx: Add error handling to lan78xx_irq_bus_sync_unlock
Update `lan78xx_irq_bus_sync_unlock` to handle errors in register
read/write operations. If an error occurs, log it and exit the function
appropriately. This ensures proper handling of failures during IRQ
synchronization.
Oleksij Rempel [Wed, 4 Dec 2024 08:41:39 +0000 (09:41 +0100)]
net: usb: lan78xx: Add error handling to lan78xx_init_ltm
Convert `lan78xx_init_ltm` to return error codes and handle errors
properly. Previously, errors during the LTM initialization process were
not propagated, potentially leading to undetected issues. This patch
ensures:
- Errors in `lan78xx_read_reg` and `lan78xx_write_reg` are checked and
handled.
- Errors are logged with detailed messages using `%pe` for clarity.
- The function exits immediately on error, returning the error code.
Oleksij Rempel [Wed, 4 Dec 2024 08:41:38 +0000 (09:41 +0100)]
net: usb: lan78xx: Improve error handling in EEPROM and OTP operations
Refine error handling in EEPROM and OTP read/write functions by:
- Return error values immediately upon detection.
- Avoid overwriting correct error codes with `-EIO`.
- Preserve initial error codes as they were appropriate for specific
failures.
- Use `-ETIMEDOUT` for timeout conditions instead of `-EIO`.
Oleksij Rempel [Wed, 4 Dec 2024 08:41:37 +0000 (09:41 +0100)]
net: usb: lan78xx: Fix error handling in MII read/write functions
Ensure proper error handling in `lan78xx_mdiobus_read` and
`lan78xx_mdiobus_write` by checking return values of register read/write
operations and returning errors to the caller.
Oleksij Rempel [Wed, 4 Dec 2024 08:41:36 +0000 (09:41 +0100)]
net: usb: lan78xx: Improve error reporting with %pe specifier
Replace integer error codes with the `%pe` format specifier in register
read and write error messages. This change provides human-readable error
strings, making logs more informative and debugging easier.
Oleksij Rempel [Wed, 4 Dec 2024 08:41:34 +0000 (09:41 +0100)]
net: usb: lan78xx: Remove KSZ9031 PHY fixup
Remove the KSZ9031RNX PHY fixup from the lan78xx driver. The fixup applied
specific RGMII pad skew configurations globally, but these settings violate the
RGMII specification and cause more harm than benefit.
Key issues with the fixup:
1. **Non-Compliant Timing**: The fixup's delay settings fall outside the RGMII
specification requirements of 1.5 ns to 2.0 ns:
- RX Path: Total delay of **2.16 ns** (PHY internal delay of 1.2 ns + 0.96
ns skew).
- TX Path: Total delay of **0.96 ns**, significantly below the RGMII minimum
of 1.5 ns.
2. **Redundant or Incorrect Configurations**:
- The RGMII skew registers written by the fixup do not meaningfully alter
the PHY's default behavior and fail to account for its internal delays.
- The TX_DATA pad skew was not configured, relying on power-on defaults
that are insufficient for RGMII compliance.
3. **Micrel Driver Support**: By setting `PHY_INTERFACE_MODE_RGMII_ID`, the
Micrel driver can calculate and assign appropriate skew values for the
KSZ9031 PHY. This ensures better timing configurations without relying on
external fixups.
4. **System Interference**: The fixup applied globally, reconfiguring all
KSZ9031 PHYs in the system, even those unrelated to the LAN78xx adapter.
This could lead to unintended and harmful behavior on unrelated interfaces.
While the fixup is removed, a better mechanism is still needed to dynamically
determine the optimal combination of PHY and MAC delays to fully meet RGMII
requirements without relying on Device Tree or global fixups. This would allow
for robust operation across different hardware configurations.
The Micrel driver is capable of using the interface mode value to calculate and
apply better skew values, providing a configuration much closer to the RGMII
specification than the fixup. Removing the fixup ensures better default
behavior and prevents harm to other system interfaces.
Oleksij Rempel [Wed, 4 Dec 2024 08:41:33 +0000 (09:41 +0100)]
net: usb: lan78xx: Remove LAN8835 PHY fixup
Remove the PHY fixup for the LAN8835 PHY in the lan78xx driver due to
the following reasons:
- There is no publicly available information about the LAN8835 PHY.
However, it appears to be the integrated PHY used in the LAN7800 and
LAN7850 USB Ethernet controllers. These PHYs use the GMII interface,
not RGMII as configured by the fixup.
- The correct driver for handling the LAN8835 PHY functionality is the
Microchip PHY driver (`drivers/net/phy/microchip.c`), which properly
supports these integrated PHYs.
- The PHY ID `0x0007C130` is actually used by the LAN8742A PHY, which
only supports RMII. This interface is incompatible with the LAN78xx
MAC, as the LAN7801 (the only LAN78xx version without an integrated
PHY) supports only RGMII.
- The mask applied for this fixup is overly broad, inadvertently
covering both Microchip LAN88xx PHYs and unrelated SMSC LAN8742A PHYs,
leading to potential conflicts with other devices.
- Testing has shown that removing this fixup for LAN7800 and LAN7850
does not result in any noticeable difference in functionality, as the
Microchip PHY driver (`drivers/net/phy/microchip.c`) handles all
necessary configurations for these integrated PHYs.
- Registering this fixup globally (not limited to USB devices) risks
conflicts by unintentionally modifying other interfaces whenever a
LAN7801 adapter is connected to the system.
Note that both LAN7800 and LAN7850 USB Ethernet controllers use an
integrated PHY with the ID `0x0007C132`. Additionally, the LAN7515, a
specialized part for Raspberry Pi, includes an integrated LAN7800 USB
Ethernet controller and USB hub in a multifunctional chip design, and it
also uses the same PHY ID (`0x0007C132`).
Jakub Kicinski [Sat, 7 Dec 2024 01:47:34 +0000 (17:47 -0800)]
Merge branch 'net-phylib-eee-cleanups'
Russell King says:
====================
net: phylib EEE cleanups
Clean up phylib's EEE support. Patches previously posted as RFC as part
of the phylink EEE series.
Patch 1 changes the Marvell driver to use the state we store in
struct phy_device, rather than manually calling
phydev->eee_cfg.eee_enabled.
Patch 2 avoids genphy_c45_ethtool_get_eee() setting ->eee_enabled, as
we copy that from phydev->eee_cfg.eee_enabled later, and after patch 3
mo one uses this after calling genphy_c45_ethtool_get_eee(). In fact,
the only caller of this function now is phy_ethtool_get_eee().
As all callers to genphy_c45_eee_is_active() now pass NULL as its
is_enabled flag, this is no longer useful. Remove the argument in
patch 3.
Patch 4 updates the phylib documentation to make it absolutely clear
that phy_ethtool_get_eee() now fills in all members of struct
ethtool_keee, which is why we now have so many buggy network drivers.
