Merge version 6.1.52-1+rpi1 and 6.1.55-1 to produce 6.1.55-1+rpi1

Merge linux (6.1.55-1) import into refs/heads/workingbranch

[PATCH 30/30] net: mana: Add gdma stats to ethtool output for mana

Extended performance counter stats in 'ethtool -S <interface>'
for MANA VF to include GDMA tx LSO packets and bytes count.

Tested-on: Ubuntu22
Testcases:
1. LISA testcase:
PERF-NETWORK-TCP-THROUGHPUT-MULTICONNECTION-NTTTCP-Synthetic
2. LISA testcase:
PERF-NETWORK-TCP-THROUGHPUT-MULTICONNECTION-NTTTCP-SRIOV
3. Validated the GDMA stat packets and byte counters
Signed-off-by: Shradha Gupta <shradhagupta@linux.microsoft.com>
Reviewed-by: Pavan Chebbi <pavan.chebbi@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit ac3899c6229649737b9d5cb86e417c98243883dc)
Signed-off-by: Bastian Blank <waldi@debian.org>
Gbp-Pq: Topic features/all/ethernet-microsoft
Gbp-Pq: Name 0030-net-mana-Add-gdma-stats-to-ethtool-output-for-mana.patch

[PATCH 29/30] net: mana: Rename mana_refill_rxoob and remove some empty lines

Rename mana_refill_rxoob for naming consistency.
And remove some empty lines between function call and error
checking.

Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
(cherry picked from commit 5c74064f43c291d9add2b436a2d70205b71a7cc7)
Signed-off-by: Bastian Blank <waldi@debian.org>
Gbp-Pq: Topic features/all/ethernet-microsoft
Gbp-Pq: Name 0029-net-mana-Rename-mana_refill_rxoob-and-remove-some-em.patch

[PATCH 28/30] net: mana: Configure hwc timeout from hardware

At present hwc timeout value is a fixed value. This patch sets the hwc
timeout from the hardware. It now uses a new hardware capability
GDMA_DRV_CAP_FLAG_1_HWC_TIMEOUT_RECONFIG to query and set the value
in hwc_timeout.

Signed-off-by: Souradeep Chakrabarti <schakrabarti@linux.microsoft.com>
Reviewed-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 62c1bff593b7e30041d0273b835af9fd6f5ee737)
Signed-off-by: Bastian Blank <waldi@debian.org>
Gbp-Pq: Topic features/all/ethernet-microsoft
Gbp-Pq: Name 0028-net-mana-Configure-hwc-timeout-from-hardware.patch

[PATCH 27/30] net: mana: Use the correct WQE count for ringing RQ doorbell

The hardware specification specifies that WQE_COUNT should set to 0 for
the Receive Queue. Although currently the hardware doesn't enforce the
check, in the future releases it may check on this value.

Reviewed-by: Haiyang Zhang <haiyangz@microsoft.com>
Reviewed-by: Dexuan Cui <decui@microsoft.com>
Signed-off-by: Long Li <longli@microsoft.com>
Link: https://lore.kernel.org/r/1689622539-5334-3-git-send-email-longli@linuxonhyperv.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
(cherry picked from commit f5e39b57124fd4715d7f0e2f841b8609b38f3e40)
Signed-off-by: Bastian Blank <waldi@debian.org>
Gbp-Pq: Topic features/all/ethernet-microsoft
Gbp-Pq: Name 0027-net-mana-Use-the-correct-WQE-count-for-ringing-RQ-do.patch

[PATCH 26/30] net: mana: Batch ringing RX queue doorbell on receiving packets

It's inefficient to ring the doorbell page every time a WQE is posted to
the received queue. Excessive MMIO writes result in CPU spending more
time waiting on LOCK instructions (atomic operations), resulting in
poor scaling performance.

Move the code for ringing doorbell page to where after we have posted all
WQEs to the receive queue during a callback from napi_poll().

With this change, tests showed an improvement from 120G/s to 160G/s on a
200G physical link, with 16 or 32 hardware queues.

Tests showed no regression in network latency benchmarks on single
connection.

Reviewed-by: Haiyang Zhang <haiyangz@microsoft.com>
Reviewed-by: Dexuan Cui <decui@microsoft.com>
Signed-off-by: Long Li <longli@microsoft.com>
Link: https://lore.kernel.org/r/1689622539-5334-2-git-send-email-longli@linuxonhyperv.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
(cherry picked from commit da4e8648079eb6f26f3a88d8c34270a057e2bfe6)
Signed-off-by: Bastian Blank <waldi@debian.org>
Gbp-Pq: Topic features/all/ethernet-microsoft
Gbp-Pq: Name 0026-net-mana-Batch-ringing-RX-queue-doorbell-on-receivin.patch

[PATCH 25/30] net: mana: use vmalloc_array and vcalloc

Use vmalloc_array and vcalloc to protect against
multiplication overflows.

