net: annotate data-races around sk->sk_lingertime

sk_getsockopt() runs locklessly. This means sk->sk_lingertime
can be read while other threads are changing its value.

Other reads also happen without socket lock being held,
and must be annotated.

Remove preprocessor logic using BITS_PER_LONG, compilers
are smart enough to figure this by themselves.

v2: fixed a clang W=1 (-Wtautological-constant-out-of-range-compare) warning
    (Jakub)

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

IPv4: add extack info for IPv4 address add/delete

Add extack info for IPv4 address add/delete, which would be useful for
users to understand the problem without having to read kernel code.

No extack message for the ifa_local checking in __inet_insert_ifa() as
it has been checked in find_matching_ifa().

Suggested-by: Ido Schimmel <idosch@idosch.org>
Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: stmmac: Check more MAC HW features for XGMAC Core 3.20

1. XGMAC Core does not have hash_filter definition, it uses
vlhash(VLAN Hash Filtering) instead, skip hash_filter when XGMAC.
2. Show exact size of Hash Table instead of raw register value.
3. Show full description of safety features defined by Synopsys Databook.
4. When safety feature is configured with no parity, or ECC only,
keep FSM Parity Checking disabled.

Signed-off-by: Furong Xu <0x1207@gmail.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'ipv6-update-route-when-delete-saddr'

Hangbin Liu says:

====================
ipv6: update route when delete source address

Currently, when remove an address, the IPv6 route will not remove the
prefer source address when the address is bond to other device. Fix this
issue and add related tests as Ido and David suggested.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

selftests: fib_test: add a test case for IPv6 source address delete

Add a test case for IPv6 source address delete.

As David suggested, add tests:
- Single device using src address
- Two devices with the same source address
- VRF with single device using src address
- VRF with two devices using src address

As Ido points out, in IPv6, the preferred source address is looked up in
the same VRF as the first nexthop device. This will give us similar results
to IPv4 if the route is installed in the same VRF as the nexthop device, but
not when the nexthop device is enslaved to a different VRF. So add tests:
- src address and nexthop dev in same VR
- src address and nexthop device in different VRF

The link local address delete logic is different from the global address.
It should only affect the associate device it bonds to. So add tests cases
for link local address testing.

Here is the test result:

IPv6 delete address route tests
    Single device using src address
    TEST: Prefsrc removed when src address removed on other device      [ OK ]
    Two devices with the same source address
    TEST: Prefsrc not removed when src address exist on other device    [ OK ]
    TEST: Prefsrc removed when src address removed on all devices       [ OK ]
    VRF with single device using src address
    TEST: Prefsrc removed when src address removed on other device      [ OK ]
    VRF with two devices using src address
    TEST: Prefsrc not removed when src address exist on other device    [ OK ]
    TEST: Prefsrc removed when src address removed on all devices       [ OK ]
    src address and nexthop dev in same VRF
    TEST: Prefsrc removed from VRF when source address deleted          [ OK ]
    TEST: Prefsrc in default VRF not removed                            [ OK ]
    TEST: Prefsrc not removed from VRF when source address exist        [ OK ]
    TEST: Prefsrc in default VRF removed                                [ OK ]
    src address and nexthop device in different VRF
    TEST: Prefsrc not removed from VRF when nexthop dev in diff VRF     [ OK ]
    TEST: Prefsrc not removed in default VRF                            [ OK ]
    TEST: Prefsrc removed from VRF when nexthop dev in diff VRF         [ OK ]
    TEST: Prefsrc removed in default VRF                                [ OK ]
    Table ID 0
    TEST: Prefsrc removed from default VRF when source address deleted  [ OK ]
    Link local source route
    TEST: Prefsrc not removed when delete ll addr from other dev        [ OK ]
    TEST: Prefsrc removed when delete ll addr                           [ OK ]
    TEST: Prefsrc not removed when delete ll addr from other dev        [ OK ]
    TEST: Prefsrc removed even ll addr still exist on other dev         [ OK ]

Tests passed:  19
Tests failed:   0

Suggested-by: Ido Schimmel <idosch@idosch.org>
Suggested-by: David Ahern <dsahern@kernel.org>
Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ipv6: do not match device when remove source route

After deleting an IPv6 address on an interface and cleaning up the
related preferred source entries, it is important to ensure that all
routes associated with the deleted address are properly cleared. The
current implementation of rt6_remove_prefsrc() only checks the preferred
source addresses bound to the current device. However, there may be
routes that are bound to other devices but still utilize the same
preferred source address.

To address this issue, it is necessary to also delete entries that are
bound to other interfaces but share the same source address with the
current device. Failure to delete these entries would leave routes that
are bound to the deleted address unclear. Here is an example reproducer
(I have omitted unrelated routes):

+ ip link add dummy1 type dummy
+ ip link add dummy2 type dummy
+ ip link set dummy1 up
+ ip link set dummy2 up
+ ip addr add 1:2:3:4::5/64 dev dummy1
+ ip route add 7:7:7:0::1 dev dummy1 src 1:2:3:4::5
+ ip route add 7:7:7:0::2 dev dummy2 src 1:2:3:4::5
+ ip -6 route show
1:2:3:4::/64 dev dummy1 proto kernel metric 256 pref medium
7:7:7::1 dev dummy1 src 1:2:3:4::5 metric 1024 pref medium
7:7:7::2 dev dummy2 src 1:2:3:4::5 metric 1024 pref medium
+ ip addr del 1:2:3:4::5/64 dev dummy1
+ ip -6 route show
7:7:7::1 dev dummy1 metric 1024 pref medium
7:7:7::2 dev dummy2 src 1:2:3:4::5 metric 1024 pref medium

As Ido reminds, in IPv6, the preferred source address is looked up in
the same VRF as the first nexthop device, which is different with IPv4.
So, while removing the device checking, we also need to add an
ipv6_chk_addr() check to make sure the address does not exist on the other
devices of the rt nexthop device's VRF.

After fix:
+ ip addr del 1:2:3:4::5/64 dev dummy1
+ ip -6 route show
7:7:7::1 dev dummy1 metric 1024 pref medium
7:7:7::2 dev dummy2 metric 1024 pref medium

Reported-by: Thomas Haller <thaller@redhat.com>
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2170513
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

selftests: vrf_route_leaking: remove ipv6_ping_frag from default testing

As the initial commit 1a01727676a8 ("selftests: Add VRF route leaking
tests") said, the IPv6 MTU test fails as source address selection
picking ::1. Every time we run the selftest this one report failed.
There seems not much meaning  to keep reporting a failure for 3 years
that no one plan to fix/update. Let't just skip this one first. We can
add it back when the issue fixed.

Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: release reference to inet6_dev pointer

addrconf_prefix_rcv returned early without releasing the inet6_dev
pointer when the PIO lifetime is less than accept_ra_min_lft.

Fixes: 5027d54a9c30 ("net: change accept_ra_min_rtr_lft to affect all RA lifetimes")
Cc: Maciej Żenczykowski <maze@google.com>
Cc: Lorenzo Colitti <lorenzo@google.com>
Cc: David Ahern <dsahern@kernel.org>
Cc: Simon Horman <horms@kernel.org>
Reviewed-by: Simon Horman <horms@kernel.org>
Reviewed-by: Maciej Żenczykowski <maze@google.com>
Signed-off-by: Patrick Rohr <prohr@google.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: selectively purge error queue in IP_RECVERR / IPV6_RECVERR

Setting IP_RECVERR and IPV6_RECVERR options to zero currently
purges the socket error queue, which was probably not expected
for zerocopy and tx_timestamp users.

I discovered this issue while preparing commit 6b5f43ea0815
("inet: move inet->recverr to inet->inet_flags"), I presume this
change does not need to be backported to stable kernels.

Add skb_errqueue_purge() helper to purge error messages only.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Willem de Bruijn <willemb@google.com>
Cc: Soheil Hassas Yeganeh <soheil@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'fixed_phy_register-return-value'

Ruan Jinjie says:

====================
net: Return PTR_ERR() for fixed_phy_register()

fixed_phy_register() returns not only -EIO or -ENODEV, but also
-EPROBE_DEFER, -EINVAL and -EBUSY. The Best practice is to return these
error codes with PTR_ERR().
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

net: lan743x: Return PTR_ERR() for fixed_phy_register()

fixed_phy_register() returns -EPROBE_DEFER, -EINVAL and -EBUSY,
etc, in addition to -EIO. The Best practice is to return these
error codes with PTR_ERR().

Signed-off-by: Ruan Jinjie <ruanjinjie@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: bcmgenet: Return PTR_ERR() for fixed_phy_register()

fixed_phy_register() returns -EPROBE_DEFER, -EINVAL and -EBUSY,
etc, in addition to -ENODEV. The Best practice is to return these
error codes with PTR_ERR().

Signed-off-by: Ruan Jinjie <ruanjinjie@huawei.com>
Acked-by: Doug Berger <opendmb@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: bgmac: Return PTR_ERR() for fixed_phy_register()

fixed_phy_register() returns -EPROBE_DEFER, -EINVAL and -EBUSY,
etc, in addition to -ENODEV. The best practice is to return
these error codes with PTR_ERR().

Signed-off-by: Ruan Jinjie <ruanjinjie@huawei.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: dsa: realtek: add phylink_get_caps implementation

The user ports use RSGMII, but we don't have that, and DT doesn't
specify a phy interface mode, so phylib defaults to GMII. These support
1G, 100M and 10M with flow control. It is unknown whether asymetric
pause is supported at all speeds.

The CPU port uses MII/GMII/RGMII/REVMII by hardware pin strapping,
and support speeds specific to each, with full duplex only supported
in some modes. Flow control may be supported again by hardware pin
strapping, and theoretically is readable through a register but no
information is given in the datasheet for that.

So, we do a best efforts - and be lenient.

Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'vcap_get_rule-return-value'

Ruan Jinjie says:

====================
net: Update and fix return value check for vcap_get_rule()

As Simon Horman suggests, update vcap_get_rule() to always
return an ERR_PTR() and update the error detection conditions to
use IS_ERR(), which would be more cleaner.

So se IS_ERR() to update the return value and fix the issue
in lan966x_ptp_add_trap().

Changes in v2:
- Update vcap_get_rule() to always return an ERR_PTR().
- Update the return value fix in lan966x_ptp_add_trap().
- Update the return value check in sparx5_tc_free_rule_resources().
====================

Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: microchip: sparx5: Update return value check for vcap_get_rule()

As Simon Horman suggests, update vcap_get_rule() to always
return an ERR_PTR() and update the error detection conditions to
use IS_ERR(), so use IS_ERR() to check the return value.