====================
All callers to genphy_c45_eee_is_active() now pass NULL as the
is_enabled argument, which means we never use the value computed
in this function. Remove the argument and clean up this function.
genphy_c45_ethtool_get_eee() is only called from phy_ethtool_get_eee(),
which then calls eeecfg_to_eee(). eeecfg_to_eee() will overwrite
keee.eee_enabled, so there's no point setting keee.eee_enabled in
genphy_c45_ethtool_get_eee(). Remove this assignment.
Joe Damato [Wed, 4 Dec 2024 16:32:39 +0000 (16:32 +0000)]
selftests: net: cleanup busy_poller.c
Fix various integer type conversions by using strtoull and a temporary
variable which is bounds checked before being casted into the
appropriate cfg_* variable for use by the test program.
While here:
- free the strdup'd cfg string for overall hygenie.
- initialize napi_id = 0 in setup_queue to avoid warnings on some
compilers.
Eric Dumazet [Wed, 4 Dec 2024 21:02:34 +0000 (21:02 +0000)]
net: tipc: remove one synchronize_net() from tipc_nametbl_stop()
tipc_exit_net() is very slow and is abused by syzbot.
tipc_nametbl_stop() is called for each netns being dismantled.
Calling synchronize_net() right before freeing tn->nametbl
is a big hammer.
Replace this with kfree_rcu().
Note that RCU is not properly used here, otherwise
tn->nametbl should be cleared before the synchronize_net()
or kfree_rcu(), or even before the cleanup loop.
We might need to fix this at some point.
Also note tipc uses other synchronize_rcu() calls,
more work is needed to make tipc_exit_net() much faster.
This is V3 of the phylink conversion for ucc_geth.
The main changes in this V3 are related to error handling in the patches
1 and 10 to report an error when the deprecated "interface" property is
found in DT. Doing so, I found and addressed some issues with the jump
labels in the error paths, impacting patches 1 and 10.
The rest of the changes are just a rebase on net-next.
Some of the V2 changes haven't been reviewed, so I stress out that I'm
still uncertain about the way WoL is handled is patches 4 and 10.
Thanks,
Maxime
Link to V1: https://lore.kernel.org/netdev/20241107170255.1058124-1-maxime.chevallier@bootlin.com/
Link to V2: https://lore.kernel.org/netdev/20241114153603.307872-1-maxime.chevallier@bootlin.com/
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
net: freescale: ucc_geth: Introduce a helper to check Reduced modes
A number of parallel MII interfaces also exist in a "Reduced" mode,
usually with higher clock rates and fewer data lines, to ease the
hardware design. This is what the 'R' stands for in RGMII, RMII, RTBI,
RXAUI, etc.
The UCC Geth controller has a special configuration bit that needs to be
set when the MII mode is one of the supported reduced modes.
Add a local helper for that.
Signed-off-by: Maxime Chevallier <maxime.chevallier@bootlin.com> Reviewed-by: Andrew Lunn <andrew@lunn.ch> Signed-off-by: David S. Miller <davem@davemloft.net>
net: freescale: ucc_geth: Move the serdes configuration around
The uec_configure_serdes() function deals with serialized linkmodes
settings. It's used during the link bringup sequence. It is planned to
be used during the phylink conversion for mac configuration, but it
needs to me moved around in the process. To make the phylink port
clearer, this commit moves the function without any feature change.
Signed-off-by: Maxime Chevallier <maxime.chevallier@bootlin.com> Reviewed-by: Andrew Lunn <andrew@lunn.ch> Signed-off-by: David S. Miller <davem@davemloft.net>
net: freescale: ucc_geth: Hardcode the preamble length to 7 bytes
The preamble length can be configured in ucc_geth, however it just
ends-up always being configured to 7 bytes, as nothing ever changes the
default value of 7.
Make that value the default value when the MACCFG2 register gets
initialized, and remove the code to configure that value altogether.
Reviewed-by: Andrew Lunn <andrew@lunn.ch> Signed-off-by: Maxime Chevallier <maxime.chevallier@bootlin.com> Signed-off-by: David S. Miller <davem@davemloft.net>
The frame length check is configured when the phy interface is setup.
However, it's configured according to an internal flag that is always
false. So, just make so that we disable the relevant bit in the MACCFG2
register upon accessing it for other MAC configuration operations.
Reviewed-by: Andrew Lunn <andrew@lunn.ch> Signed-off-by: Maxime Chevallier <maxime.chevallier@bootlin.com> Signed-off-by: David S. Miller <davem@davemloft.net>
net: freescale: ucc_geth: Use the correct type to store WoL opts
The WoL opts are represented through a bitmask stored in a u32. As this
mask is copied as-is in the driver, make sure we use the exact same type
to store them internally.
Signed-off-by: Maxime Chevallier <maxime.chevallier@bootlin.com> Reviewed-by: Andrew Lunn <andrew@lunn.ch> Signed-off-by: David S. Miller <davem@davemloft.net>
The get/set_wol ethtool ops rely on querying the PHY for its WoL
capabilities, checking for the presence of a PHY and a PHY interrupts
isn't enough. Address that by cleaning up the WoL configuration
sequence.
Signed-off-by: Maxime Chevallier <maxime.chevallier@bootlin.com> Reviewed-by: Andrew Lunn <andrew@lunn.ch> Signed-off-by: David S. Miller <davem@davemloft.net>
net: freescale: ucc_geth: Use netdev->phydev to access the PHY
As this driver pre-dates phylib, it uses a private pointer to get a
reference to the attached phy_device. Drop that pointer and use the
netdev's pointer instead.
Reviewed-by: Andrew Lunn <andrew@lunn.ch> Signed-off-by: Maxime Chevallier <maxime.chevallier@bootlin.com> Signed-off-by: David S. Miller <davem@davemloft.net>
net: freescale: ucc_geth: split adjust_link for phylink conversion
Preparing the phylink conversion, split the adjust_link callbaclk, by
clearly separating the mac configuration, link_up and link_down phases.
Reviewed-by: Andrew Lunn <andrew@lunn.ch> Signed-off-by: Maxime Chevallier <maxime.chevallier@bootlin.com> Signed-off-by: David S. Miller <davem@davemloft.net>
net: freescale: ucc_geth: Drop support for the "interface" DT property
In april 2007, ucc_geth was converted to phylib with :
commit 728de4c927a3 ("ucc_geth: migrate ucc_geth to phylib").
In that commit, the device-tree property "interface", that could be used to
retrieve the PHY interface mode was deprecated.
DTS files that still used that property were converted along the way, in
the following commit, also dating from april 2007 :
commit 0fd8c47cccb1 ("[POWERPC] Replace undocumented interface properties in dts files")
17 years later, there's no users of that property left and I hope it's
safe to say we can remove support from that in the ucc_geth driver,
making the probe() function a bit simpler.
Should there be any users that have a DT that was generated when 2.6.21 was
cutting-edge, print an error message with hints on how to convert the
devicetree if the 'interface' property is found.
With that property gone, we can greatly simplify the parsing of the
phy-interface-mode from the devicetree by using of_get_phy_mode(),
allowing the removal of the open-coded parsing in the driver.