The changes were done using the following Coccinelle
semantic patch:

// <smpl>
@initialize:ocaml@
@@

let rename alloc =
  match alloc with
    "vmalloc" -> "vmalloc_array"
  | "vzalloc" -> "vcalloc"
  | _ -> failwith "unknown"

@@
    size_t e1,e2;
    constant C1, C2;
    expression E1, E2, COUNT, x1, x2, x3;
    typedef u8;
    typedef __u8;
    type t = {u8,__u8,char,unsigned char};
    identifier alloc = {vmalloc,vzalloc};
    fresh identifier realloc = script:ocaml(alloc) { rename alloc };
@@

(
      alloc(x1*x2*x3)
|
      alloc(C1 * C2)
|
      alloc((sizeof(t)) * (COUNT), ...)
|
-     alloc((e1) * (e2))
+     realloc(e1, e2)
|
-     alloc((e1) * (COUNT))
+     realloc(COUNT, e1)
|
-     alloc((E1) * (E2))
+     realloc(E1, E2)
)
// </smpl>

Signed-off-by: Julia Lawall <Julia.Lawall@inria.fr>
Link: https://lore.kernel.org/r/20230627144339.144478-23-Julia.Lawall@inria.fr
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
(cherry picked from commit e9c74f8b8a31f77f8e9d7bbed5fc9f2eacbf32a5)
Signed-off-by: Bastian Blank <waldi@debian.org>
Gbp-Pq: Topic features/all/ethernet-microsoft
Gbp-Pq: Name 0025-net-mana-use-vmalloc_array-and-vcalloc.patch

[PATCH 24/30] RDMA/mana_ib: Use v2 version of cfg_rx_steer_req to enable RX coalescing

With RX coalescing, one CQE entry can be used to indicate multiple packets
on the receive queue. This saves processing time and PCI bandwidth over
the CQ.

The MANA Ethernet driver also uses the v2 version of the protocol. It
doesn't use RX coalescing and its behavior is not changed.

Link: https://lore.kernel.org/r/1684045095-31228-1-git-send-email-longli@linuxonhyperv.com
Signed-off-by: Long Li <longli@microsoft.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
(cherry picked from commit 2145328515c8fa9b8a9f7889250bc6c032f2a0e6)
Signed-off-by: Bastian Blank <waldi@debian.org>
Gbp-Pq: Topic features/all/ethernet-microsoft
Gbp-Pq: Name 0024-RDMA-mana_ib-Use-v2-version-of-cfg_rx_steer_req-to-e.patch

[PATCH 23/23] net: mana: Add support for vlan tagging

To support vlan, use MANA_LONG_PKT_FMT if vlan tag is present in TX
skb. Then extract the vlan tag from the skb struct, and save it to
tx_oob for the NIC to transmit. For vlan tags on the payload, they
are accepted by the NIC too.

For RX, extract the vlan tag from CQE and put it into skb.

Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit b803d1fded4085d268507a432dac8077ead68971)
Signed-off-by: Bastian Blank <waldi@debian.org>
Gbp-Pq: Topic features/all/ethernet-microsoft
Gbp-Pq: Name 0023-net-mana-Add-support-for-vlan-tagging.patch

[PATCH 22/23] net: mana: Fix perf regression: remove rx_cqes, tx_cqes counters

The apc->eth_stats.rx_cqes is one per NIC (vport), and it's on the
frequent and parallel code path of all queues. So, r/w into this
single shared variable by many threads on different CPUs creates a
lot caching and memory overhead, hence perf regression. And, it's
not accurate due to the high volume concurrent r/w.

For example, a workload is iperf with 128 threads, and with RPS
enabled. We saw perf regression of 25% with the previous patch
adding the counters. And this patch eliminates the regression.

Since the error path of mana_poll_rx_cq() already has warnings, so
keeping the counter and convert it to a per-queue variable is not
necessary. So, just remove this counter from this high frequency
code path.

Also, remove the tx_cqes counter for the same reason. We have
warnings & other counters for errors on that path, and don't need
to count every normal cqe processing.

Cc: stable@vger.kernel.org
Fixes: bd7fc6e1957c ("net: mana: Add new MANA VF performance counters for easier troubleshooting")
Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com>
Reviewed-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Link: https://lore.kernel.org/r/1685115537-31675-1-git-send-email-haiyangz@microsoft.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
(cherry picked from commit 1919b39fc6eabb9a6f9a51706ff6d03865f5df29)
Signed-off-by: Bastian Blank <waldi@debian.org>
Gbp-Pq: Topic features/all/ethernet-microsoft
Gbp-Pq: Name 0022-net-mana-Fix-perf-regression-remove-rx_cqes-tx_cqes-.patch

[PATCH 21/23] net: mana: Check if netdev/napi_alloc_frag returns single page

netdev/napi_alloc_frag() may fall back to single page which is smaller
than the requested size.
Add error checking to avoid memory overwritten.

Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
(cherry picked from commit df18f2da302f169e1a29098c6ca3b474f1b0269e)
Signed-off-by: Bastian Blank <waldi@debian.org>
Gbp-Pq: Topic features/all/ethernet-microsoft
Gbp-Pq: Name 0021-net-mana-Check-if-netdev-napi_alloc_frag-returns-sin.patch

[PATCH 20/23] net: mana: Add support for jumbo frame

During probe, get the hardware-allowed max MTU by querying the device
configuration. Users can select MTU up to the device limit.
When XDP is in use, limit MTU settings so the buffer size is within
one page. And, when MTU is set to a too large value, XDP is not allowed
to run.
Also, to prevent changing MTU fails, and leaves the NIC in a bad state,
pre-allocate all buffers before starting the change. So in low memory
condition, it will return error, without affecting the NIC.

Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com>
Reviewed-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 80f6215b450eb8e92d8b1f117abf5ecf867f963e)
Signed-off-by: Bastian Blank <waldi@debian.org>
Gbp-Pq: Topic features/all/ethernet-microsoft
Gbp-Pq: Name 0020-net-mana-Add-support-for-jumbo-frame.patch

[PATCH 19/23] net: mana: Enable RX path to handle various MTU sizes

Update RX data path to allocate and use RX queue DMA buffers with
proper size based on potentially various MTU sizes.

Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com>
Reviewed-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 2fbbd712baf1c60996554326728bbdbef5616e12)
Signed-off-by: Bastian Blank <waldi@debian.org>
Gbp-Pq: Topic features/all/ethernet-microsoft
Gbp-Pq: Name 0019-net-mana-Enable-RX-path-to-handle-various-MTU-sizes.patch

[PATCH 18/23] net: mana: Refactor RX buffer allocation code to prepare for various MTU

Move out common buffer allocation code from mana_process_rx_cqe() and
mana_alloc_rx_wqe() to helper functions.
Refactor related variables so they can be changed in one place, and buffer
sizes are in sync.

Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com>
Reviewed-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit a2917b23497e4205db32271e4e06e142a9f8a6aa)
Signed-off-by: Bastian Blank <waldi@debian.org>
Gbp-Pq: Topic features/all/ethernet-microsoft
Gbp-Pq: Name 0018-net-mana-Refactor-RX-buffer-allocation-code-to-prepa.patch

[PATCH 17/23] net: mana: Use napi_build_skb in RX path

Use napi_build_skb() instead of build_skb() to take advantage of the
NAPI percpu caches to obtain skbuff_head.

Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com>
Reviewed-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit ce518bc3e9ca342309995c9270c3ec4892963695)
Signed-off-by: Bastian Blank <waldi@debian.org>
Gbp-Pq: Topic features/all/ethernet-microsoft
Gbp-Pq: Name 0017-net-mana-Use-napi_build_skb-in-RX-path.patch

[PATCH 16/23] net: mana: Remove redundant pci_clear_master

Remove pci_clear_master to simplify the code,
the bus-mastering is also cleared in do_pci_disable_device,
like this:
./drivers/pci/pci.c:2197
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 dev->is_busmaster is set to 0 in pci_disable_device.

Signed-off-by: Cai Huoqing <cai.huoqing@linux.dev>
Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit 2d59af8307526f2829fdec9c5c5898a857d55180)
Signed-off-by: Bastian Blank <waldi@debian.org>
Gbp-Pq: Topic features/all/ethernet-microsoft
Gbp-Pq: Name 0016-net-mana-Remove-redundant-pci_clear_master.patch

[PATCH 15/23] net: mana: Add new MANA VF performance counters for easier troubleshooting

Extended performance counter stats in 'ethtool -S <interface>' output
for MANA VF to facilitate troubleshooting.

Tested-on: Ubuntu22
Signed-off-by: Shradha Gupta <shradhagupta@linux.microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
(cherry picked from commit bd7fc6e1957c2102866f9e464c1f2302e891b7e9)
Signed-off-by: Bastian Blank <waldi@debian.org>
Gbp-Pq: Topic features/all/ethernet-microsoft
Gbp-Pq: Name 0015-net-mana-Add-new-MANA-VF-performance-counters-for-ea.patch

[PATCH 14/23] net: mana: Fix accessing freed irq affinity_hint

After calling irq_set_affinity_and_hint(), the cpumask pointer is
saved in desc->affinity_hint, and will be used later when reading
/proc/irq/<num>/affinity_hint. So the cpumask variable needs to be
persistent. Otherwise, we are accessing freed memory when reading
the affinity_hint file.

Also, need to clear affinity_hint before free_irq(), otherwise there
is a one-time warning and stack trace during module unloading:

 [  243.948687] WARNING: CPU: 10 PID: 1589 at kernel/irq/manage.c:1913 free_irq+0x318/0x360
 ...
 [  243.948753] Call Trace:
 [  243.948754]  <TASK>
 [  243.948760]  mana_gd_remove_irqs+0x78/0xc0 [mana]
 [  243.948767]  mana_gd_remove+0x3e/0x80 [mana]
 [  243.948773]  pci_device_remove+0x3d/0xb0
 [  243.948778]  device_remove+0x46/0x70
 [  243.948782]  device_release_driver_internal+0x1fe/0x280
 [  243.948785]  driver_detach+0x4e/0xa0
 [  243.948787]  bus_remove_driver+0x70/0xf0
 [  243.948789]  driver_unregister+0x35/0x60
 [  243.948792]  pci_unregister_driver+0x44/0x90
 [  243.948794]  mana_driver_exit+0x14/0x3fe [mana]
 [  243.948800]  __do_sys_delete_module.constprop.0+0x185/0x2f0

To fix the bug, use the persistent mask, cpumask_of(cpu#), and set
affinity_hint to NULL before freeing the IRQ, as required by free_irq().