Signed-off-by: Ruan Jinjie <ruanjinjie@huawei.com>
Suggested-by: Simon Horman <horms@kernel.org>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: lan966x: Fix return value check for vcap_get_rule()

As Simon Horman suggests, update vcap_get_rule() to always
return an ERR_PTR() and update the error detection conditions to
use IS_ERR(), so use IS_ERR() to fix the return value issue.

Fixes: 72df3489fb10 ("net: lan966x: Add ptp trap rules")
Signed-off-by: Ruan Jinjie <ruanjinjie@huawei.com>
Suggested-by: Simon Horman <horms@kernel.org>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: microchip: vcap api: Always return ERR_PTR for vcap_get_rule()

As Simon Horman suggests, update vcap_get_rule() to always
return an ERR_PTR() and update the error detection conditions to
use IS_ERR(), which would be more cleaner in this case.

Signed-off-by: Ruan Jinjie <ruanjinjie@huawei.com>
Suggested-by: Simon Horman <horms@kernel.org>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: mdio: xgene: remove useless xgene_mdio_status

xgene_mdio_status is declared static, and is only written once by the
driver. It appears to have been this way since the driver was first
added to the kernel tree. No other users can be found, so let's remove
it.

Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

tools: ynl-gen: use temporary file for rendering

Currently any error during render leads to output an empty file.
That is quite annoying when using tools/net/ynl/ynl-regen.sh
which git greps files with content of "YNL-GEN.." and therefore ignores
empty files. So once you fail to regen, you have to checkout the file.

Avoid that by rendering to a temporary file first, only at the end
copy the content to the actual destination.

Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

stmmac: intel: Enable correction of MAC propagation delay

All captured timestamps should be corrected by PHY, MAC and CDC introduced
latency/errors. The CDC correction is already used. Enable MAC propagation delay
correction as well which is available since commit 26cfb838aa00 ("net: stmmac:
correct MAC propagation delay").

Before:
|ptp4l[390.458]: rms    7 max   21 freq   +177 +/-  14 delay   357 +/-   1

After:
|ptp4l[620.012]: rms    7 max   20 freq   +195 +/-  14 delay   345 +/-   1

Tested on Intel Elkhart Lake.

Signed-off-by: Kurt Kanzenbach <kurt@linutronix.de>
Reviewed-by: Johannes Zink <j.zink@pengutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: add skb_queue_purge_reason and __skb_queue_purge_reason

skb_queue_purge() and __skb_queue_purge() become wrappers
around the new generic functions.

New SKB_DROP_REASON_QUEUE_PURGE drop reason is added,
but users can start adding more specific reasons.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'smc-features'

Guangguan Wang says:

====================
net/smc: several features's implementation for smc v2.1

This patch set implement several new features in SMC v2.1(https://
www.ibm.com/support/pages/node/7009315), including vendor unique
experimental options, max connections per lgr negotiation, max links
per lgr negotiation.

v1 - v2:
 - rename field fce_v20 to fce_v2_base in struct
   smc_clc_first_contact_ext_v2x
 - use smc_get_clc_first_contact_ext in smc_connect
   _rdma_v2_prepare
 - adding comment about field vendor_oui in struct
   smc_clc_msg_smcd
 - remove comment about SMC_CONN_PER_LGR_MAX in smc_
   clc_srv_v2x_features_validate
 - rename smc_clc_clnt_v2x_features_validate

RFC v2 - v1:
 - more description in commit message
 - modify SMC_CONN_PER_LGR_xxx and SMC_LINKS_ADD_LNK_xxx
   macro defination and usage
 - rename field release_ver to release_nr
 - remove redundant release version check in client
 - explicitly set the rc value in smc_llc_cli/srv_add_link

RFC v1 - RFC v2:
 - Remove ini pointer NULL check and fix code style in
   smc_clc_send_confirm_accept.
 - Optimize the max_conns check in smc_clc_xxx_v2x_features_validate.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

net/smc: Extend SMCR v2 linkgroup netlink attribute

Add SMC_NLA_LGR_R_V2_MAX_CONNS and SMC_NLA_LGR_R_V2_MAX_LINKS
to SMCR v2 linkgroup netlink attribute SMC_NLA_LGR_R_V2 for
linkgroup's detail info showing.

Signed-off-by: Guangguan Wang <guangguan.wang@linux.alibaba.com>
Reviewed-by: Jan Karcher <jaka@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/smc: support max links per lgr negotiation in clc handshake

Support max links per lgr negotiation in clc handshake for SMCR v2.1,
which is one of smc v2.1 features. Server makes decision for the final
value of max links based on the client preferred max links and
self-preferred max links. Here use the minimum value of the client
preferred max links and server preferred max links.

Client                                       Server
     Proposal(max links(client preferred))
     -------------------------------------->

     Accept(max links(accepted value))
accepted value=min(client preferred, server preferred)
     <-------------------------------------

      Confirm(max links(accepted value))
     ------------------------------------->

Signed-off-by: Guangguan Wang <guangguan.wang@linux.alibaba.com>
Reviewed-by: Tony Lu <tonylu@linux.alibaba.com>
Reviewed-by: Jan Karcher <jaka@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/smc: support max connections per lgr negotiation

Support max connections per lgr negotiation for SMCR v2.1,
which is one of smc v2.1 features. Server makes decision for
the final value of max conns based on the client preferred
max conns and self-preferred max conns. Here use the minimum
value of client preferred max conns and server preferred max
conns.

Client                                     Server
     Proposal(max conns(client preferred))
     ------------------------------------>

     Accept(max conns(accepted value))
accepted value=min(client preferred, server preferred)
     <-----------------------------------

     Confirm(max conns(accepted value))
     ----------------------------------->

Signed-off-by: Guangguan Wang <guangguan.wang@linux.alibaba.com>
Reviewed-by: Tony Lu <tonylu@linux.alibaba.com>
Reviewed-by: Jan Karcher <jaka@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/smc: support smc v2.x features validate

Support SMC v2.x features validate for SMC v2.1. This is the frame
code for SMC v2.x features validate, and will take effects only when
the negotiated release version is v2.1 or later.

For Server, v2.x features' validation should be done in smc_clc_srv_
v2x_features_validate when receiving v2.1 or later CLC Proposal Message,
such as max conns, max links negotiation, the decision of the final
value of max conns and max links should be made in this function.
And final check for server when receiving v2.1 or later CLC Confirm
Message should be done in smc_clc_v2x_features_confirm_check.

For client, v2.x features' validation should be done in smc_clc_clnt_
v2x_features_validate when receiving v2.1 or later CLC Accept Message,
for example, the decision to accpt the accepted value or to decline
should be made in this function.

Signed-off-by: Guangguan Wang <guangguan.wang@linux.alibaba.com>
Reviewed-by: Tony Lu <tonylu@linux.alibaba.com>
Reviewed-by: Jan Karcher <jaka@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/smc: add vendor unique experimental options area in clc handshake

Add vendor unique experimental options area in clc handshake. In clc
accept and confirm msg, vendor unique experimental options use the
16-Bytes reserved field, which defined in struct smc_clc_fce_gid_ext
in previous version. Because of the struct smc_clc_first_contact_ext
is widely used and limit the scope of modification, this patch moves
the 16-Bytes reserved field out of struct smc_clc_fce_gid_ext, and
followed with the struct smc_clc_first_contact_ext in a new struct
names struct smc_clc_first_contact_ext_v2x.

For SMC-R first connection, in previous version, the struct smc_clc_
first_contact_ext and the 16-Bytes reserved field has already been
included in clc accept and confirm msg. Thus, this patch use struct
smc_clc_first_contact_ext_v2x instead of the struct smc_clc_first_
contact_ext and the 16-Bytes reserved field in SMC-R clc accept and
confirm msg is compatible with previous version.

For SMC-D first connection, in previous version, only the struct smc_
clc_first_contact_ext is included in clc accept and confirm msg, and
the 16-Bytes reserved field is not included. Thus, when the negotiated
smc release version is the version before v2.1, we still use struct
smc_clc_first_contact_ext for compatible consideration. If the negotiated
smc release version is v2.1 or later, use struct smc_clc_first_contact_
ext_v2x instead.

Signed-off-by: Guangguan Wang <guangguan.wang@linux.alibaba.com>
Reviewed-by: Tony Lu <tonylu@linux.alibaba.com>
Reviewed-by: Jan Karcher <jaka@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/smc: support smc release version negotiation in clc handshake

Support smc release version negotiation in clc handshake based on
SMC v2, where no negotiation process for different releases, but
for different versions. The latest smc release version was updated
to v2.1. And currently there are two release versions of SMCv2, v2.0
and v2.1. In the release version negotiation, client sends the preferred
release version by CLC Proposal Message, server makes decision for which
release version to use based on the client preferred release version and
self-supported release version (here choose the minimum release version
of the client preferred and server latest supported), then the decision
returns to client by CLC Accept Message. Client confirms the decision by
CLC Confirm Message.

Client                                    Server
      Proposal(preferred release version)
     ------------------------------------>

      Accept(accpeted release version)
 min(client preferred, server latest supported)
     <------------------------------------

      Confirm(accpeted release version)
     ------------------------------------>

Signed-off-by: Guangguan Wang <guangguan.wang@linux.alibaba.com>
Reviewed-by: Tony Lu <tonylu@linux.alibaba.com>
Reviewed-by: Jan Karcher <jaka@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: freescale: Remove unused declarations

Commit 5d93cfcf7360 ("net: dpaa: Convert to phylink") removed
fman_set_mac_active_pause()/fman_get_pause_cfg() but not declarations.
Commit 48257c4f168e ("Add fs_enet ethernet network driver, for several
embedded platforms.") declared but never implemented
fs_enet_platform_init() and fs_enet_platform_cleanup().

Signed-off-by: Yue Haibing <yuehaibing@huawei.com>
Reviewed-by: Sean Anderson <sean.anderson@seco.com>
Acked-by: Madalin Bucur <madalin.bucur@oss.nxp.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Link: https://lore.kernel.org/r/20230817134159.38484-1-yuehaibing@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

tcp: refine skb->ooo_okay setting

Enabling BIG TCP on a low end platform apparently increased
chances of getting flows locked on one busy TX queue.

A similar problem was handled in commit 9b462d02d6dd
("tcp: TCP Small Queues and strange attractors"),
but the strategy worked for either bulk flows,
or 'large enough' RPC. BIG TCP changed how large
RPC needed to be to enable the work around:
If RPC fits in a single skb, TSQ never triggers.

Root cause for the problem is a busy TX queue,
with delayed TX completions.