Reviewed-by: Andrew Lunn <andrew@lunn.ch> Signed-off-by: Maxime Chevallier <maxime.chevallier@bootlin.com> Signed-off-by: David S. Miller <davem@davemloft.net>
====================
xdp: a fistful of generic changes pt. I
XDP for idpf is currently 6 chapters:
* convert Rx to libeth;
* convert Tx and stats to libeth;
* generic XDP and XSk code changes (you are here);
* generic XDP and XSk code additions;
* actual XDP for idpf via new libeth_xdp;
* XSk for idpf (via ^).
Part III does the following:
* improve &xdp_buff_xsk cacheline placement;
* does some cleanups with marking read-only bpf_prog and xdp_buff
arguments const for some generic functions;
* allows attaching already registered XDP memory model to RxQ info;
* makes system percpu page_pools valid XDP memory models;
* starts using netmems in the XDP core code (1 function);
* allows mixing pages from several page_pools within one XDP frame;
* optimizes &xdp_frame layout and removes no-more-used field.
Bullets 4-6 are the most important ones. All of them are prereqs to
libeth_xdp.
====================
page_pool: make page_pool_put_page_bulk() handle array of netmems
Currently, page_pool_put_page_bulk() indeed takes an array of pointers
to the data, not pages, despite the name. As one side effect, when
you're freeing frags from &skb_shared_info, xdp_return_frame_bulk()
converts page pointers to virtual addresses and then
page_pool_put_page_bulk() converts them back. Moreover, data pointers
assume every frag is placed in the host memory, making this function
non-universal.
Make page_pool_put_page_bulk() handle array of netmems. Pass frag
netmems directly and use virt_to_netmem() when freeing xdpf->data,
so that the PP core will then get the compound netmem and take care
of the rest.
They do the same as their non-underscored buddies, but assume the netmem
is always page-backed. When working with header &page_pools, you don't
need to check whether netmem belongs to the host memory and you can
never get NULL instead of &page. Checks for the LSB, clearing the LSB,
branches take cycles and increase object code size, sometimes
significantly. When you're sure your PP is always host, you can avoid
this by using the underscored counterparts.
xdp: register system page pool as an XDP memory model
To make the system page pool usable as a source for allocating XDP
frames, we need to register it with xdp_reg_mem_model(), so that page
return works correctly. This is done in preparation for using the system
page_pool to convert XDP_PASS XSk frames to skbs; for the same reason,
make the per-cpu variable non-static so we can access it from other
source files as well (but w/o exporting).
xsk: allow attaching XSk pool via xdp_rxq_info_reg_mem_model()
When you register an XSk pool as XDP Rxq info memory model, you then
need to manually attach it after the registration.
Let the user combine both actions into one by just passing a pointer
to the pool directly to xdp_rxq_info_reg_mem_model(), which will take
care of calling xsk_pool_set_rxq_info(). This looks similar to how a
&page_pool gets registered and reduce repeating driver code.
Acked-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com> Reviewed-by: Toke Høiland-Jørgensen <toke@redhat.com> Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com> Link: https://patch.msgid.link/20241203173733.3181246-6-aleksander.lobakin@intel.com Signed-off-by: Jakub Kicinski <kuba@kernel.org>
xdp: allow attaching already registered memory model to xdp_rxq_info
One may need to register memory model separately from xdp_rxq_info. One
simple example may be XDP test run code, but in general, it might be
useful when memory model registering is managed by one layer and then
XDP RxQ info by a different one.
Allow such scenarios by adding a simple helper which "attaches"
already registered memory model to the desired xdp_rxq_info. As this
is mostly needed for Page Pool, add a special function to do that for
a &page_pool pointer.
xdp, xsk: constify read-only arguments of some static inline helpers
Lots of read-only helpers for &xdp_buff and &xdp_frame, such as getting
the frame length, skb_shared_info etc., don't have their arguments
marked with `const` for no reason. Add the missing annotations to leave
less place for mistakes and more for optimization.
bpf, xdp: constify some bpf_prog * function arguments
In lots of places, bpf_prog pointer is used only for tracing or other
stuff that doesn't modify the structure itself. Same for net_device.
Address at least some of them and add `const` attributes there. The
object code didn't change, but that may prevent unwanted data
modifications and also allow more helpers to have const arguments.
Reviewed-by: Toke Høiland-Jørgensen <toke@redhat.com> Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
After the series "XSk buff on a diet" by Maciej, the greatest pow-2
which &xdp_buff_xsk can be divided got reduced from 16 to 8 on x86_64.
Also, sizeof(xdp_buff_xsk) now is 120 bytes, which, taking the previous
sentence into account, leads to that it leaves 8 bytes at the end of
cacheline, which means an array of buffs will have its elements
messed between the cachelines chaotically.
Use __aligned_largest for this struct. This alignment is usually 16
bytes, which makes it fill two full cachelines and align an array
nicely. ___cacheline_aligned may be excessive here, especially on
arches with 128-256 byte CLs, as well as 32-bit arches (76 -> 96
bytes on MIPS32R2), while not doing better than _largest.
====================
net_sched: sch_sfq: reject limit of 1
The implementation does not properly support limits of 1. Add an
in-kernel check, in addition to existing iproute2 check, since other
tools may be used for configuration.
This patch set also adds a selfcheck to test that a limit of 1 is
rejected.
An alternative (or in addition) we could fix the implementation by
setting q->tail to NULL in sfq_drop if this is the last slot we marked
empty, e.g.:
--- a/net/sched/sch_sfq.c
+++ b/net/sched/sch_sfq.c
@@ -317,8 +317,11 @@ static unsigned int sfq_drop(struct Qdisc *sch, struct sk_buff **to_free)
/* It is difficult to believe, but ALL THE SLOTS HAVE LENGTH 1. */
x = q->tail->next;
slot = &q->slots[x];
- q->tail->next = slot->next;
q->ht[slot->hash] = SFQ_EMPTY_SLOT;
+ if (x == slot->next)
+ q->tail = NULL; /* no more active slots */
+ else
+ q->tail->next = slot->next;
goto drop;
}
====================
The current implementation does not work correctly with a limit of
1. iproute2 actually checks for this and this patch adds the check in
kernel as well.
This fixes the following syzkaller reported crash:
The crash can be also be reproduced with the following (with a tc
recompiled to allow for sfq limits of 1):
tc qdisc add dev dummy0 handle 1: root tbf rate 1Kbit burst 100b lat 1s
../iproute2-6.9.0/tc/tc qdisc add dev dummy0 handle 2: parent 1:10 sfq limit 1
ifconfig dummy0 up
ping -I dummy0 -f -c2 -W0.1 8.8.8.8
sleep 1
Scenario that triggers the crash:
* the first packet is sent and queued in TBF and SFQ; qdisc qlen is 1
* TBF dequeues: it peeks from SFQ which moves the packet to the
gso_skb list and keeps qdisc qlen set to 1. TBF is out of tokens so
it schedules itself for later.