Cc: stable@vger.kernel.org
Fixes: 71fa6887eeca ("net: mana: Assign interrupts to CPUs based on NUMA nodes")
Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com>
Reviewed-by: Michael Kelley <mikelley@microsoft.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Link: https://lore.kernel.org/r/1675718929-19565-1-git-send-email-haiyangz@microsoft.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
(cherry picked from commit 18a048370b06a3a521219e9e5b10bdc2178ef19c)
Signed-off-by: Bastian Blank <waldi@debian.org>
Gbp-Pq: Topic features/all/ethernet-microsoft
Gbp-Pq: Name 0014-net-mana-Fix-accessing-freed-irq-affinity_hint.patch

[PATCH 13/23] net: mana: Fix return type of mana_start_xmit()

The ndo_start_xmit field in net_device_ops is expected to be of type
netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb, struct net_device *dev).

The mismatched return type breaks forward edge kCFI since the underlying
function definition does not match the function hook definition. A new
warning in clang will catch this at compile time:

  drivers/net/ethernet/microsoft/mana/mana_en.c:382:21: error: incompatible function pointer types initializing 'netdev_tx_t (*)(struct sk_buff *, struct net_device *)' (aka 'enum netdev_tx (*)(struct sk_buff *, struct net_device *)') with an expression of type 'int (struct sk_buff *, struct net_device *)' [-Werror,-Wincompatible-function-pointer-types-strict]
          .ndo_start_xmit         = mana_start_xmit,
                                    ^~~~~~~~~~~~~~~
  1 error generated.

The return type of mana_start_xmit should be changed from int to
netdev_tx_t.

Reported-by: Dan Carpenter <error27@gmail.com>
Link: https://github.com/ClangBuiltLinux/linux/issues/1703
Link: https://github.com/ClangBuiltLinux/linux/issues/1750
Signed-off-by: Nathan Huckleberry <nhuck@google.com>
Reviewed-by: Dexuan Cui <decui@microsoft.com>
[nathan: Rebase on net-next and resolve conflicts
         Add note about new clang warning]
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Link: https://lore.kernel.org/r/20221109002629.1446680-1-nathan@kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
(cherry picked from commit 0c9ef08a4d0fd6c5e6000597b506235d71a85a61)
Signed-off-by: Bastian Blank <waldi@debian.org>
Gbp-Pq: Topic features/all/ethernet-microsoft
Gbp-Pq: Name 0013-net-mana-Fix-return-type-of-mana_start_xmit.patch

[PATCH 12/23] net: mana: Define data structures for protection domain and memory registration

The MANA hardware support protection domain and memory registration for use
in RDMA environment. Add those definitions and expose them for use by the
RDMA driver.

Signed-off-by: Ajay Sharma <sharmaajay@microsoft.com>
Signed-off-by: Long Li <longli@microsoft.com>
Link: https://lore.kernel.org/r/1667502990-2559-12-git-send-email-longli@linuxonhyperv.com
Reviewed-by: Dexuan Cui <decui@microsoft.com>
Acked-by: Haiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
(cherry picked from commit 28c66cfa45388af1126985d1114e0ed762eb2abd)
Signed-off-by: Bastian Blank <waldi@debian.org>
Gbp-Pq: Topic features/all/ethernet-microsoft
Gbp-Pq: Name 0012-net-mana-Define-data-structures-for-protection-domai.patch

[PATCH 11/23] net: mana: Define and process GDMA response code GDMA_STATUS_MORE_ENTRIES

When doing memory registration, the PF may respond with
GDMA_STATUS_MORE_ENTRIES to indicate a follow request is needed. This is
not an error and should be processed as expected.

Signed-off-by: Ajay Sharma <sharmaajay@microsoft.com>
Reviewed-by: Dexuan Cui <decui@microsoft.com>
Signed-off-by: Long Li <longli@microsoft.com>
Link: https://lore.kernel.org/r/1667502990-2559-10-git-send-email-longli@linuxonhyperv.com
Acked-by: Haiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
(cherry picked from commit de372f2a9ca7ada2698ecac7df8f02407cd98fa0)
Signed-off-by: Bastian Blank <waldi@debian.org>
Gbp-Pq: Topic features/all/ethernet-microsoft
Gbp-Pq: Name 0011-net-mana-Define-and-process-GDMA-response-code-GDMA_.patch

[PATCH 10/23] net: mana: Define max values for SGL entries

The number of maximum SGl entries should be computed from the maximum
WQE size for the intended queue type and the corresponding OOB data
size. This guarantees the hardware queue can successfully queue requests
up to the queue depth exposed to the upper layer.

Reviewed-by: Dexuan Cui <decui@microsoft.com>
Signed-off-by: Long Li <longli@microsoft.com>
Link: https://lore.kernel.org/r/1667502990-2559-9-git-send-email-longli@linuxonhyperv.com
Acked-by: Haiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
(cherry picked from commit aa56549792fb348892fbbae67f6f0c71bb750b65)
Signed-off-by: Bastian Blank <waldi@debian.org>
Gbp-Pq: Topic features/all/ethernet-microsoft
Gbp-Pq: Name 0010-net-mana-Define-max-values-for-SGL-entries.patch

[PATCH 09/23] net: mana: Move header files to a common location

In preparation to add MANA RDMA driver, move all the required header files
to a common location for use by both Ethernet and RDMA drivers.