This patch changes how we set skb->ooo_okay to detect
the case TX completion was not done, but incoming ACK
already was processed and emptied rtx queue.

Update the comment to explain the tricky details.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/20230817182353.2523746-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge branch 'bnxt_en-update-for-net-next'

Michael Chan says:

====================
bnxt_en: Update for net-next

This patchset contains 2 features:

- The page pool implementation for the normal RX path (non-XDP) for
paged buffers in the aggregation ring.

- Saving of the ring error counters across reset.
====================

Link: https://lore.kernel.org/r/20230817231911.165035-1-michael.chan@broadcom.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

bnxt_en: Add tx_resets ring counter

Add a new tx_resets ring counter.  This counter will be saved as
tx_total_resets across any reset.  Since we currently do a full reset
in bnxt_sched_reset_txr(), the per ring counter will always be cleared
during reset.  Only the tx_total_resets count will be meaningful and we
only display this under ethtool -S.

Link: https://lore.kernel.org/netdev/CACKFLimD-bKmJ1tGZOLYRjWzEwxkri-Mw7iFme1x2Dr0twdCeg@mail.gmail.com/
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Link: https://lore.kernel.org/r/20230817231911.165035-7-michael.chan@broadcom.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

bnxt_en: Display the ring error counters under ethtool -S

The existing driver displays the sum of 4 ring counters under ethtool -S.
These counters are in the array bnxt_sw_func_stats.  These counters are
summed at the time of ethtool -S and will be lost when the device is reset.

Replace these counters with the new total ring error counters added in the
last patch.  These new counters are saved before reset.  ethtool -S will
now display the sum of the saved counters plus the current counters.

Link: https://lore.kernel.org/netdev/CACKFLimD-bKmJ1tGZOLYRjWzEwxkri-Mw7iFme1x2Dr0twdCeg@mail.gmail.com/
Reviewed-by: Andy Gospodarek <andrew.gospodarek@broadcom.com>
Reviewed-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: Somnath Kotur <somnath.kotur@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Link: https://lore.kernel.org/r/20230817231911.165035-6-michael.chan@broadcom.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

bnxt_en: Save ring error counters across reset

Currently, the ring counters are stored in the per ring datastructure.
During reset, all the rings are freed together with the associated
datastructures.  As a result, all the ring error counters will be reset
to zero.

Add logic to keep track of the total error counts of all the rings
and save them before reset (including ifdown).  The next patch will
display these total ring error counters under ethtool -S.

Link: https://lore.kernel.org/netdev/CACKFLimD-bKmJ1tGZOLYRjWzEwxkri-Mw7iFme1x2Dr0twdCeg@mail.gmail.com/
Reviewed-by: Ajit Khaparde <ajit.khaparde@broadcom.com>
Reviewed-by: Andy Gospodarek <andrew.gospodarek@broadcom.com>
Reviewed-by: Somnath Kotur <somnath.kotur@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Link: https://lore.kernel.org/r/20230817231911.165035-5-michael.chan@broadcom.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

bnxt_en: Increment rx_resets counter in bnxt_disable_napi()

If we are doing a complete reset with irq_re_init set to true in
bnxt_close_nic(), all the ring structures will be freed.  New
structures will be allocated in bnxt_open_nic().  The current code
increments rx_resets counter in bnxt_enable_napi() if bnapi->in_reset
is true.  In a complete reset, bnapi->in_reset will never be true
since the structure is just allocated.

Increment the rx_resets counter in bnxt_disable_napi() instead.  This
will allow us to save all the ring error counters including the
rx_resets counters in the next patch.

Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Link: https://lore.kernel.org/r/20230817231911.165035-4-michael.chan@broadcom.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

bnxt_en: Let the page pool manage the DMA mapping

Use the page pool's ability to maintain DMA mappings for us.
This avoids re-mapping of the recycled pages.

Link: https://lore.kernel.org/netdev/20230728231829.235716-4-michael.chan@broadcom.com/
Signed-off-by: Somnath Kotur <somnath.kotur@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Link: https://lore.kernel.org/r/20230817231911.165035-3-michael.chan@broadcom.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

bnxt_en: Use the unified RX page pool buffers for XDP and non-XDP

Convert to use the page pool buffers for the aggregation ring when
running in non-XDP mode.  This simplifies the driver and we benefit
from the recycling of pages.  Adjust the page pool size to account
for the aggregation ring size.

Signed-off-by: Somnath Kotur <somnath.kotur@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Link: https://lore.kernel.org/r/20230817231911.165035-2-michael.chan@broadcom.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge branch '100GbE' of git://git./linux/kernel/git/tnguy/next-queue

Tony Nguyen says:

====================
Intel Wired LAN Driver Updates 2023-08-17 (ice)

This series contains updates to ice driver only.

Jan removes unused functions and refactors code to make, possible,
functions static.

Jake rearranges some functions to be logically grouped.

Marcin removes an unnecessary call to disable VLAN stripping.

Yang Yingliang utilizes list_for_each_entry() helper for a couple list
traversals.

Przemek removes some parameters from ice_aq_alloc_free_res() which were
always the same and reworks ice_aq_wait_for_event() to reduce chance of
race.

* '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next-queue:
  ice: split ice_aq_wait_for_event() func into two
  ice: embed &ice_rq_event_info event into struct ice_aq_task
  ice: ice_aq_check_events: fix off-by-one check when filling buffer
  ice: drop two params from ice_aq_alloc_free_res()
  ice: use list_for_each_entry() helper
  ice: Remove redundant VSI configuration in eswitch setup
  ice: move E810T functions to before device agnostic ones
  ice: refactor ice_vsi_is_vlan_pruning_ena
  ice: refactor ice_ptp_hw to make functions static
  ice: refactor ice_sched to make functions static
  ice: Utilize assign_bit() helper
  ice: refactor ice_vf_lib to make functions static
  ice: refactor ice_lib to make functions static
  ice: refactor ice_ddp to make functions static
  ice: remove unused methods
====================

Link: https://lore.kernel.org/r/20230817212239.2601543-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: dm9051: Use PTR_ERR_OR_ZERO() to simplify code

Return PTR_ERR_OR_ZERO() instead of return 0 or PTR_ERR() to
simplify code.

Signed-off-by: Ruan Jinjie <ruanjinjie@huawei.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Link: https://lore.kernel.org/r/20230817022418.3588831-1-ruanjinjie@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

sky2: Remove redundant NULL check for debugfs_create_dir

Since debugfs_create_dir() returns ERR_PTR, IS_ERR() is enough to
check whether the directory is successfully created. So remove the
redundant NULL check.

Signed-off-by: Ruan Jinjie <ruanjinjie@huawei.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Link: https://lore.kernel.org/r/20230817073017.350002-1-ruanjinjie@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

pds_core: remove redundant pci_clear_master()

do_pci_disable_device() disable PCI bus-mastering as following:
static void do_pci_disable_device(struct pci_dev *dev)
{
u16 pci_command;

pci_read_config_word(dev, PCI_COMMAND, &pci_command);
if (pci_command & PCI_COMMAND_MASTER) {
pci_command &= ~PCI_COMMAND_MASTER;
pci_write_config_word(dev, PCI_COMMAND, pci_command);
}

pcibios_disable_device(dev);
}
And pci_disable_device() sets dev->is_busmaster to 0.

pci_enable_device() is called only once before calling to
pci_disable_device() and such pci_clear_master() is not needed. So remove
redundant pci_clear_master().

Also rename goto label 'err_out_clear_master' to 'err_out_disable_device'.

Signed-off-by: Yu Liao <liaoyu15@huawei.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Acked-by: Shannon Nelson <shannon.nelson@amd.com>
Link: https://lore.kernel.org/r/20230817025709.2023553-1-liaoyu15@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge branch '40GbE' of git://git./linux/kernel/git/tnguy/next-queue

Tony Nguyen says:

====================
virtchnl: fix fake 1-elem arrays

Alexander Lobakin says:

6.5-rc1 started spitting warning splats when composing virtchnl
messages, precisely on virtchnl_rss_key and virtchnl_lut:

[   84.167709] memcpy: detected field-spanning write (size 52) of single
field "vrk->key" at drivers/net/ethernet/intel/iavf/iavf_virtchnl.c:1095
(size 1)
[   84.169915] WARNING: CPU: 3 PID: 11 at drivers/net/ethernet/intel/
iavf/iavf_virtchnl.c:1095 iavf_set_rss_key+0x123/0x140 [iavf]
...
[   84.191982] Call Trace:
[   84.192439]  <TASK>
[   84.192900]  ? __warn+0xc9/0x1a0
[   84.193353]  ? iavf_set_rss_key+0x123/0x140 [iavf]
[   84.193818]  ? report_bug+0x12c/0x1b0
[   84.194266]  ? handle_bug+0x42/0x70
[   84.194714]  ? exc_invalid_op+0x1a/0x50
[   84.195149]  ? asm_exc_invalid_op+0x1a/0x20
[   84.195592]  ? iavf_set_rss_key+0x123/0x140 [iavf]
[   84.196033]  iavf_watchdog_task+0xb0c/0xe00 [iavf]
...
[   84.225476] memcpy: detected field-spanning write (size 64) of single
field "vrl->lut" at drivers/net/ethernet/intel/iavf/iavf_virtchnl.c:1127
(size 1)
[   84.227190] WARNING: CPU: 27 PID: 1044 at drivers/net/ethernet/intel/
iavf/iavf_virtchnl.c:1127 iavf_set_rss_lut+0x123/0x140 [iavf]
...
[   84.246601] Call Trace:
[   84.247228]  <TASK>
[   84.247840]  ? __warn+0xc9/0x1a0
[   84.248263]  ? iavf_set_rss_lut+0x123/0x140 [iavf]
[   84.248698]  ? report_bug+0x12c/0x1b0
[   84.249122]  ? handle_bug+0x42/0x70
[   84.249549]  ? exc_invalid_op+0x1a/0x50
[   84.249970]  ? asm_exc_invalid_op+0x1a/0x20
[   84.250390]  ? iavf_set_rss_lut+0x123/0x140 [iavf]
[   84.250820]  iavf_watchdog_task+0xb16/0xe00 [iavf]

Gustavo already tried to fix those back in 2021[0][1]. Unfortunately,
a VM can run a different kernel than the host, meaning that those
structures are sorta ABI.
However, it is possible to have proper flex arrays + struct_size()
calculations and still send the very same messages with the same sizes.
The common rule is:

elem[1] -> elem[]
size = struct_size() + <difference between the old and the new msg size>

The "old" size in the current code is calculated 3 different ways for
10 virtchnl structures total. Each commit addresses one of the ways
cumulatively instead of per-structure.