* the second packet is sent and TBF tries to queues it to SFQ. qdisc
qlen is now 2 and because the SFQ limit is 1 the packet is dropped
by SFQ. At this point qlen is 1, and all of the SFQ slots are empty,
however q->tail is not NULL.
At this point, assuming no more packets are queued, when sch_dequeue
runs again it will decrement the qlen for the current empty slot
causing an underflow and the subsequent out of bounds access.
Reported-by: syzbot <syzkaller@googlegroups.com> Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2") Signed-off-by: Octavian Purdila <tavip@google.com> Reviewed-by: Eric Dumazet <edumazet@google.com> Link: https://patch.msgid.link/20241204030520.2084663-2-tavip@google.com Signed-off-by: Jakub Kicinski <kuba@kernel.org>
====================
ethtool: generate uapi header from the spec
We keep expanding ethtool netlink api surface and this leads to
constantly playing catchup on the ynl spec side. There are a couple
of things that prevent us from fully converting to generating
the header from the spec (stats and cable tests), but we can
generate 95% of the header which is still better than maintaining
c header and spec separately. The series adds a couple of missing
features on the ynl-gen-c side and separates the parts
that we can generate into new ethtool_netlink_generated.h.
====================
ethtool: separate definitions that are gonna be generated
Reshuffle definitions that are gonna be generated into
ethtool_netlink_generated.h and match ynl spec order.
This should make it easier to compare the output of the ynl-gen-c
to the existing uapi header. No functional changes.
Things that are still remaining to be manually defined:
- ETHTOOL_FLAG_ALL - probably no good way to add to spec?
- some of the cable test bits (not sure whether it's possible to move to
spec)
- some of the stats definitions (no way currently to move to spec)
ynl: add missing pieces to ethtool spec to better match uapi header
- __ETHTOOL_UDP_TUNNEL_TYPE_CNT and render max
- skip rendering stringset (empty enum)
- skip rendering c33-pse-ext-state (defined in ethtool.h)
- rename header flags to ethtool-flag-
- add attr-cnt-name to each attribute to use XXX_CNT instead of XXX_MAX
- add unspec 0 entry to each attribute
- carry some doc entries from the existing header
- tcp-header-split
The intent is to generate ethtool uapi headers. For now, some of the
things are hard-coded:
- <FAMILY>_MSG_{USER,KERNEL}_MAX
- the split between USER and KERNEL messages
- ipv6:
- avoid possible NULL deref in modify_prefix_route()
- release expired exception dst cached in socket
- smc: fix LGR and link use-after-free issue
- hsr: avoid potential out-of-bound access in fill_frame_info()
- can: hi311x: fix potential use-after-free
- eth: ice: fix VLAN pruning in switchdev mode
Previous releases - always broken:
- netfilter:
- ipset: hold module reference while requesting a module
- nft_inner: incorrect percpu area handling under softirq
- can: j1939: fix skb reference counting
- eth:
- mlxsw: use correct key block on Spectrum-4
- mlx5: fix memory leak in mlx5hws_definer_calc_layout"
* tag 'net-6.13-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (76 commits)
net :mana :Request a V2 response version for MANA_QUERY_GF_STAT
net: avoid potential UAF in default_operstate()
vsock/test: verify socket options after setting them
vsock/test: fix parameter types in SO_VM_SOCKETS_* calls
vsock/test: fix failures due to wrong SO_RCVLOWAT parameter
net/mlx5e: Remove workaround to avoid syndrome for internal port
net/mlx5e: SD, Use correct mdev to build channel param
net/mlx5: E-Switch, Fix switching to switchdev mode in MPV
net/mlx5: E-Switch, Fix switching to switchdev mode with IB device disabled
net/mlx5: HWS: Properly set bwc queue locks lock classes
net/mlx5: HWS: Fix memory leak in mlx5hws_definer_calc_layout
bnxt_en: handle tpa_info in queue API implementation
bnxt_en: refactor bnxt_alloc_rx_rings() to call bnxt_alloc_rx_agg_bmap()
bnxt_en: refactor tpa_info alloc/free into helpers
geneve: do not assume mac header is set in geneve_xmit_skb()
mlxsw: spectrum_acl_flex_keys: Use correct key block on Spectrum-4
ethtool: Fix wrong mod state in case of verbose and no_mask bitset
ipmr: tune the ipmr_can_free_table() checks.
netfilter: nft_set_hash: skip duplicated elements pending gc run
netfilter: ipset: Hold module reference while requesting a module
...
Linus Torvalds [Thu, 5 Dec 2024 18:17:55 +0000 (10:17 -0800)]
Merge tag 'trace-v6.13-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace
Pull tracing fixes from Steven Rostedt:
- Fix trace histogram sort function cmp_entries_dup()
The sort function cmp_entries_dup() returns either 1 or 0, and not -1
if parameter "a" is less than "b" by memcmp().
- Fix archs that call trace_hardirqs_off() without RCU watching
Both x86 and arm64 no longer call any tracepoints with RCU not
watching. It was assumed that it was safe to get rid of
trace_*_rcuidle() version of the tracepoint calls. This was needed to
get rid of the SRCU protection and be able to implement features like
faultable traceponits and add rust tracepoints.
Unfortunately, there were a few architectures that still relied on
that logic. There's only one file that has tracepoints that are
called without RCU watching. Add macro logic around the tracepoints
for architectures that do not have CONFIG_ARCH_WANTS_NO_INSTR defined
will check if the code is in the idle path (the only place RCU isn't
watching), and enable RCU around calling the tracepoint, but only do
it if the tracepoint is enabled.