Reviewed-by: Dexuan Cui <decui@microsoft.com>
Signed-off-by: Long Li <longli@microsoft.com>
Link: https://lore.kernel.org/r/1667502990-2559-8-git-send-email-longli@linuxonhyperv.com
Acked-by: Haiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
(cherry picked from commit fd325cd648f15eb9a8b32a68de3bafc72bcfe753)
Signed-off-by: Bastian Blank <waldi@debian.org>
Gbp-Pq: Topic features/all/ethernet-microsoft
Gbp-Pq: Name 0009-net-mana-Move-header-files-to-a-common-location.patch

[PATCH 08/23] net: mana: Record port number in netdev

The port number is useful for user-mode application to identify this
net device based on port index. Set to the correct value in ndev.

Reviewed-by: Dexuan Cui <decui@microsoft.com>
Signed-off-by: Long Li <longli@microsoft.com>
Link: https://lore.kernel.org/r/1667502990-2559-7-git-send-email-longli@linuxonhyperv.com
Acked-by: Haiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
(cherry picked from commit d44089e555ffe63a49cc6e94d0c03d933e413059)
Signed-off-by: Bastian Blank <waldi@debian.org>
Gbp-Pq: Topic features/all/ethernet-microsoft
Gbp-Pq: Name 0008-net-mana-Record-port-number-in-netdev.patch

[PATCH 07/23] net: mana: Export Work Queue functions for use by RDMA driver

RDMA device may need to create Ethernet device queues for use by Queue
Pair type RAW. This allows a user-mode context accesses Ethernet hardware
queues. Export the supporting functions for use by the RDMA driver.

Reviewed-by: Dexuan Cui <decui@microsoft.com>
Signed-off-by: Long Li <longli@microsoft.com>
Link: https://lore.kernel.org/r/1667502990-2559-6-git-send-email-longli@linuxonhyperv.com
Acked-by: Haiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
(cherry picked from commit 4c0ff7a106e16ab63e0b597557255c012f179578)
Signed-off-by: Bastian Blank <waldi@debian.org>
Gbp-Pq: Topic features/all/ethernet-microsoft
Gbp-Pq: Name 0007-net-mana-Export-Work-Queue-functions-for-use-by-RDMA.patch

[PATCH 06/23] net: mana: Set the DMA device max segment size

MANA hardware doesn't have any restrictions on the DMA segment size, set it
to the max allowed value.

Signed-off-by: Ajay Sharma <sharmaajay@microsoft.com>
Reviewed-by: Dexuan Cui <decui@microsoft.com>
Signed-off-by: Long Li <longli@microsoft.com>
Link: https://lore.kernel.org/r/1667502990-2559-5-git-send-email-longli@linuxonhyperv.com
Acked-by: Haiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
(cherry picked from commit 6fe254160bd033a1e62dbad9b734183b31144678)
Signed-off-by: Bastian Blank <waldi@debian.org>
Gbp-Pq: Topic features/all/ethernet-microsoft
Gbp-Pq: Name 0006-net-mana-Set-the-DMA-device-max-segment-size.patch

[PATCH 05/23] net: mana: Handle vport sharing between devices

For outgoing packets, the PF requires the VF to configure the vport with
corresponding protection domain and doorbell ID for the kernel or user
context. The vport can't be shared between different contexts.

Implement the logic to exclusively take over the vport by either the
Ethernet device or RDMA device.

Reviewed-by: Dexuan Cui <decui@microsoft.com>
Signed-off-by: Long Li <longli@microsoft.com>
Link: https://lore.kernel.org/r/1667502990-2559-4-git-send-email-longli@linuxonhyperv.com
Acked-by: Haiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
(cherry picked from commit b5c1c9855be3b5b978fde975a63df3cabc273faa)
Signed-off-by: Bastian Blank <waldi@debian.org>
Gbp-Pq: Topic features/all/ethernet-microsoft
Gbp-Pq: Name 0005-net-mana-Handle-vport-sharing-between-devices.patch

[PATCH 04/23] net: mana: Record the physical address for doorbell page region

For supporting RDMA device with multiple user contexts with their
individual doorbell pages, record the start address of doorbell page
region for use by the RDMA driver to allocate user context doorbell IDs.

Reviewed-by: Dexuan Cui <decui@microsoft.com>
Signed-off-by: Long Li <longli@microsoft.com>
Link: https://lore.kernel.org/r/1667502990-2559-3-git-send-email-longli@linuxonhyperv.com
Acked-by: Haiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
(cherry picked from commit f3dc096246091048677c45cfc0e24ad512927b52)
Signed-off-by: Bastian Blank <waldi@debian.org>
Gbp-Pq: Topic features/all/ethernet-microsoft
Gbp-Pq: Name 0004-net-mana-Record-the-physical-address-for-doorbell-pa.patch

[PATCH 03/23] net: mana: Add support for auxiliary device

In preparation for supporting MANA RDMA driver, add support for auxiliary
device in the Ethernet driver. The RDMA device is modeled as an auxiliary
device to the Ethernet device.