I was planning to send it to -net initially, but given that virtchnl was
renamed from i40evf and got some fat style cleanup commits in the past,
it's not very straightforward to even pick appropriate SHAs, not
speaking of automatic portability. I may send manual backports for
a couple of the latest supported kernels later on if anyone needs it
at all.

[0] https://lore.kernel.org/all/20210525230912.GA175802@embeddedor
[1] https://lore.kernel.org/all/20210525231851.GA176647@embeddedor

* '40GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next-queue:
  virtchnl: fix fake 1-elem arrays for structures allocated as `nents`
  virtchnl: fix fake 1-elem arrays in structures allocated as `nents + 1`
  virtchnl: fix fake 1-elem arrays in structs allocated as `nents + 1` - 1
====================

Link: https://lore.kernel.org/r/20230816210657.1326772-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge tag 'batadv-next-pullrequest-20230816' of git://git.open-mesh.org/linux-merge

Simon Wunderlich says:

====================
This cleanup patchset includes the following patches:

 - bump version strings, by Simon Wunderlich

 - Remove unused declarations, by Yue Haibing

 - Clean up MTU handling, by Sven Eckelmann (2 patches)

 - Clean up/remove (obsolete) functions, by Sven Eckelmann (3 patches)

* tag 'batadv-next-pullrequest-20230816' of git://git.open-mesh.org/linux-merge:
  batman-adv: Drop per algo GW section class code
  batman-adv: Keep batadv_netlink_notify_* static
  batman-adv: Drop unused function batadv_gw_bandwidth_set
  batman-adv: Check hardif MTU against runtime MTU
  batman-adv: Avoid magic value for minimum MTU
  batman-adv: Remove unused declarations
  batman-adv: Start new development cycle
====================

Link: https://lore.kernel.org/r/20230816164000.190884-1-sw@simonwunderlich.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: fec: use napi_consume_skb() in fec_enet_tx_queue()

Now that the "budget" is passed into fec_enet_tx_queue(), one
optimization we can do is to use napi_consume_skb() to instead
of dev_kfree_skb_any().

Signed-off-by: Wei Fang <wei.fang@nxp.com>
Suggested-by: Alexander H Duyck <alexander.duyck@gmail.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Link: https://lore.kernel.org/r/20230816090242.463822-1-wei.fang@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

sfc: Remove unneeded semicolon

./drivers/net/ethernet/sfc/tc_conntrack.c:464:2-3: Unneeded semicolon

Signed-off-by: Yang Li <yang.lee@linux.alibaba.com>
Acked-by: Martin Habets <habetsm.xilinx@gmail.com>
Link: https://lore.kernel.org/r/20230816004944.10841-1-yang.lee@linux.alibaba.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: use SLAB_NO_MERGE for kmem_cache skbuff_head_cache

Since v6.5-rc1 MM-tree is merged and contains a new flag SLAB_NO_MERGE
in commit d0bf7d5759c1 ("mm/slab: introduce kmem_cache flag SLAB_NO_MERGE")
now is the time to use this flag for networking as proposed
earlier see link.

The SKB (sk_buff) kmem_cache slab is critical for network performance.
Network stack uses kmem_cache_{alloc,free}_bulk APIs to gain
performance by amortising the alloc/free cost.

For the bulk API to perform efficiently the slub fragmentation need to
be low. Especially for the SLUB allocator, the efficiency of bulk free
API depend on objects belonging to the same slab (page).

When running different network performance microbenchmarks, I started
to notice that performance was reduced (slightly) when machines had
longer uptimes. I believe the cause was 'skbuff_head_cache' got
aliased/merged into the general slub for 256 bytes sized objects (with
my kernel config, without CONFIG_HARDENED_USERCOPY).

For SKB kmem_cache network stack have other various reasons for
not merging, but it varies depending on kernel config (e.g.
CONFIG_HARDENED_USERCOPY). We want to explicitly set SLAB_NO_MERGE
for this kmem_cache to get most out of kmem_cache_{alloc,free}_bulk APIs.

When CONFIG_SLUB_TINY is configured the bulk APIs are essentially
disabled. Thus, for this case drop the SLAB_NO_MERGE flag.

Link: https://lore.kernel.org/all/167396280045.539803.7540459812377220500.stgit@firesoul/
Signed-off-by: Jesper Dangaard Brouer <hawk@kernel.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Link: https://lore.kernel.org/r/169211265663.1491038.8580163757548985946.stgit@firesoul
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge git://git./linux/kernel/git/netdev/net

Cross-merge networking fixes after downstream PR.

Conflicts:

drivers/net/ethernet/sfc/tc.c
  fa165e194997 ("sfc: don't unregister flow_indr if it was never registered")
  3bf969e88ada ("sfc: add MAE table machinery for conntrack table")
https://lore.kernel.org/all/20230818112159.7430e9b4@canb.auug.org.au/

No adjacent changes.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge tag 'net-6.5-rc7' of git://git./linux/kernel/git/netdev/net

Pull networking fixes from Jakub Kicinski:
 "Including fixes from ipsec and netfilter.

  No known outstanding regressions.

  Fixes to fixes:

   - virtio-net: set queues after driver_ok, avoid a potential race
     added by recent fix

   - Revert "vlan: Fix VLAN 0 memory leak", it may lead to a warning
     when VLAN 0 is registered explicitly

   - nf_tables:
      - fix false-positive lockdep splat in recent fixes
      - don't fail inserts if duplicate has expired (fix test failures)
      - fix races between garbage collection and netns dismantle

  Current release - new code bugs:

   - mlx5: Fix mlx5_cmd_update_root_ft() error flow

  Previous releases - regressions:

   - phy: fix IRQ-based wake-on-lan over hibernate / power off

  Previous releases - always broken:

   - sock: fix misuse of sk_under_memory_pressure() preventing system
     from exiting global TCP memory pressure if a single cgroup is under
     pressure

   - fix the RTO timer retransmitting skb every 1ms if linear option is
     enabled

   - af_key: fix sadb_x_filter validation, amment netlink policy

   - ipsec: fix slab-use-after-free in decode_session6()

   - macb: in ZynqMP resume always configure PS GTR for non-wakeup
     source

  Misc:

   - netfilter: set default timeout to 3 secs for sctp shutdown send and
     recv state (from 300ms), align with protocol timers"

* tag 'net-6.5-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (49 commits)
  ice: Block switchdev mode when ADQ is active and vice versa
  qede: fix firmware halt over suspend and resume
  net: do not allow gso_size to be set to GSO_BY_FRAGS
  sock: Fix misuse of sk_under_memory_pressure()
  sfc: don't fail probe if MAE/TC setup fails
  sfc: don't unregister flow_indr if it was never registered
  net: dsa: mv88e6xxx: Wait for EEPROM done before HW reset
  net/mlx5: Fix mlx5_cmd_update_root_ft() error flow
  net/mlx5e: XDP, Fix fifo overrun on XDP_REDIRECT
  i40e: fix misleading debug logs
  iavf: fix FDIR rule fields masks validation
  ipv6: fix indentation of a config attribute
  mailmap: add entries for Simon Horman
  broadcom: b44: Use b44_writephy() return value
  net: openvswitch: reject negative ifindex
  team: Fix incorrect deletion of ETH_P_8021AD protocol vid from slaves
  net: phy: broadcom: stub c45 read/write for 54810
  netfilter: nft_dynset: disallow object maps
  netfilter: nf_tables: GC transaction race with netns dismantle
  netfilter: nf_tables: fix GC transaction races with netns and netlink event exit path
  ...

Merge tag 'drm-fixes-2023-08-18-1' of git://anongit.freedesktop.org/drm/drm

Pull drm fixes from Dave Airlie:
 "Regular enough week, mostly the usual amdgpu and i915 fixes.  Also
  qaic, nouveau, qxl and a revert for an EDID patch that had some side
  effects, along with a couple of panel fixes.

  edid:
   - revert mode parsing fix that had side effects.

  i915:
   - Fix the flow for ignoring GuC SLPC efficient frequency selection
   - Fix SDVO panel_type initialization
   - Fix display probe for IVB Q and IVB D GT2 server

  nouveau:
   - fix use-after-free in connector code

  qaic:
   - integer overflow check fix
   - fix slicing memory leak

  panel:
   - fix JDI LT070ME05000 probing
   - fix AUO G121EAN01 timings

  amdgpu:
   - SMU 13.x fixes
   - Fix mcbp parameter for gfx9
   - SMU 11.x fixes
   - Temporary fix for large numbers of XCP partitions
   - S0ix fixes
   - DCN 2.0 fix

  qxl:
   - fix use after free race in dumb object allocation"

* tag 'drm-fixes-2023-08-18-1' of git://anongit.freedesktop.org/drm/drm:
  drm/qxl: fix UAF on handle creation
  Revert "drm/edid: Fix csync detailed mode parsing"
  drm/nouveau/disp: fix use-after-free in error handling of nouveau_connector_create
  Revert "Revert "drm/amdgpu/display: change pipe policy for DCN 2.0""
  drm/amd: flush any delayed gfxoff on suspend entry
  drm/amdgpu: skip fence GFX interrupts disable/enable for S0ix
  drm/amdgpu: skip xcp drm device allocation when out of drm resource
  drm/amd/pm: Update pci link width for smu v13.0.6
  drm/amd/pm: Fix temperature unit of SMU v13.0.6
  drm/amdgpu/pm: fix throttle_status for other than MP1 11.0.7
  drm/amdgpu: disable mcbp if parameter zero is set
  drm/amd/pm: disallow the fan setting if there is no fan on smu 13.0.0
  accel/qaic: Clean up integer overflow checking in map_user_pages()
  accel/qaic: Fix slicing memory leak
  drm/i915: fix display probe for IVB Q and IVB D GT2 server
  drm/i915/sdvo: fix panel_type initialization
  drm/i915/guc/slpc: Restore efficient freq earlier
  drm/panel: simple: Fix AUO G121EAN01 panel timings according to the docs
  drm/panel: JDI LT070ME05000 simplify with dev_err_probe()

Merge branch 'netconsole-enable-compile-time-configuration'

Breno Leitao says:

====================
netconsole: Enable compile time configuration

Enable netconsole features to be set at compilation time. Create two
Kconfig options that allow users to set extended logs and release
prepending features at compilation time.

The first patch de-duplicates the initialization code, and the second
patch adds the support in the de-duplicated code, avoiding touching two
different functions with the same change.
====================

Link: https://lore.kernel.org/r/20230811093158.1678322-1-leitao@debian.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

netconsole: Enable compile time configuration

Enable netconsole features to be set at compilation time. Create two
Kconfig options that allow users to set extended logs and release
prepending features at compilation time.