* tag 'trace-v6.13-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace:
tracing: Fix archs that still call tracepoints without RCU watching
tracing: Fix cmp_entries_dup() to respect sort() comparison rules
Linus Torvalds [Thu, 5 Dec 2024 18:06:47 +0000 (10:06 -0800)]
Merge tag 'hid-for-linus-2024120501' of git://git.kernel.org/pub/scm/linux/kernel/git/hid/hid
Pull HID fixes from Benjamin Tissoires:
- regression fix in suspend/resume for i2c-hid (Kenny Levinsen)
- fix wacom driver assuming a name can not be null (WangYuli)
- a couple of constify changes/fixes (Thomas Weißschuh)
- a couple of selftests/hid fixes (Maximilian Heyne & Benjamin
Tissoires)
* tag 'hid-for-linus-2024120501' of git://git.kernel.org/pub/scm/linux/kernel/git/hid/hid:
selftests/hid: fix kfunc inclusions with newer bpftool
HID: bpf: drop unneeded casts discarding const
HID: bpf: constify hid_ops
selftests: hid: fix typo and exit code
HID: wacom: fix when get product name maybe null pointer
HID: i2c-hid: Revert to using power commands to wake on resume
Linus Torvalds [Thu, 5 Dec 2024 18:03:43 +0000 (10:03 -0800)]
Merge tag 'linux-watchdog-6.13-rc1' of git://www.linux-watchdog.org/linux-watchdog
Pull watchdog updates from Wim Van Sebroeck:
- Add support for exynosautov920 SoC
- Add support for Airoha EN7851 watchdog
- Add support for MT6735 TOPRGU/WDT
- Delete the cpu5wdt driver
- Always print when registering watchdog fails
- Several other small fixes and improvements
* tag 'linux-watchdog-6.13-rc1' of git://www.linux-watchdog.org/linux-watchdog: (36 commits)
watchdog: rti: of: honor timeout-sec property
watchdog: s3c2410_wdt: add support for exynosautov920 SoC
dt-bindings: watchdog: Document ExynosAutoV920 watchdog bindings
watchdog: mediatek: Add support for MT6735 TOPRGU/WDT
watchdog: mediatek: Make sure system reset gets asserted in mtk_wdt_restart()
dt-bindings: watchdog: fsl-imx-wdt: Add missing 'big-endian' property
dt-bindings: watchdog: Document Qualcomm QCS8300
docs: ABI: Fix spelling mistake in pretimeout_avaialable_governors
Revert "watchdog: s3c2410_wdt: use exynos_get_pmu_regmap_by_phandle() for PMU regs"
watchdog: rzg2l_wdt: Power on the watchdog domain in the restart handler
watchdog: Switch back to struct platform_driver::remove()
watchdog: it87_wdt: add PWRGD enable quirk for Qotom QCML04
watchdog: da9063: Remove __maybe_unused notations
watchdog: da9063: Do not use a global variable
watchdog: Delete the cpu5wdt driver
watchdog: Add support for Airoha EN7851 watchdog
dt-bindings: watchdog: airoha: document watchdog for Airoha EN7581
watchdog: sl28cpld_wdt: don't print out if registering watchdog fails
watchdog: rza_wdt: don't print out if registering watchdog fails
watchdog: rti_wdt: don't print out if registering watchdog fails
...
Steven Rostedt [Wed, 4 Dec 2024 15:04:14 +0000 (10:04 -0500)]
tracing: Fix archs that still call tracepoints without RCU watching
Tracepoints require having RCU "watching" as it uses RCU to do updates to
the tracepoints. There are some cases that would call a tracepoint when
RCU was not "watching". This was usually in the idle path where RCU has
"shutdown". For the few locations that had tracepoints without RCU
watching, there was an trace_*_rcuidle() variant that could be used. This
used SRCU for protection.
There are tracepoints that trace when interrupts and preemption are
enabled and disabled. In some architectures, these tracepoints are called
in a path where RCU is not watching. When x86 and arm64 removed these
locations, it was incorrectly assumed that it would be safe to remove the
trace_*_rcuidle() variant and also remove the SRCU logic, as it made the
code more complex and harder to implement new tracepoint features (like
faultable tracepoints and tracepoints in rust).
Instead of bringing back the trace_*_rcuidle(), as it will not be trivial
to do as new code has already been added depending on its removal, add a
workaround to the one file that still requires it (trace_preemptirq.c). If
the architecture does not define CONFIG_ARCH_WANTS_NO_INSTR, then check if
the code is in the idle path, and if so, call ct_irq_enter/exit() which
will enable RCU around the tracepoint.
Cc: Masami Hiramatsu <mhiramat@kernel.org> Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mark Rutland <mark.rutland@arm.com> Link: https://lore.kernel.org/20241204100414.4d3e06d0@gandalf.local.home Reported-by: Geert Uytterhoeven <geert@linux-m68k.org> Fixes: 48bcda684823 ("tracing: Remove definition of trace_*_rcuidle()") Closes: https://lore.kernel.org/all/bddb02de-957a-4df5-8e77-829f55728ea2@roeck-us.net/ Acked-by: Paul E. McKenney <paulmck@kernel.org> Tested-by: Guenter Roeck <linux@roeck-us.net> Tested-by: Geert Uytterhoeven <geert+renesas@glider.be> Tested-by: Madhavan Srinivasan <maddy@linux.ibm.com> Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Shradha Gupta [Wed, 4 Dec 2024 05:48:20 +0000 (21:48 -0800)]
net :mana :Request a V2 response version for MANA_QUERY_GF_STAT
The current requested response version(V1) for MANA_QUERY_GF_STAT query
results in STATISTICS_FLAGS_TX_ERRORS_GDMA_ERROR value being set to
0 always.
In order to get the correct value for this counter we request the response
version to be V2.
Cc: stable@vger.kernel.org Fixes: e1df5202e879 ("net :mana :Add remaining GDMA stats for MANA to ethtool") Signed-off-by: Shradha Gupta <shradhagupta@linux.microsoft.com> Reviewed-by: Haiyang Zhang <haiyangz@microsoft.com> Link: https://patch.msgid.link/1733291300-12593-1-git-send-email-shradhagupta@linux.microsoft.com Signed-off-by: Paolo Abeni <pabeni@redhat.com>
The buggy address belongs to the object at ffff888043eba000
which belongs to the cache kmalloc-2k of size 2048
The buggy address is located 432 bytes inside of
freed 2048-byte region [ffff888043eba000, ffff888043eba800)
Fixes: 8c55facecd7a ("net: linkwatch: only report IF_OPER_LOWERLAYERDOWN if iflink is actually down") Reported-by: syzbot+1939f24bdb783e9e43d9@syzkaller.appspotmail.com Closes: https://lore.kernel.org/netdev/674f3a18.050a0220.48a03.0041.GAE@google.com/T/#u Signed-off-by: Eric Dumazet <edumazet@google.com> Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com> Link: https://patch.msgid.link/20241203170933.2449307-1-edumazet@google.com Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Paolo Abeni [Thu, 5 Dec 2024 10:49:14 +0000 (11:49 +0100)]
Merge tag 'nf-24-12-05' of git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf
Pablo Neira Ayuso says:
====================
Netfilter fixes for net
The following patchset contains Netfilter fixes for net:
1) Fix esoteric undefined behaviour due to uninitialized stack access
in ip_vs_protocol_init(), from Jinghao Jia.
2) Fix iptables xt_LED slab-out-of-bounds due to incorrect sanitization
of the led string identifier, reported by syzbot. Patch from
Dmitry Antipov.
3) Remove WARN_ON_ONCE reachable from userspace to check for the maximum
cgroup level, nft_socket cgroup matching is restricted to 255 levels,
but cgroups allow for INT_MAX levels by default. Reported by syzbot.
4) Fix nft_inner incorrect use of percpu area to store tunnel parser
context with softirqs, resulting in inconsistent inner header
offsets that could lead to bogus rule mismatches, reported by syzbot.
5) Grab module reference on ipset core while requesting set type modules,
otherwise kernel crash is possible by removing ipset core module,
patch from Phil Sutter.
6) Fix possible double-free in nft_hash garbage collector due to unstable
walk interator that can provide twice the same element. Use a sequence
number to skip expired/dead elements that have been already scheduled
for removal. Based on patch from Laurent Fasnach
netfilter pull request 24-12-05
* tag 'nf-24-12-05' of git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf:
netfilter: nft_set_hash: skip duplicated elements pending gc run
netfilter: ipset: Hold module reference while requesting a module
netfilter: nft_inner: incorrect percpu area handling under softirq
netfilter: nft_socket: remove WARN_ON_ONCE on maximum cgroup level
netfilter: x_tables: fix LED ID check in led_tg_check()
ipvs: fix UB due to uninitialized stack access in ip_vs_protocol_init()
====================
Parameters were created using wrong C types, which caused them to be of
wrong size on some architectures, causing problems.