Reviewed-by: Dexuan Cui <decui@microsoft.com>
Signed-off-by: Long Li <longli@microsoft.com>
Link: https://lore.kernel.org/r/1667502990-2559-2-git-send-email-longli@linuxonhyperv.com
Acked-by: Haiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
(cherry picked from commit a69839d4327d053b18d8e1b0e7ddeee78db78f4f)
Signed-off-by: Bastian Blank <waldi@debian.org>
Gbp-Pq: Topic features/all/ethernet-microsoft
Gbp-Pq: Name 0003-net-mana-Add-support-for-auxiliary-device.patch

[PATCH 02/23] net: mana: Assign interrupts to CPUs based on NUMA nodes

In large VMs with multiple NUMA nodes, network performance is usually
best if network interrupts are all assigned to the same virtual NUMA
node. This patch assigns online CPU according to a numa aware policy,
local cpus are returned first, followed by non-local ones, then it wraps
around.

Signed-off-by: Saurabh Sengar <ssengar@linux.microsoft.com>
Reviewed-by: Haiyang Zhang <haiyangz@microsoft.com>
Link: https://lore.kernel.org/r/1667282761-11547-1-git-send-email-ssengar@linux.microsoft.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
(cherry picked from commit 71fa6887eeca7b631528f9c7a39815498de8028c)
Signed-off-by: Bastian Blank <waldi@debian.org>
Gbp-Pq: Topic features/all/ethernet-microsoft
Gbp-Pq: Name 0002-net-mana-Assign-interrupts-to-CPUs-based-on-NUMA-nod.patch

[PATCH 01/23] net: Remove the obsolte u64_stats_fetch_*_irq() users (drivers).

Now that the 32bit UP oddity is gone and 32bit uses always a sequence
count, there is no need for the fetch_irq() variants anymore.

Convert to the regular interface.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
(cherry picked from commit 068c38ad88ccb09e5e966d4db5cedab0e02b3b95)
Signed-off-by: Bastian Blank <waldi@debian.org>
Gbp-Pq: Topic features/all/ethernet-microsoft
Gbp-Pq: Name 0001-net-Remove-the-obsolte-u64_stats_fetch_-_irq-users-d.patch

fbdev/offb: Update expected device name

Origin: https://git.kernel.org/pub/scm/linux/kernel/git/deller/linux-fbdev.git/commit?id=27c74ea74be805ccba1bc1a0a03cc79c51dca6ea

Since commit 241d2fb56a18 ("of: Make OF framebuffer device names unique"),
as spotted by Frédéric Bonnard, the historical "of-display" device is
gone: the updated logic creates "of-display.0" instead, then as many
"of-display.N" as required.

This means that offb no longer finds the expected device, which prevents
the Debian Installer from setting up its interface, at least on ppc64el.

It might be better to iterate on all possible nodes, but updating the
hardcoded device from "of-display" to "of-display.0" is confirmed to fix
the Debian Installer at the very least.

Link: https://bugzilla.kernel.org/show_bug.cgi?id=217328
Link: https://bugs.debian.org/1033058
Fixes: 241d2fb56a18 ("of: Make OF framebuffer device names unique")
Cc: stable@vger.kernel.org
Signed-off-by: Cyril Brulebois <cyril@debamax.com>
Signed-off-by: Helge Deller <deller@gmx.de>
Gbp-Pq: Topic bugfix/powerpc
Gbp-Pq: Name fbdev-offb-Update-expected-device-name.patch

tools/perf: Fix missing LDFLAGS for some programs

Signed-off-by: Ben Hutchings <benh@debian.org>
Gbp-Pq: Topic bugfix/all
Gbp-Pq: Name tools-perf-fix-missing-ldflags-for-some-programs.patch

libapi: Define _FORTIFY_SOURCE as 2, not empty

Signed-off-by: Ben Hutchings <benh@debian.org>
Gbp-Pq: Topic bugfix/all
Gbp-Pq: Name libapi-define-_fortify_source-as-2-not-empty.patch

tools/perf: pmu-events: Fix reproducibility

Forwarded: https://lore.kernel.org/lkml/20190825131329.naqzd5kwg7mw5d3f@decadent.org.uk/T/#u

jevents.py enumerates files and outputs the corresponding C structs in
the order they are found.  This makes it sensitive to directory
ordering, so that the perf executable is not reproducible.

To avoid this, sort the entries returned by os.scandir() before
processing them.

References: https://tests.reproducible-builds.org/debian/dbdtxt/bullseye/i386/linux_4.19.37-6.diffoscope.txt.gz
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Gbp-Pq: Topic bugfix/all
Gbp-Pq: Name tools-perf-pmu-events-fix-reproducibility.patch

cpupower: Fix checks for CPU existence

Forwarded: https://marc.info/?l=linux-pm&m=149248268214265

Calls to cpufreq_cpu_exists(cpu) were converted to
cpupower_is_cpu_online(cpu) when libcpupower was introduced and the
former function was deleted.  However, cpupower_is_cpu_online() does
not distinguish physically absent and offline CPUs, and does not set
errno.

cpufreq-set has already been fixed (commit c25badc9ceb6).

In cpufreq-bench, which prints an error message for offline CPUs,
properly distinguish and report the zero and negative cases.