Right now, the user needs to pass command line parameters to netconsole,
such as "+"/"r" to enable extended logs and version prepending features.

With these two options, the user could set the default values for the
features at compile time, and don't need to pass it in the command line
to get them enabled, simplifying the command line.

Signed-off-by: Breno Leitao <leitao@debian.org>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/20230811093158.1678322-3-leitao@debian.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

netconsole: Create a allocation helper

De-duplicate the initialization and allocation code for struct
netconsole_target.

The same allocation and initialization code is duplicated in two
different places in the netconsole subsystem, when the netconsole target
is initialized by command line parameters (alloc_param_target()), and
dynamically by sysfs (make_netconsole_target()).

Create a helper function, and call it from the two different functions.

Suggested-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Breno Leitao <leitao@debian.org>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/20230811093158.1678322-2-leitao@debian.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: mdio: fix -Wvoid-pointer-to-enum-cast warning

When building with clang 18 I see the following warning:
|       drivers/net/mdio/mdio-xgene.c:338:13: warning: cast to smaller integer
|               type 'enum xgene_mdio_id' from 'const void *' [-Wvoid-pointer-to-enum-cast]
|         338 |                 mdio_id = (enum xgene_mdio_id)of_id->data;

This is due to the fact that `of_id->data` is a void* while `enum
xgene_mdio_id` has the size of an int. This leads to truncation and
possible data loss.

Link: https://github.com/ClangBuiltLinux/linux/issues/1910
Reported-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Justin Stitt <justinstitt@google.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Link: https://lore.kernel.org/r/20230815-void-drivers-net-mdio-mdio-xgene-v1-1-5304342e0659@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge branch 'netem-use-a-seeded-prng-for-loss-and-corruption-events'

François Michel says:

====================
netem: use a seeded PRNG for loss and corruption events

In order to reproduce bugs or performance evaluation of
network protocols and applications, it is useful to have
reproducible test suites and tools. This patch adds
a way to specify a PRNG seed through the
TCA_NETEM_PRNG_SEED attribute for generating netem
loss and corruption events. Initializing the qdisc
with the same seed leads to the exact same loss
and corruption patterns. If no seed is explicitly
specified, the qdisc generates a random seed using
get_random_u64().

This patch can be and has been tested using tc from
the following iproute2-next fork:
https://github.com/francoismichel/iproute2-next

For instance, setting the seed 42424242 on the loopback
with a loss rate of 10% will systematically drop the 5th,
12th and 24th packet when sending 25 packets.
====================

Link: https://lore.kernel.org/r/20230815092348.1449179-1-francois.michel@uclouvain.be
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

netem: use seeded PRNG for correlated loss events

Use prandom_u32_state() instead of get_random_u32() to generate
the correlated loss events of netem.

Signed-off-by: François Michel <francois.michel@uclouvain.be>
Reviewed-by: Simon Horman <horms@kernel.org>
Acked-by: Stephen Hemminger <stephen@networkplumber.org>
Link: https://lore.kernel.org/r/20230815092348.1449179-4-francois.michel@uclouvain.be
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

netem: use a seeded PRNG for generating random losses

Use prandom_u32_state() instead of get_random_u32() to generate
the random loss events of netem. The state of the prng is part
of the prng attribute of struct netem_sched_data.

Signed-off-by: François Michel <francois.michel@uclouvain.be>
Reviewed-by: Simon Horman <horms@kernel.org>
Acked-by: Stephen Hemminger <stephen@networkplumber.org>
Link: https://lore.kernel.org/r/20230815092348.1449179-3-francois.michel@uclouvain.be
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

netem: add prng attribute to netem_sched_data

Add prng attribute to struct netem_sched_data and
allows setting the seed of the PRNG through netlink
using the new TCA_NETEM_PRNG_SEED attribute.
The PRNG attribute is not actually used yet.

Signed-off-by: François Michel <francois.michel@uclouvain.be>
Reviewed-by: Simon Horman <horms@kernel.org>
Acked-by: Stephen Hemminger <stephen@networkplumber.org>
Link: https://lore.kernel.org/r/20230815092348.1449179-2-francois.michel@uclouvain.be
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: ena: Use pci_dev_id() to simplify the code

PCI core API pci_dev_id() can be used to get the BDF number for a pci
device. We don't need to compose it manually. Use pci_dev_id() to
simplify the code a little bit.

Signed-off-by: Jialin Zhang <zhangjialin11@huawei.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Reviewed-by: Shay Agroskin <shayagr@amazon.com>
Link: https://lore.kernel.org/r/20230815024248.3519068-1-zhangjialin11@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

tun: add __exit annotations to module exit func tun_cleanup()

Add missing __exit annotations to module exit func tun_cleanup().

Signed-off-by: Ziyang Xuan <william.xuanziyang@huawei.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Link: https://lore.kernel.org/r/20230814083000.3893589-1-william.xuanziyang@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge branch '40GbE' of git://git./linux/kernel/git/tnguy/net-queue

Tony Nguyen says:

====================
Intel Wired LAN Driver Updates 2023-08-16 (iavf, i40e)

This series contains updates to iavf and i40e drivers.

Piotr adds checks for unsupported Flow Director rules on iavf.

Andrii replaces incorrect 'write' messaging on read operations for i40e.

* '40GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue:
  i40e: fix misleading debug logs
  iavf: fix FDIR rule fields masks validation
====================

Link: https://lore.kernel.org/r/20230816193308.1307535-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

ice: split ice_aq_wait_for_event() func into two

Mitigate race between registering on wait list and receiving
AQ Response from FW.

ice_aq_prep_for_event() should be called before sending AQ command,
ice_aq_wait_for_event() should be called after sending AQ command,
to wait for AQ Response.

Please note, that this was found by reading the code,
an actual race has not yet materialized.

Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Signed-off-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Tested-by: Pucha Himasekhar Reddy <himasekharx.reddy.pucha@intel.com> (A Contingent worker at Intel)
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>

ice: embed &ice_rq_event_info event into struct ice_aq_task

Expose struct ice_aq_task to callers,
what takes burden of memory ownership out from AQ-wait family of functions,
and reduces need for heap-based allocations.

Embed struct ice_rq_event_info event into struct ice_aq_task
(instead of it being a ptr) to remove some more code from the callers.

Subsequent commit will improve more based on this one.

Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Signed-off-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Tested-by: Pucha Himasekhar Reddy <himasekharx.reddy.pucha@intel.com> (A Contingent worker at Intel)
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>

ice: ice_aq_check_events: fix off-by-one check when filling buffer

Allow task's event buffer to be filled also in the case that it's size
is exactly the size of the message.

Fixes: d69ea414c9b4 ("ice: implement device flash update via devlink")
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Signed-off-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Tested-by: Pucha Himasekhar Reddy <himasekharx.reddy.pucha@intel.com> (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>

ice: drop two params from ice_aq_alloc_free_res()

Drop @num_entries and @cd params, latter of which was always NULL.

Number of entities to alloc is passed in internal buffer, the outer layer
(that @num_entries was assigned to) meaning is closer to "the number of
requests", which was =1 in all cases.
ice_free_hw_res() was always called with 1 as its @num arg.

Signed-off-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Tested-by: Pucha Himasekhar Reddy <himasekharx.reddy.pucha@intel.com> (A Contingent worker at Intel)
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>

ice: use list_for_each_entry() helper

Convert list_for_each() to list_for_each_entry() where applicable.
No functional changed.

Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Tested-by: Sujai Buvaneswaran <sujai.buvaneswaran@intel.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>

ice: Remove redundant VSI configuration in eswitch setup

Remove a call to disable VLAN stripping on switchdev control plane VSI, as
it is disabled by default.

Signed-off-by: Marcin Szycik <marcin.szycik@linux.intel.com>
Tested-by: Sujai Buvaneswaran <sujai.buvaneswaran@intel.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>

ice: move E810T functions to before device agnostic ones

Commit 885fe6932a11 ("ice: Add support for SMA control multiplexer")
accidentally placed all of the E810T SMA control functions in the middle of
the device agnostic functions section of ice_ptp_hw.c

This works fine, but makes it harder for readers to follow. The
ice_ptp_hw.c file is laid out such that each hardware family has the
specific functions in one block, with the access functions placed at the
end of the file.

Move the E810T functions so that they are in a block just after the E810
functions. Also move the ice_get_phy_tx_tstamp_ready_e810 which got added
at the end of the E810T block.

This keeps the functions laid out in a logical order and avoids intermixing
the generic access functions with the device specific implementations.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Tested-by: Arpana Arland <arpanax.arland@intel.com> (A Contingent worker at Intel)
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>

ice: refactor ice_vsi_is_vlan_pruning_ena

As this method became static, and is already called
with check for vsi being non-null, an unnecessary check along
with superfluous parentheses is removed.

Signed-off-by: Jan Sokolowski <jan.sokolowski@intel.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>

ice: refactor ice_ptp_hw to make functions static

As following methods are not used outside ice_ptp_hw,
they can be made static:
ice_read_phy_reg_e822
ice_write_phy_reg_e822
ice_ptp_prep_port_adj_e822

Signed-off-by: Jan Sokolowski <jan.sokolowski@intel.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>

ice: refactor ice_sched to make functions static

As ice_sched_set_node_bw_lmt_per_tc is not used
outside of ice_sched, it can be made static.

Signed-off-by: Jan Sokolowski <jan.sokolowski@intel.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>

ice: Utilize assign_bit() helper

The if/else check for bit setting can be replaced by using the
assign_bit() helper so do so.

Suggested-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>

drm/qxl: fix UAF on handle creation

qxl_mode_dumb_create() dereferences the qobj returned by
qxl_gem_object_create_with_handle(), but the handle is the only one
holding a reference to it.

A potential attacker could guess the returned handle value and closes it
between the return of qxl_gem_object_create_with_handle() and the qobj
usage, triggering a use-after-free scenario.