The problem with SO_RCVLOWAT was found on s390 (big endian), while x86-64
didn't show it. After the fix, all tests pass on s390.
Then Stefano Garzarella pointed out that SO_VM_SOCKETS_* calls might have
a similar problem, which turned out to be true, hence, the second patch.
Changes for v8:
- Fix whitespace warnings from "checkpatch.pl --strict"
- Add maintainers to Cc:
Changes for v7:
- Rebase on top of https://git.kernel.org/pub/scm/linux/kernel/git/netdev/net.git
- Add the "net" tags to the subjects
Changes for v6:
- rework the patch #3 to avoid creating a new file for new functions,
and exclude vsock_perf from calling the new functions.
- add "Reviewed-by:" to the patch #2.
Changes for v5:
- in the patch #2 replace the introduced uint64_t with unsigned long long
to match documentation
- add a patch #3 that verifies every setsockopt() call.
Changes for v4:
- add "Reviewed-by:" to the first patch, and add a second patch fixing
SO_VM_SOCKETS_* calls, which depends on the first one (hence, it's now
a patch series.)
Changes for v3:
- fix the same problem in vsock_perf and update commit message
Changes for v2:
- add "Fixes:" lines to the commit message
====================
vsock/test: verify socket options after setting them
Replace setsockopt() calls with calls to functions that follow
setsockopt() with getsockopt() and check that the returned value and its
size are the same as have been set. (Except in vsock_perf.)
Signed-off-by: Konstantin Shkolnyy <kshk@linux.ibm.com> Reviewed-by: Stefano Garzarella <sgarzare@redhat.com> Signed-off-by: Paolo Abeni <pabeni@redhat.com>
vsock/test: fix parameter types in SO_VM_SOCKETS_* calls
Change parameters of SO_VM_SOCKETS_* to unsigned long long as documented
in the vm_sockets.h, because the corresponding kernel code requires them
to be at least 64-bit, no matter what architecture. Otherwise they are
too small on 32-bit machines.
Fixes: 5c338112e48a ("test/vsock: rework message bounds test") Fixes: 685a21c314a8 ("test/vsock: add big message test") Fixes: 542e893fbadc ("vsock/test: two tests to check credit update logic") Fixes: 8abbffd27ced ("test/vsock: vsock_perf utility") Signed-off-by: Konstantin Shkolnyy <kshk@linux.ibm.com> Reviewed-by: Stefano Garzarella <sgarzare@redhat.com> Signed-off-by: Paolo Abeni <pabeni@redhat.com>
vsock/test: fix failures due to wrong SO_RCVLOWAT parameter
This happens on 64-bit big-endian machines.
SO_RCVLOWAT requires an int parameter. However, instead of int, the test
uses unsigned long in one place and size_t in another. Both are 8 bytes
long on 64-bit machines. The kernel, having received the 8 bytes, doesn't
test for the exact size of the parameter, it only cares that it's >=
sizeof(int), and casts the 4 lower-addressed bytes to an int, which, on
a big-endian machine, contains 0. 0 doesn't trigger an error, SO_RCVLOWAT
returns with success and the socket stays with the default SO_RCVLOWAT = 1,
which results in vsock_test failures, while vsock_perf doesn't even notice
that it's failed to change it.
Fixes: b1346338fbae ("vsock_test: POLLIN + SO_RCVLOWAT test") Fixes: 542e893fbadc ("vsock/test: two tests to check credit update logic") Fixes: 8abbffd27ced ("test/vsock: vsock_perf utility") Signed-off-by: Konstantin Shkolnyy <kshk@linux.ibm.com> Reviewed-by: Stefano Garzarella <sgarzare@redhat.com> Signed-off-by: Paolo Abeni <pabeni@redhat.com>
====================
Mitigate the two-reallocations issue for iptunnels
RESEND v5:
- v5 was sent just when net-next closed
v5:
- address Paolo's comments
- s/int dst_dev_overhead()/unsigned int dst_dev_overhead()/
v4:
- move static inline function to include/net/dst.h
v3:
- fix compilation error in seg6_iptunnel
v2:
- add missing "static" keywords in seg6_iptunnel
- use a static-inline function to return the dev overhead (as suggested
by Olek, thanks)
The same pattern is found in ioam6, rpl6, and seg6. Basically, it first
makes sure there is enough room for inserting a new header:
(1) err = skb_cow_head(skb, len + skb->mac_len);
Then, when the insertion (encap or inline) is performed, the input and
output handlers respectively make sure there is enough room for layer 2:
skb_cow_head() does nothing when there is enough room. Otherwise, it
reallocates more room, which depends on the architecture. Briefly,
skb_cow_head() calls __skb_cow() which then calls pskb_expand_head() as
follows:
"delta" represents the number of bytes to be added. This value is
aligned with NET_SKB_PAD, which is defined as follows:
NET_SKB_PAD = max(32, L1_CACHE_BYTES)
... where L1_CACHE_BYTES also depends on the architecture. In our case
(x86), it is defined as follows:
L1_CACHE_BYTES = (1 << CONFIG_X86_L1_CACHE_SHIFT)
... where (again, in our case) CONFIG_X86_L1_CACHE_SHIFT equals 6
(=X86_GENERIC).
All this to say, skb_cow_head() would reallocate to the next multiple of
NET_SKB_PAD (in our case a 64-byte multiple) when there is not enough
room.
Back to the main issue with the pattern: in some cases, two
reallocations are triggered, resulting in a performance drop (i.e.,
lines (1) and (2) would both trigger an implicit reallocation). How's
that possible? Well, this is kind of bad luck as we hit an exact
NET_SKB_PAD boundary and when skb->mac_len (=14) is smaller than
LL_RESERVED_SPACE(dst->dev) (=16 in our case). For an x86 arch, it
happens in the following cases (with the default needed_headroom):
- ioam6:
- (inline mode) pre-allocated data trace of 236 or 240 bytes
- (encap mode) pre-allocated data trace of 196 or 200 bytes
- seg6:
- (encap mode) for 13, 17, 21, 25, 29, 33, ...(+4)... prefixes
Let's illustrate the problem, i.e., when we fall on the exact
NET_SKB_PAD boundary. In the case of ioam6, for the above problematic
values, the total overhead is 256 bytes for both modes. Based on line
(1), skb->mac_len (=14) is added, therefore passing 270 bytes to
skb_cow_head(). At that moment, the headroom has 206 bytes available (in
our case). Since 270 > 206, skb_cow_head() performs a reallocation and
the new headroom is now 206 + 64 (NET_SKB_PAD) = 270. Which is exactly
the room we needed. After the insertion, the headroom has 0 byte
available. But, there's line (2) where 16 bytes are still needed. Which,
again, triggers another reallocation.