Fixes: ac5a181d065d ("cpupower: Add cpuidle parts into library")
Fixes: 53d1cd6b125f ("cpupowerutils: bench - Fix cpu online check")
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
[carnil: Update/Refresh patch for 4.14.17: The issue with the
incorrect check has been fixed with upstream commit 53d1cd6b125f.
Keep in the patch the distinction and report for the zero and
negative cases.]

Gbp-Pq: Topic bugfix/all
Gbp-Pq: Name cpupower-fix-checks-for-cpu-existence.patch

cpupower: Bump soname version

Forwarded: http://mid.gmane.org/20160610005619.GQ7555@decadent.org.uk

Several functions in the libcpupower API are renamed or removed in
Linux 4.7.  This is an backward-incompatible ABI change, so the
library soname should change from libcpupower.so.0 to
libcpupower.so.1.

Fixes: ac5a181d065d ("cpupower: Add cpuidle parts into library")
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Gbp-Pq: Topic bugfix/all
Gbp-Pq: Name cpupower-bump-soname-version.patch

tools/build: Remove bpf() run-time check at build time

Forwarded: no

It is not correct to test that a syscall works on the build system's
kernel.  We might be building on an earlier kernel version or with
security restrictions that block bpf().

Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Gbp-Pq: Topic bugfix/all
Gbp-Pq: Name tools-build-remove-bpf-run-time-check-at-build-time.patch

Revert "perf build: Fix libunwind feature detection on 32-bit x86"

Forwarded: no

This reverts commit 05b41775e2edd69a83f592e3534930c934d4038e.
It broke feature detection that was working just fine for us.

Gbp-Pq: Topic bugfix/x86
Gbp-Pq: Name revert-perf-build-fix-libunwind-feature-detection-on.patch

tools/perf: Remove shebang lines from perf scripts

Forwarded: no

perf scripts need to be invoked through perf, not directly through
perl (or other language interpreter).  So including shebang lines in
them is useless and possibly misleading.

Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Gbp-Pq: Topic bugfix/all
Gbp-Pq: Name tools-perf-remove-shebangs.patch

perf tools: Use $KBUILD_BUILD_TIMESTAMP as man page date

Forwarded: http://mid.gmane.org/20160517132809.GE7555@decadent.org.uk

This allows man pages to be built reproducibly.

Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Gbp-Pq: Topic bugfix/all
Gbp-Pq: Name tools-perf-man-date.patch

kbuild: Fix recordmcount dependency for OOT modules

Forwarded: no

We never rebuild anything in-tree when building an out-of-tree
modules, so external modules should not depend on the recordmcount
sources.

Gbp-Pq: Topic bugfix/all
Gbp-Pq: Name kbuild-fix-recordmcount-dependency.patch

usbip: Document TCP wrappers

Forwarded: no

Add references to TCP wrappers configuration in the manual page.

Gbp-Pq: Topic bugfix/all
Gbp-Pq: Name usbip-document-tcp-wrappers.patch

module: Disable matching missing version CRC

Forwarded: not-needed

This partly reverts commit cd3caefb4663e3811d37cc2afad3cce642d60061.
We want to fail closed if a symbol version CRC is missing, as the
alternative may allow subverting module signing.

Gbp-Pq: Topic bugfix/all
Gbp-Pq: Name module-disable-matching-missing-version-crc.patch

ipv4: fix null-deref in ipv4_link_failure

Origin: https://git.kernel.org/pub/scm/linux/kernel/git/stable/stable-queue.git/tree/queue-6.1/ipv4-fix-null-deref-in-ipv4_link_failure.patch

[ Upstream commit 0113d9c9d1ccc07f5a3710dac4aa24b6d711278c ]

Currently, we assume the skb is associated with a device before calling
__ip_options_compile, which is not always the case if it is re-routed by
ipvs.
When skb->dev is NULL, dev_net(skb->dev) will become null-dereference.
This patch adds a check for the edge case and switch to use the net_device
from the rtable when skb->dev is NULL.

Fixes: ed0de45a1008 ("ipv4: recompile ip options in ipv4_link_failure")
Suggested-by: David Ahern <dsahern@kernel.org>
Signed-off-by: Kyle Zeng <zengyhkyle@gmail.com>
Cc: Stephen Suryaputra <ssuryaextr@gmail.com>
Cc: Vadim Fedorenko <vfedorenko@novek.ru>
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
Gbp-Pq: Topic bugfix/all
Gbp-Pq: Name ipv4-fix-null-deref-in-ipv4_link_failure.patch

netfilter: ipset: Fix race between IPSET_CMD_CREATE and IPSET_CMD_SWAP

[ Upstream commit 7433b6d2afd512d04398c73aa984d1e285be125b ]

Kyle Zeng reported that there is a race between IPSET_CMD_ADD and IPSET_CMD_SWAP
in netfilter/ip_set, which can lead to the invocation of `__ip_set_put` on a
wrong `set`, triggering the `BUG_ON(set->ref == 0);` check in it.

The race is caused by using the wrong reference counter, i.e. the ref counter instead
of ref_netlink.