Reproducer:

int dri_fd =-1;
struct drm_mode_create_dumb arg = {0};

void gem_close(int handle);

void* trigger(void* ptr)
{
int ret;
arg.width = arg.height = 0x20;
arg.bpp = 32;
ret = ioctl(dri_fd, DRM_IOCTL_MODE_CREATE_DUMB, &arg);
if(ret)
{
perror("[*] DRM_IOCTL_MODE_CREATE_DUMB Failed");
exit(-1);
}
gem_close(arg.handle);
while(1) {
struct drm_mode_create_dumb args = {0};
args.width = args.height = 0x20;
args.bpp = 32;
ret = ioctl(dri_fd, DRM_IOCTL_MODE_CREATE_DUMB, &args);
if (ret) {
perror("[*] DRM_IOCTL_MODE_CREATE_DUMB Failed");
exit(-1);
}

printf("[*] DRM_IOCTL_MODE_CREATE_DUMB created, %d\n", args.handle);
gem_close(args.handle);
}
return NULL;
}

void gem_close(int handle)
{
struct drm_gem_close args;
args.handle = handle;
int ret = ioctl(dri_fd, DRM_IOCTL_GEM_CLOSE, &args); // gem close handle
if (!ret)
printf("gem close handle %d\n", args.handle);
}

int main(void)
{
dri_fd= open("/dev/dri/card0", O_RDWR);
printf("fd:%d\n", dri_fd);

if(dri_fd == -1)
return -1;

pthread_t tid1;

if(pthread_create(&tid1,NULL,trigger,NULL)){
perror("[*] thread_create tid1\n");
return -1;
}
while (1)
{
gem_close(arg.handle);
}
return 0;
}

This is a KASAN report:

==================================================================
BUG: KASAN: slab-use-after-free in qxl_mode_dumb_create+0x3c2/0x400 linux/drivers/gpu/drm/qxl/qxl_dumb.c:69
Write of size 1 at addr ffff88801136c240 by task poc/515

CPU: 1 PID: 515 Comm: poc Not tainted 6.3.0 #3
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.0-debian-1.16.0-4 04/01/2014
Call Trace:
<TASK>
__dump_stack linux/lib/dump_stack.c:88
dump_stack_lvl+0x48/0x70 linux/lib/dump_stack.c:106
print_address_description linux/mm/kasan/report.c:319
print_report+0xd2/0x660 linux/mm/kasan/report.c:430
kasan_report+0xd2/0x110 linux/mm/kasan/report.c:536
__asan_report_store1_noabort+0x17/0x30 linux/mm/kasan/report_generic.c:383
qxl_mode_dumb_create+0x3c2/0x400 linux/drivers/gpu/drm/qxl/qxl_dumb.c:69
drm_mode_create_dumb linux/drivers/gpu/drm/drm_dumb_buffers.c:96
drm_mode_create_dumb_ioctl+0x1f5/0x2d0 linux/drivers/gpu/drm/drm_dumb_buffers.c:102
drm_ioctl_kernel+0x21d/0x430 linux/drivers/gpu/drm/drm_ioctl.c:788
drm_ioctl+0x56f/0xcc0 linux/drivers/gpu/drm/drm_ioctl.c:891
vfs_ioctl linux/fs/ioctl.c:51
__do_sys_ioctl linux/fs/ioctl.c:870
__se_sys_ioctl linux/fs/ioctl.c:856
__x64_sys_ioctl+0x13d/0x1c0 linux/fs/ioctl.c:856
do_syscall_x64 linux/arch/x86/entry/common.c:50
do_syscall_64+0x5b/0x90 linux/arch/x86/entry/common.c:80
entry_SYSCALL_64_after_hwframe+0x72/0xdc linux/arch/x86/entry/entry_64.S:120
RIP: 0033:0x7ff5004ff5f7
Code: 00 00 00 48 8b 05 99 c8 0d 00 64 c7 00 26 00 00 00 48 c7 c0 ff ff ff ff c3 66 2e 0f 1f 84 00 00 00 00 00 b8 10 00 00 00 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d 69 c8 0d 00 f7 d8 64 89 01 48

RSP: 002b:00007ff500408ea8 EFLAGS: 00000286 ORIG_RAX: 0000000000000010
RAX: ffffffffffffffda RBX: 0000000000000000 RCX: 00007ff5004ff5f7
RDX: 00007ff500408ec0 RSI: 00000000c02064b2 RDI: 0000000000000003
RBP: 00007ff500408ef0 R08: 0000000000000000 R09: 000000000000002a
R10: 0000000000000000 R11: 0000000000000286 R12: 00007fff1c6cdafe
R13: 00007fff1c6cdaff R14: 00007ff500408fc0 R15: 0000000000802000
</TASK>

Allocated by task 515:
kasan_save_stack+0x38/0x70 linux/mm/kasan/common.c:45
kasan_set_track+0x25/0x40 linux/mm/kasan/common.c:52
kasan_save_alloc_info+0x1e/0x40 linux/mm/kasan/generic.c:510
____kasan_kmalloc linux/mm/kasan/common.c:374
__kasan_kmalloc+0xc3/0xd0 linux/mm/kasan/common.c:383
kasan_kmalloc linux/./include/linux/kasan.h:196
kmalloc_trace+0x48/0xc0 linux/mm/slab_common.c:1066
kmalloc linux/./include/linux/slab.h:580
kzalloc linux/./include/linux/slab.h:720
qxl_bo_create+0x11a/0x610 linux/drivers/gpu/drm/qxl/qxl_object.c:124
qxl_gem_object_create+0xd9/0x360 linux/drivers/gpu/drm/qxl/qxl_gem.c:58
qxl_gem_object_create_with_handle+0xa1/0x180 linux/drivers/gpu/drm/qxl/qxl_gem.c:89
qxl_mode_dumb_create+0x1cd/0x400 linux/drivers/gpu/drm/qxl/qxl_dumb.c:63
drm_mode_create_dumb linux/drivers/gpu/drm/drm_dumb_buffers.c:96
drm_mode_create_dumb_ioctl+0x1f5/0x2d0 linux/drivers/gpu/drm/drm_dumb_buffers.c:102
drm_ioctl_kernel+0x21d/0x430 linux/drivers/gpu/drm/drm_ioctl.c:788
drm_ioctl+0x56f/0xcc0 linux/drivers/gpu/drm/drm_ioctl.c:891
vfs_ioctl linux/fs/ioctl.c:51
__do_sys_ioctl linux/fs/ioctl.c:870
__se_sys_ioctl linux/fs/ioctl.c:856
__x64_sys_ioctl+0x13d/0x1c0 linux/fs/ioctl.c:856
do_syscall_x64 linux/arch/x86/entry/common.c:50
do_syscall_64+0x5b/0x90 linux/arch/x86/entry/common.c:80
entry_SYSCALL_64_after_hwframe+0x72/0xdc linux/arch/x86/entry/entry_64.S:120

Freed by task 515:
kasan_save_stack+0x38/0x70 linux/mm/kasan/common.c:45
kasan_set_track+0x25/0x40 linux/mm/kasan/common.c:52
kasan_save_free_info+0x2e/0x60 linux/mm/kasan/generic.c:521
____kasan_slab_free linux/mm/kasan/common.c:236
____kasan_slab_free+0x180/0x1f0 linux/mm/kasan/common.c:200
__kasan_slab_free+0x12/0x30 linux/mm/kasan/common.c:244
kasan_slab_free linux/./include/linux/kasan.h:162
slab_free_hook linux/mm/slub.c:1781
slab_free_freelist_hook+0xd2/0x1a0 linux/mm/slub.c:1807
slab_free linux/mm/slub.c:3787
__kmem_cache_free+0x196/0x2d0 linux/mm/slub.c:3800
kfree+0x78/0x120 linux/mm/slab_common.c:1019
qxl_ttm_bo_destroy+0x140/0x1a0 linux/drivers/gpu/drm/qxl/qxl_object.c:49
ttm_bo_release+0x678/0xa30 linux/drivers/gpu/drm/ttm/ttm_bo.c:381
kref_put linux/./include/linux/kref.h:65
ttm_bo_put+0x50/0x80 linux/drivers/gpu/drm/ttm/ttm_bo.c:393
qxl_gem_object_free+0x3e/0x60 linux/drivers/gpu/drm/qxl/qxl_gem.c:42
drm_gem_object_free+0x5c/0x90 linux/drivers/gpu/drm/drm_gem.c:974
kref_put linux/./include/linux/kref.h:65
__drm_gem_object_put linux/./include/drm/drm_gem.h:431
drm_gem_object_put linux/./include/drm/drm_gem.h:444
qxl_gem_object_create_with_handle+0x151/0x180 linux/drivers/gpu/drm/qxl/qxl_gem.c:100
qxl_mode_dumb_create+0x1cd/0x400 linux/drivers/gpu/drm/qxl/qxl_dumb.c:63
drm_mode_create_dumb linux/drivers/gpu/drm/drm_dumb_buffers.c:96
drm_mode_create_dumb_ioctl+0x1f5/0x2d0 linux/drivers/gpu/drm/drm_dumb_buffers.c:102
drm_ioctl_kernel+0x21d/0x430 linux/drivers/gpu/drm/drm_ioctl.c:788
drm_ioctl+0x56f/0xcc0 linux/drivers/gpu/drm/drm_ioctl.c:891
vfs_ioctl linux/fs/ioctl.c:51
__do_sys_ioctl linux/fs/ioctl.c:870
__se_sys_ioctl linux/fs/ioctl.c:856
__x64_sys_ioctl+0x13d/0x1c0 linux/fs/ioctl.c:856
do_syscall_x64 linux/arch/x86/entry/common.c:50
do_syscall_64+0x5b/0x90 linux/arch/x86/entry/common.c:80
entry_SYSCALL_64_after_hwframe+0x72/0xdc linux/arch/x86/entry/entry_64.S:120

The buggy address belongs to the object at ffff88801136c000
which belongs to the cache kmalloc-1k of size 1024
The buggy address is located 576 bytes inside of
freed 1024-byte region [ffff88801136c000, ffff88801136c400)

The buggy address belongs to the physical page:
page:0000000089fc329b refcount:1 mapcount:0 mapping:0000000000000000 index:0x0 pfn:0x11368
head:0000000089fc329b order:3 entire_mapcount:0 nr_pages_mapped:0 pincount:0
flags: 0xfffffc0010200(slab|head|node=0|zone=1|lastcpupid=0x1fffff)
raw: 000fffffc0010200 ffff888007841dc0 dead000000000122 0000000000000000
raw: 0000000000000000 0000000080100010 00000001ffffffff 0000000000000000
page dumped because: kasan: bad access detected

Memory state around the buggy address:
ffff88801136c100: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
ffff88801136c180: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
>ffff88801136c200: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
^
ffff88801136c280: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
ffff88801136c300: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
==================================================================
Disabling lock debugging due to kernel taint

Instead of returning a weak reference to the qxl_bo object, return the
created drm_gem_object and let the caller decrement the reference count
when it no longer needs it. As a convenience, if the caller is not
interested in the gobj object, it can pass NULL to the parameter and the
reference counting is descremented internally.

The bug and the reproducer were originally found by the Zero Day Initiative project (ZDI-CAN-20940).