The same logic is applied to seg6 (although it does not happen with the
inline mode, i.e., -40 bytes). It happens with other L1 cache shifts too
(the larger the cache shift, the less often it happens). For example,
with a +32 cache shift (instead of +64), the following number of
segments would trigger two reallocations: 11, 15, 19, ... With a +128
cache shift, the following number of segments would trigger two
reallocations: 17, 25, 33, ... And so on and so forth. Note that it is
the same for both the "encap" and "l2encap" modes. For the "encap.red"
and "l2encap.red" modes, it is the same logic but with "segs+1" (e.g.,
14, 18, 22, 26, etc for a +64 cache shift). Note also that it may happen
with rpl6 (based on some calculations), although it did not in our case.
This series provides a solution to mitigate the aforementioned issue for
ioam6, seg6, and rpl6. It provides the dst_entry (in the cache) to
skb_cow_head() **before** the insertion (line (1)). As a result, the
very first iteration would still trigger two reallocations (i.e., empty
cache), while next iterations would only trigger a single reallocation.
====================
Justin Iurman [Tue, 3 Dec 2024 12:49:45 +0000 (13:49 +0100)]
net: ipv6: rpl_iptunnel: mitigate 2-realloc issue
This patch mitigates the two-reallocations issue with rpl_iptunnel by
providing the dst_entry (in the cache) to the first call to
skb_cow_head(). As a result, the very first iteration would still
trigger two reallocations (i.e., empty cache), while next iterations
would only trigger a single reallocation.
Performance tests before/after applying this patch, which clearly shows
there is no impact (it even shows improvement):
- before: https://ibb.co/nQJhqwc
- after: https://ibb.co/4ZvW6wV
Signed-off-by: Justin Iurman <justin.iurman@uliege.be> Cc: Alexander Aring <aahringo@redhat.com> Signed-off-by: Paolo Abeni <pabeni@redhat.com>
This patch mitigates the two-reallocations issue with seg6_iptunnel by
providing the dst_entry (in the cache) to the first call to
skb_cow_head(). As a result, the very first iteration would still
trigger two reallocations (i.e., empty cache), while next iterations
would only trigger a single reallocation.
Performance tests before/after applying this patch, which clearly shows
the improvement:
- before: https://ibb.co/3Cg4sNH
- after: https://ibb.co/8rQ350r
Signed-off-by: Justin Iurman <justin.iurman@uliege.be> Cc: David Lebrun <dlebrun@google.com> Signed-off-by: Paolo Abeni <pabeni@redhat.com>
This patch mitigates the two-reallocations issue with ioam6_iptunnel by
providing the dst_entry (in the cache) to the first call to
skb_cow_head(). As a result, the very first iteration may still trigger
two reallocations (i.e., empty cache), while next iterations would only
trigger a single reallocation.
Performance tests before/after applying this patch, which clearly shows
the improvement:
- inline mode:
- before: https://ibb.co/LhQ8V63
- after: https://ibb.co/x5YT2bS
- encap mode:
- before: https://ibb.co/3Cjm5m0
- after: https://ibb.co/TwpsxTC
- encap mode with tunsrc:
- before: https://ibb.co/Gpy9QPg
- after: https://ibb.co/PW1bZFT
This patch also fixes an incorrect behavior: after the insertion, the
second call to skb_cow_head() makes sure that the dev has enough
headroom in the skb for layer 2 and stuff. In that case, the "old"
dst_entry was used, which is now fixed. After discussing with Paolo, it
appears that both patches can be merged into a single one -this one-
(for the sake of readability) and target net-next.
Signed-off-by: Justin Iurman <justin.iurman@uliege.be> Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Justin Iurman [Tue, 3 Dec 2024 12:49:42 +0000 (13:49 +0100)]
include: net: add static inline dst_dev_overhead() to dst.h
Add static inline dst_dev_overhead() function to include/net/dst.h. This
helper function is used by ioam6_iptunnel, rpl_iptunnel and
seg6_iptunnel to get the dev's overhead based on a cache entry
(dst_entry). If the cache is empty, the default and generic value
skb->mac_len is returned. Otherwise, LL_RESERVED_SPACE() over dst's dev
is returned.
Signed-off-by: Justin Iurman <justin.iurman@uliege.be> Cc: Alexander Lobakin <aleksander.lobakin@intel.com> Cc: Vadim Fedorenko <vadim.fedorenko@linux.dev> Signed-off-by: Paolo Abeni <pabeni@redhat.com>
This series contains updates to ice, idpf, ixgbe, ixgbevf, and igb
drivers.
For ice:
Arkadiusz corrects search for determining whether PHY clock recovery is
supported on the device.
Przemyslaw corrects mask used for PHY timestamps on ETH56G devices.
Wojciech adds missing virtchnl ops which caused NULL pointer
dereference.
Marcin fixes VLAN filter settings for uplink VSI in switchdev mode.
For idpf:
Josh restores setting of completion tag for empty buffers.
For ixgbevf:
Jake removes incorrect initialization/support of IPSEC for mailbox
version 1.5.
For ixgbe:
Jake rewords and downgrades misleading message when negotiation
of VF mailbox version is not supported.
Tore Amundsen corrects value for BASE-BX10 capability.
For igb:
Yuan Can adds proper teardown on failed pci_register_driver() call.
* '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue:
igb: Fix potential invalid memory access in igb_init_module()
ixgbe: Correct BASE-BX10 compliance code
ixgbe: downgrade logging of unsupported VF API version to debug
ixgbevf: stop attempting IPSEC offload on Mailbox API 1.5
idpf: set completion tag for "empty" bufs associated with a packet
ice: Fix VLAN pruning in switchdev mode
ice: Fix NULL pointer dereference in switchdev
ice: fix PHY timestamp extraction for ETH56G
ice: fix PHY Clock Recovery availability check
====================
Jianbo Liu [Tue, 3 Dec 2024 20:49:20 +0000 (22:49 +0200)]
net/mlx5e: Remove workaround to avoid syndrome for internal port
Previously a workaround was added to avoid syndrome 0xcdb051. It is
triggered when offload a rule with tunnel encapsulation, and
forwarding to another table, but not matching on the internal port in
firmware steering mode. The original workaround skips internal tunnel
port logic, which is not correct as not all cases are considered. As
an example, if vlan is configured on the uplink port, traffic can't
pass because vlan header is not added with this workaround. Besides,
there is no such issue for software steering. So, this patch removes
that, and returns error directly if trying to offload such rule for
firmware steering.
Fixes: 06b4eac9c4be ("net/mlx5e: Don't offload internal port if filter device is out device") Signed-off-by: Jianbo Liu <jianbol@nvidia.com> Tested-by: Frode Nordahl <frode.nordahl@canonical.com> Reviewed-by: Chris Mi <cmi@nvidia.com> Reviewed-by: Ariel Levkovich <lariel@nvidia.com> Signed-off-by: Tariq Toukan <tariqt@nvidia.com> Link: https://patch.msgid.link/20241203204920.232744-7-tariqt@nvidia.com Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Tariq Toukan [Tue, 3 Dec 2024 20:49:19 +0000 (22:49 +0200)]
net/mlx5e: SD, Use correct mdev to build channel param
In a multi-PF netdev, each traffic channel creates its own resources
against a specific PF.