Link: https://www.zerodayinitiative.com/
Signed-off-by: Wander Lairson Costa <wander@redhat.com>
Cc: stable@vger.kernel.org
Reviewed-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230814165119.90847-1-wander@redhat.com

Merge tag 'amd-drm-fixes-6.5-2023-08-16' of https://gitlab.freedesktop.org/agd5f/linux into drm-fixes

amd-drm-fixes-6.5-2023-08-16:

amdgpu:
- SMU 13.x fixes
- Fix mcbp parameter for gfx9
- SMU 11.x fixes
- Temporary fix for large numbers of XCP partitions
- S0ix fixes
- DCN 2.0 fix

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Alex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230816200226.10771-1-alexander.deucher@amd.com

Merge tag 'drm-misc-fixes-2023-08-17' of git://anongit.freedesktop.org/drm/drm-misc into drm-fixes

One EPROBE_DEFER handling fix for the JDI LT070ME05000, a timing fix for
the AUO G121EAN01 panel, an integer overflow and a memory leak fixes for
the qaic accel, a use-after-free fix for nouveau and a revert for an
alleged fix in EDID parsing.

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Maxime Ripard <mripard@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/3olqt33em5uhxzjbqghwcwnvmw73h7bxkbdxookmnkecymd4vc@7ogm6gewpprq

Merge tag 'drm-intel-fixes-2023-08-17' of git://anongit.freedesktop.org/drm/drm-intel into drm-fixes

- Fix the flow for ignoring GuC SLPC efficient frequency selection (Vinay)
- Fix SDVO panel_type initialization (Jani)
- Fix display probe for IVB Q and IVB D GT2 server (Jani)

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/ZN4yduyBU1Ev9dc7@intel.com

Merge tag 'mlx5-fixes-2023-08-16' of git://git./linux/kernel/git/saeed/linux

Saeed Mahameed says:

====================
mlx5 fixes 2023-08-16

This series provides bug fixes to mlx5 driver.

* tag 'mlx5-fixes-2023-08-16' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux:
  net/mlx5: Fix mlx5_cmd_update_root_ft() error flow
  net/mlx5e: XDP, Fix fifo overrun on XDP_REDIRECT
====================

Link: https://lore.kernel.org/r/20230816204108.53819-1-saeed@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

ice: Block switchdev mode when ADQ is active and vice versa

ADQ and switchdev are not supported simultaneously. Enabling both at the
same time can result in nullptr dereference.

To prevent this, check if ADQ is active when changing devlink mode to
switchdev mode, and check if switchdev is active when enabling ADQ.

Fixes: fbc7b27af0f9 ("ice: enable ndo_setup_tc support for mqprio_qdisc")
Signed-off-by: Marcin Szycik <marcin.szycik@linux.intel.com>
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Tested-by: Sujai Buvaneswaran <sujai.buvaneswaran@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/20230816193405.1307580-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

qede: fix firmware halt over suspend and resume

While performing certain power-off sequences, PCI drivers are
called to suspend and resume their underlying devices through
PCI PM (power management) interface. However this NIC hardware
does not support PCI PM suspend/resume operations so system wide
suspend/resume leads to bad MFW (management firmware) state which
causes various follow-up errors in driver when communicating with
the device/firmware afterwards.

To fix this driver implements PCI PM suspend handler to indicate
unsupported operation to the PCI subsystem explicitly, thus avoiding
system to go into suspended/standby mode.

Without this fix device/firmware does not recover unless system
is power cycled.

Fixes: 2950219d87b0 ("qede: Add basic network device support")
Signed-off-by: Manish Chopra <manishc@marvell.com>
Signed-off-by: Alok Prasad <palok@marvell.com>
Reviewed-by: John Meneghini <jmeneghi@redhat.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/20230816150711.59035-1-manishc@marvell.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: do not allow gso_size to be set to GSO_BY_FRAGS

One missing check in virtio_net_hdr_to_skb() allowed
syzbot to crash kernels again [1]

Do not allow gso_size to be set to GSO_BY_FRAGS (0xffff),
because this magic value is used by the kernel.

[1]
general protection fault, probably for non-canonical address 0xdffffc000000000e: 0000 [#1] PREEMPT SMP KASAN
KASAN: null-ptr-deref in range [0x0000000000000070-0x0000000000000077]
CPU: 0 PID: 5039 Comm: syz-executor401 Not tainted 6.5.0-rc5-next-20230809-syzkaller #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 07/26/2023
RIP: 0010:skb_segment+0x1a52/0x3ef0 net/core/skbuff.c:4500
Code: 00 00 00 e9 ab eb ff ff e8 6b 96 5d f9 48 8b 84 24 00 01 00 00 48 8d 78 70 48 b8 00 00 00 00 00 fc ff df 48 89 fa 48 c1 ea 03 <0f> b6 04 02 84 c0 74 08 3c 03 0f 8e ea 21 00 00 48 8b 84 24 00 01
RSP: 0018:ffffc90003d3f1c8 EFLAGS: 00010202
RAX: dffffc0000000000 RBX: 000000000001fffe RCX: 0000000000000000
RDX: 000000000000000e RSI: ffffffff882a3115 RDI: 0000000000000070
RBP: ffffc90003d3f378 R08: 0000000000000005 R09: 000000000000ffff
R10: 000000000000ffff R11: 5ee4a93e456187d6 R12: 000000000001ffc6
R13: dffffc0000000000 R14: 0000000000000008 R15: 000000000000ffff
FS: 00005555563f2380(0000) GS:ffff8880b9800000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000000020020000 CR3: 000000001626d000 CR4: 00000000003506f0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
Call Trace:
<TASK>
udp6_ufo_fragment+0x9d2/0xd50 net/ipv6/udp_offload.c:109
ipv6_gso_segment+0x5c4/0x17b0 net/ipv6/ip6_offload.c:120
skb_mac_gso_segment+0x292/0x610 net/core/gso.c:53
__skb_gso_segment+0x339/0x710 net/core/gso.c:124
skb_gso_segment include/net/gso.h:83 [inline]
validate_xmit_skb+0x3a5/0xf10 net/core/dev.c:3625
__dev_queue_xmit+0x8f0/0x3d60 net/core/dev.c:4329
dev_queue_xmit include/linux/netdevice.h:3082 [inline]
packet_xmit+0x257/0x380 net/packet/af_packet.c:276
packet_snd net/packet/af_packet.c:3087 [inline]
packet_sendmsg+0x24c7/0x5570 net/packet/af_packet.c:3119
sock_sendmsg_nosec net/socket.c:727 [inline]
sock_sendmsg+0xd9/0x180 net/socket.c:750
____sys_sendmsg+0x6ac/0x940 net/socket.c:2496
___sys_sendmsg+0x135/0x1d0 net/socket.c:2550
__sys_sendmsg+0x117/0x1e0 net/socket.c:2579
do_syscall_x64 arch/x86/entry/common.c:50 [inline]
do_syscall_64+0x38/0xb0 arch/x86/entry/common.c:80
entry_SYSCALL_64_after_hwframe+0x63/0xcd
RIP: 0033:0x7ff27cdb34d9

Fixes: 3953c46c3ac7 ("sk_buff: allow segmenting based on frag sizes")
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Xin Long <lucien.xin@gmail.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Jason Wang <jasowang@redhat.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Reviewed-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Reviewed-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Link: https://lore.kernel.org/r/20230816142158.1779798-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

sock: Fix misuse of sk_under_memory_pressure()

The status of global socket memory pressure is updated when:

  a) __sk_mem_raise_allocated():

enter: sk_memory_allocated(sk) >  sysctl_mem[1]
leave: sk_memory_allocated(sk) <= sysctl_mem[0]

  b) __sk_mem_reduce_allocated():

leave: sk_under_memory_pressure(sk) &&
sk_memory_allocated(sk) < sysctl_mem[0]

So the conditions of leaving global pressure are inconstant, which
may lead to the situation that one pressured net-memcg prevents the
global pressure from being cleared when there is indeed no global
pressure, thus the global constrains are still in effect unexpectedly
on the other sockets.

This patch fixes this by ignoring the net-memcg's pressure when
deciding whether should leave global memory pressure.

Fixes: e1aab161e013 ("socket: initial cgroup code.")
Signed-off-by: Abel Wu <wuyun.abel@bytedance.com>
Acked-by: Shakeel Butt <shakeelb@google.com>
Link: https://lore.kernel.org/r/20230816091226.1542-1-wuyun.abel@bytedance.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

sfc: don't fail probe if MAE/TC setup fails

Existing comment in the source explains why we don't want efx_init_tc()
 failure to be fatal.  Cited commit erroneously consolidated failure
 paths causing the probe to be failed in this case.

Fixes: 7e056e2360d9 ("sfc: obtain device mac address based on firmware handle for ef100")
Reviewed-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
Link: https://lore.kernel.org/r/aa7f589dd6028bd1ad49f0a85f37ab33c09b2b45.1692114888.git.ecree.xilinx@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

sfc: don't unregister flow_indr if it was never registered

In efx_init_tc(), move the setting of efx->tc->up after the
 flow_indr_dev_register() call, so that if it fails, efx_fini_tc()
 won't call flow_indr_dev_unregister().

Fixes: 5b2e12d51bd8 ("sfc: bind indirect blocks for TC offload on EF100")
Suggested-by: Pieter Jansen van Vuuren <pieter.jansen-van-vuuren@amd.com>
Reviewed-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
Link: https://lore.kernel.org/r/a81284d7013aba74005277bd81104e4cfbea3f6f.1692114888.git.ecree.xilinx@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

ice: refactor ice_vf_lib to make functions static

As following methods are not used outside ice_vf_lib,
they can be made static:
ice_vf_rebuild_host_vlan_cfg
ice_vf_rebuild_host_tx_rate_cfg
ice_vf_set_host_trust_cfg
ice_vf_rebuild_host_mac_cfg
ice_vf_rebuild_aggregator_node_cfg
ice_vf_rebuild_host_cfg
ice_set_vf_state_qs_dis
ice_vf_set_initialized

In order to achieve that, the order in which these
were defined was reorganized.

Signed-off-by: Jan Sokolowski <jan.sokolowski@intel.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>

ice: refactor ice_lib to make functions static

As following methods are not used outside of ice_lib,
they can be made static:
ice_vsi_is_vlan_pruning_ena
ice_vsi_cfg_frame_size

Signed-off-by: Jan Sokolowski <jan.sokolowski@intel.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>

ice: refactor ice_ddp to make functions static

As following methods are not used outside of ice_ddp,
they can be made static:
ice_verify_pgk
ice_pkg_val_buf
ice_aq_download_pkg
ice_aq_update_pkg
ice_find_seg_in_pkg

Signed-off-by: Jan Sokolowski <jan.sokolowski@intel.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>

ice: remove unused methods

Following methods were found to no longer be in use:
ice_is_pca9575_present
ice_mac_fltr_exist
ice_napi_del

Remove them.