In the cited commit, where this support was added, the channel_param
logic was mistakenly kept unchanged, so it always used the primary PF
which is found at priv->mdev.
In this patch we fix this by moving the logic to be per-channel, and
passing the correct mdev instance.
This bug happened to be usually harmless, as the resulting cparam
structures would be the same for all channels, due to identical FW logic
and decisions.
However, in some use cases, like fwreset, this gets broken.
This could lead to different symptoms. Example:
Error cqe on cqn 0x428, ci 0x0, qn 0x10a9, opcode 0xe, syndrome 0x4,
vendor syndrome 0x32
Fixes: e4f9686bdee7 ("net/mlx5e: Let channels be SD-aware") Signed-off-by: Tariq Toukan <tariqt@nvidia.com> Reviewed-by: Lama Kayal <lkayal@nvidia.com> Reviewed-by: Gal Pressman <gal@nvidia.com> Link: https://patch.msgid.link/20241203204920.232744-6-tariqt@nvidia.com Signed-off-by: Jakub Kicinski <kuba@kernel.org>
net/mlx5: E-Switch, Fix switching to switchdev mode in MPV
Fix the mentioned commit change for MPV mode, since in MPV mode the IB
device is shared between different core devices, so under this change
when moving both devices simultaneously to switchdev mode the IB device
removal and re-addition can race with itself causing unexpected behavior.
In such case do rescan_drivers() only once in order to add the ethernet
representor auxiliary device, and skip adding and removing IB devices.
net/mlx5: E-Switch, Fix switching to switchdev mode with IB device disabled
In case that IB device is already disabled when moving to switchdev mode,
which can happen when working with LAG, need to do rescan_drivers()
before leaving in order to add ethernet representor auxiliary device.
====================
bnxt_en: support header page pool in queue API
Commit 7ed816be35ab ("eth: bnxt: use page pool for head frags") added a
separate page pool for header frags. Now, frags are allocated from this
header page pool e.g. rxr->tpa_info.data.
The queue API did not properly handle rxr->tpa_info and so using the
queue API to i.e. reset any queues will result in pages being returned
to the incorrect page pool, causing inflight != 0 warnings.
Fix this bug by properly allocating/freeing tpa_info and copying/freeing
head_pool in the queue API implementation.
The 1st patch is a prep patch that refactors helpers out to be used by
the implementation patch later.
The 2nd patch is a drive-by refactor. Happy to take it out and re-send
to net-next if there are any objections.
The 3rd patch is the implementation patch that will properly alloc/free
rxr->tpa_info.
====================
David Wei [Wed, 4 Dec 2024 04:10:22 +0000 (20:10 -0800)]
bnxt_en: handle tpa_info in queue API implementation
Commit 7ed816be35ab ("eth: bnxt: use page pool for head frags") added a
page pool for header frags, which may be distinct from the existing pool
for the aggregation ring. Prior to this change, frags used in the TPA
ring rx_tpa were allocated from system memory e.g. napi_alloc_frag()
meaning their lifetimes were not associated with a page pool. They can
be returned at any time and so the queue API did not alloc or free
rx_tpa.
But now frags come from a separate head_pool which may be different to
page_pool. Without allocating and freeing rx_tpa, frags allocated from
the old head_pool may be returned to a different new head_pool which
causes a mismatch between the pp hold/release count.
Fix this problem by properly freeing and allocating rx_tpa in the queue
API implementation.
====================
net: add negotiation of in-band capabilities
This is a repost without RFC for this series, shrunk down to 13 patches
by removing the non-Marvell PCS.
Phylink's handling of in-band has been deficient for a long time, and
people keep hitting problems with it. Notably, situations with the way-
to-late standardized 2500Base-X and whether that should or should not
have in-band enabled. We have also been carrying a hack in the form of
phylink_phy_no_inband() for a PHY that has been used on a SFP module,
but has no in-band capabilities, not even for SGMII.
When phylink is trying to operate in in-band mode, this series will look
at the capabilities of the MAC-side PCS and PHY, and work out whether
in-band can or should be used, programming the PHY as appropriate. This
includes in-band bypass mode at the PHY.
We don't... yet... support bypass on the MAC side PCS, because that
requires yet more complexity.
Patch 1 passes struct phylink and struct phylink_pcs into
phylink_pcs_neg_mode() so we can look at more state in this function in
a future patch.
Patch 2 splits "cur_link_an_mode" (the MLO_AN_* mode) into two separate
purposes - a requested and an active mode. The active mode is the one
we will be using for the MAC, which becomes dependent on the result of
in-band negotiation.
Patch 3 adds debug to phylink_major_config() so we can see what is going
on with the requested and active AN modes.
Patch 4 adds to phylib a method to get the in-band capabilities of the
PHY from phylib. Patches 5 and 6 add implementations for BCM84881 and
some Marvell PHYs found on SFPs.
Patch 7 adds to phylib a method to configure the PHY in-band signalling,
and patch 8 implements it for those Marvell PHYs that support the method
in patch 4.
Patch 9 does the same as patch 4 but for the MAC-side PCS, with patches
10 and 11 adding support to Marvell NETA and PP2.
Patch 12 adds the code to phylink_pcs_neg_mode() which looks at the
capabilities, and works out whether to use in-band or out-band mode for
driving the link between the MAC PCS and PHY.
Patch 13 removes the phylink_phy_no_inband() hack now that we are
publishing the in-band capabilities from the BCM84881 PHY driver.
Three more PCS, omitted from this series due to the limit of 15 patches,
will be sent once this has been merged.
====================
Remove phylink_phy_no_inband() now that we are handling the lack of
inband negotiation by querying the capabilities of the PHY and PCS,
and the BCM84881 PHY driver provides us the information necessary to
make the decision.
net: phylink: add negotiation of in-band capabilities
Support for in-band signalling with Serdes links is uncertain. Some
PHYs do not support in-band for e.g. SGMII. Some PCS do not support
in-band for 2500Base-X. Some PCS require in-band for Base-X protocols.
Simply using what is in DT is insufficient when we have hot-pluggable
PHYs e.g. in the form of SFP modules, which may not provide the
in-band signalling.
In order to address this, we have introduced phy_inband_caps() and
pcs_inband_caps() functions to allow phylink to retrieve the
capabilities from each end of the PCS/PHY link. This commit adds code
to resolve whether in-band will be used in the various scenarios that
we have: In-band not being used, PHY present using SGMII or Base-X,
PHY not present. We also deal with no capabilties provided.
Add a pcs_inband_caps() method to query the PCS for its inband link
capabilities, and use this to determine whether link modes used with
optical SFPs can be supported.
When a PCS does not provide a method, we allow inband negotiation to
be either on or off, making this a no-op until the pcs_inband_caps()
method is implemented by a PCS driver.
projects / thirdparty / linux.git / log