Signed-off-by: Jan Sokolowski <jan.sokolowski@intel.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>

Merge tag 'nfsd-6.5-4' of git://git./linux/kernel/git/cel/linux

Pull nfsd fix from Chuck Lever:

 - Fix new MSG_SPLICE_PAGES support in server's TCP sendmsg helper

* tag 'nfsd-6.5-4' of git://git.kernel.org/pub/scm/linux/kernel/git/cel/linux:
  sunrpc: set the bv_offset of first bvec in svc_tcp_sendmsg

Revert "drm/edid: Fix csync detailed mode parsing"

This reverts commit ca62297b2085b5b3168bd891ca24862242c635a1.

Commit ca62297b2085 ("drm/edid: Fix csync detailed mode parsing") fixed
EDID detailed mode sync parsing. Unfortunately, there are quite a few
displays out there that have bogus (zero) sync field that are broken by
the change. Zero means analog composite sync, which is not right for
digital displays, and the modes get rejected. Regardless, it used to
work, and it needs to continue to work. Revert the change.

Rejecting modes with analog composite sync was the part that fixed the
gitlab issue 8146 [1]. We'll need to get back to the drawing board with
that.

[1] https://gitlab.freedesktop.org/drm/intel/-/issues/8146

Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/8789
Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/8930
Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/9044
Fixes: ca62297b2085 ("drm/edid: Fix csync detailed mode parsing")
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: dri-devel@lists.freedesktop.org
Cc: <stable@vger.kernel.org> # v6.4+
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Acked-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230815101907.2900768-1-jani.nikula@intel.com

net: dsa: mv88e6xxx: Wait for EEPROM done before HW reset

If the switch is reset during active EEPROM transactions, as in
just after an SoC reset after power up, the I2C bus transaction
may be cut short leaving the EEPROM internal I2C state machine
in the wrong state.  When the switch is reset again, the bad
state machine state may result in data being read from the wrong
memory location causing the switch to enter unexpected mode
rendering it inoperational.

Fixes: a3dcb3e7e70c ("net: dsa: mv88e6xxx: Wait for EEPROM done after HW reset")
Signed-off-by: Alfred Lee <l00g33k@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://lore.kernel.org/r/20230815001323.24739-1-l00g33k@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge tag 'for-netdev' of https://git./linux/kernel/git/bpf/bpf-next

Daniel Borkmann says:

====================
pull-request: bpf-next 2023-08-16

We've added 17 non-merge commits during the last 6 day(s) which contain
a total of 20 files changed, 1179 insertions(+), 37 deletions(-).

The main changes are:

1) Add a BPF hook in sys_socket() to change the protocol ID
   from IPPROTO_TCP to IPPROTO_MPTCP to cover migration for legacy
   applications, from Geliang Tang.

2) Follow-up/fallout fix from the SO_REUSEPORT + bpf_sk_assign work
   to fix a splat on non-fullsock sks in inet[6]_steal_sock,
   from Lorenz Bauer.

3) Improvements to struct_ops links to avoid forcing presence of
   update/validate callbacks. Also add bpf_struct_ops fields documentation,
   from David Vernet.

4) Ensure libbpf sets close-on-exec flag on gzopen, from Marco Vedovati.

5) Several new tcx selftest additions and bpftool link show support for
   tcx and xdp links, from Daniel Borkmann.

6) Fix a smatch warning on uninitialized symbol in
   bpf_perf_link_fill_kprobe, from Yafang Shao.

7) BPF selftest fixes e.g. misplaced break in kfunc_call test,
   from Yipeng Zou.

8) Small cleanup to remove unused declaration bpf_link_new_file,
   from Yue Haibing.

9) Small typo fix to bpftool's perf help message, from Daniel T. Lee.

* tag 'for-netdev' of https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next:
  selftests/bpf: Add mptcpify test
  selftests/bpf: Fix error checks of mptcp open_and_load
  selftests/bpf: Add two mptcp netns helpers
  bpf: Add update_socket_protocol hook
  bpftool: Implement link show support for xdp
  bpftool: Implement link show support for tcx
  selftests/bpf: Add selftest for fill_link_info
  bpf: Fix uninitialized symbol in bpf_perf_link_fill_kprobe()
  net: Fix slab-out-of-bounds in inet[6]_steal_sock
  bpf: Document struct bpf_struct_ops fields
  bpf: Support default .validate() and .update() behavior for struct_ops links
  selftests/bpf: Add various more tcx test cases
  selftests/bpf: Clean up fmod_ret in bench_rename test script
  selftests/bpf: Fix repeat option when kfunc_call verification fails
  libbpf: Set close-on-exec flag on gzopen
  bpftool: fix perf help message
  bpf: Remove unused declaration bpf_link_new_file()
====================

Link: https://lore.kernel.org/r/20230816212840.1539-1-daniel@iogearbox.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Revert "net: ethernet: ti: am65-cpsw: add mqprio qdisc offload in channel mode"

This reverts commit 90bc21aaef4adaefceda2d385756138fc247c0c2.

Patch was merged too hastily, Vladimir requested changes in:
https://lore.kernel.org/all/20230816121305.5dio5tk3chge2ndh@skbuf/

Signed-off-by: Jakub Kicinski <kuba@kernel.org>

drm/nouveau/disp: fix use-after-free in error handling of nouveau_connector_create

We can't simply free the connector after calling drm_connector_init on it.
We need to clean up the drm side first.

It might not fix all regressions from commit 2b5d1c29f6c4
("drm/nouveau/disp: PIOR DP uses GPIO for HPD, not PMGR AUX interrupts"),
but at least it fixes a memory corruption in error handling related to
that commit.

Link: https://lore.kernel.org/lkml/20230806213107.GFZNARG6moWpFuSJ9W@fat_crate.local/
Fixes: 95983aea8003 ("drm/nouveau/disp: add connector class")
Signed-off-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Lyude Paul <lyude@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230814144933.3956959-1-kherbst@redhat.com

net/mlx5: Fix mlx5_cmd_update_root_ft() error flow

The cited patch change mlx5_cmd_update_root_ft() to work with multiple
peer devices. However, it didn't align the error flow as well.
Hence, Fix the error code to work with multiple peer devices.

Fixes: 222dd185833e ("{net/RDMA}/mlx5: introduce lag_for_each_peer")
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Roi Dayan <roid@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

net/mlx5e: XDP, Fix fifo overrun on XDP_REDIRECT

Before this fix, running high rate traffic through XDP_REDIRECT
with multibuf could overrun the fifo used to release the
xdp frames after tx completion. This resulted in corrupted data
being consumed on the free side.

The culplirt was a miscalculation of the fifo size: the maximum ratio
between fifo entries / data segments was incorrect. This ratio serves to
calculate the max fifo size for a full sq where each packet uses the
worst case number of entries in the fifo.

This patch fixes the formula and names the constant. It also makes sure
that future values will use a power of 2 number of entries for the fifo
mask to work.

Signed-off-by: Dragos Tatulea <dtatulea@nvidia.com>
Fixes: 3f734b8c594b ("net/mlx5e: XDP, Use multiple single-entry objects in xdpi_fifo")
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

Revert "Revert "drm/amdgpu/display: change pipe policy for DCN 2.0""

This reverts commit 27dd79c00aeab36cd7542c7a4481a32549038659.

It appears MPC_SPLIT_DYNAMIC still causes problems with multiple
displays on DCN2.0 hardware.  Switch back to MPC_SPLIT_AVOID_MULT_DISP.
This increases power usage with multiple displays, but avoids hangs.

Link: https://gitlab.freedesktop.org/drm/amd/-/issues/2475
Cc: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org # 6.4.x

drm/amd: flush any delayed gfxoff on suspend entry

DCN 3.1.4 is reported to hang on s2idle entry if graphics activity
is happening during entry.  This is because GFXOFF was scheduled as
delayed but RLC gets disabled in s2idle entry sequence which will
hang GFX IP if not already in GFXOFF.

To help this problem, flush any delayed work for GFXOFF early in
s2idle entry sequence to ensure that it's off when RLC is changed.

commit 4b31b92b143f ("drm/amdgpu: complete gfxoff allow signal during
suspend without delay") modified power gating flow so that if called
in s0ix that it ensured that GFXOFF wasn't put in work queue but
instead processed immediately.

This is dead code due to commit 10cb67eb8a1b ("drm/amdgpu: skip
CG/PG for gfx during S0ix") because GFXOFF will now not be explicitly
called as part of the suspend entry code.  Remove that dead code.

Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Tim Huang <tim.huang@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org

drm/amdgpu: skip fence GFX interrupts disable/enable for S0ix

GFX v11.0.1 reported fence fallback timer expired issue on
SDMA and GFX rings after S0ix resume. This is generated by
EOP interrupts are disabled when S0ix suspend but fails to
re-enable when resume because of the GFX is in GFXOFF.

[  203.349571] [drm] Fence fallback timer expired on ring sdma0
[  203.349572] [drm] Fence fallback timer expired on ring gfx_0.0.0
[  203.861635] [drm] Fence fallback timer expired on ring gfx_0.0.0

For S0ix, GFX is in GFXOFF state, avoid to touch the GFX registers
to configure the fence driver interrupts for rings that belong to GFX.
The interrupts configuration will be restored by GFXOFF exit.

Signed-off-by: Tim Huang <Tim.Huang@amd.com>
Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org

drm/amdgpu: skip xcp drm device allocation when out of drm resource

Return 0 when drm device alloc failed with -ENOSPC in
order to  allow amdgpu drive loading. But the xcp without
drm device node assigned won't be visiable in user space.
This helps amdgpu driver loading on system which has more
than 64 nodes, the current limitation.

The proposal to add more drm nodes is discussed in public,
which will support up to 2^20 nodes totally.
kernel drm:
https://lore.kernel.org/lkml/20230724211428.3831636-1-michal.winiarski@intel.com/T/
libdrm:
https://gitlab.freedesktop.org/mesa/drm/-/merge_requests/305

Signed-off-by: James Zhu <James.Zhu@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/pm: Update pci link width for smu v13.0.6

Update addresses of PCIE link width registers,
& link width format used to populate gpu metrics
table for smu v13.0.6

v2:
Removed ESM register update

v3:
Updated patch subject and message

Signed-off-by: Asad Kamal <asad.kamal@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>