Fixes: 24e227896bbf ("netfilter: ipset: Add schedule point in call_ad().")
Reported-by: Kyle Zeng <zengyhkyle@gmail.com>
Closes: https://lore.kernel.org/netfilter-devel/ZPZqetxOmH+w%2Fmyc@westworld/#r
Tested-by: Kyle Zeng <zengyhkyle@gmail.com>
Signed-off-by: Jozsef Kadlecsik <kadlec@netfilter.org>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
Gbp-Pq: Topic bugfix/all
Gbp-Pq: Name netfilter-ipset-fix-race-between-ipset_cmd_create-an.patch

netfilter: nf_tables: disable toggling dormant table state more than once

[ Upstream commit c9bd26513b3a11b3adb3c2ed8a31a01a87173ff1 ]

nft -f -<<EOF
add table ip t
add table ip t { flags dormant; }
add chain ip t c { type filter hook input priority 0; }
add table ip t
EOF

Triggers a splat from nf core on next table delete because we lose
track of right hook register state:

WARNING: CPU: 2 PID: 1597 at net/netfilter/core.c:501 __nf_unregister_net_hook
RIP: 0010:__nf_unregister_net_hook+0x41b/0x570
 nf_unregister_net_hook+0xb4/0xf0
 __nf_tables_unregister_hook+0x160/0x1d0
[..]

The above should have table in *active* state, but in fact no
hooks were registered.

Reject on/off/on games rather than attempting to fix this.

Fixes: 179d9ba5559a ("netfilter: nf_tables: fix table flag updates")
Reported-by: "Lee, Cherie-Anne" <cherie.lee@starlabs.sg>
Cc: Bing-Jhong Billy Jheng <billy@starlabs.sg>
Cc: info@starlabs.sg
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
Gbp-Pq: Topic bugfix/all
Gbp-Pq: Name netfilter-nf_tables-disable-toggling-dormant-table-s.patch

netfilter, bpf: Adjust timeouts of non-confirmed CTs in bpf_ct_insert_entry()

[ Upstream commit 837723b22a63cfbff584655b009b9d488d0e9087 ]

bpf_nf testcase fails on s390x: bpf_skb_ct_lookup() cannot find the entry
that was added by bpf_ct_insert_entry() within the same BPF function.

The reason is that this entry is deleted by nf_ct_gc_expired().

The CT timeout starts ticking after the CT confirmation; therefore
nf_conn.timeout is initially set to the timeout value, and
__nf_conntrack_confirm() sets it to the deadline value.

bpf_ct_insert_entry() sets IPS_CONFIRMED_BIT, but does not adjust the
timeout, making its value meaningless and causing false positives.

Fix the problem by making bpf_ct_insert_entry() adjust the timeout,
like __nf_conntrack_confirm().

Fixes: 2cdaa3eefed8 ("netfilter: conntrack: restore IPS_CONFIRMED out of nf_conntrack_hash_check_insert()")
Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Cc: Florian Westphal <fw@strlen.de>
Link: https://lore.kernel.org/bpf/20230830011128.1415752-3-iii@linux.ibm.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
Gbp-Pq: Topic bugfix/all
Gbp-Pq: Name netfilter-bpf-adjust-timeouts-of-non-confirmed-cts-i.patch

netfilter: conntrack: fix extension size table

[ Upstream commit 4908d5af16676b9d2901830551c2af911e452524 ]

The size table is incorrect due to copypaste error,
this reserves more size than needed.

TSTAMP reserved 32 instead of 16 bytes.
TIMEOUT reserved 16 instead of 8 bytes.

Fixes: 5f31edc0676b ("netfilter: conntrack: move extension sizes into core")
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
Gbp-Pq: Topic bugfix/all
Gbp-Pq: Name netfilter-conntrack-fix-extension-size-table.patch

netfilter: nf_tables: disallow element removal on anonymous sets

[ Upstream commit 23a3bfd4ba7acd36abf52b78605f61b21bdac216 ]

Anonymous sets need to be populated once at creation and then they are
bound to rule since 938154b93be8 ("netfilter: nf_tables: reject unbound
anonymous set before commit phase"), otherwise transaction reports
EINVAL.

Userspace does not need to delete elements of anonymous sets that are
not yet bound, reject this with EOPNOTSUPP.

From flush command path, skip anonymous sets, they are expected to be
bound already. Otherwise, EINVAL is hit at the end of this transaction
for unbound sets.

Fixes: 96518518cc41 ("netfilter: add nftables")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
Gbp-Pq: Topic bugfix/all
Gbp-Pq: Name netfilter-nf_tables-disallow-element-removal-on-anon.patch

netfilter: nf_tables: fix memleak when more than 255 elements expired

commit cf5000a7787cbc10341091d37245a42c119d26c5 upstream.

When more than 255 elements expired we're supposed to switch to a new gc
container structure.

This never happens: u8 type will wrap before reaching the boundary
and nft_trans_gc_space() always returns true.

This means we recycle the initial gc container structure and
lose track of the elements that came before.

While at it, don't deref 'gc' after we've passed it to call_rcu.

Fixes: 5f68718b34a5 ("netfilter: nf_tables: GC transaction API to avoid race with control plane")
Reported-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
Gbp-Pq: Topic bugfix/all
Gbp-Pq: Name netfilter-nf_tables-fix-memleak-when-more-than-255-e.patch

netfilter: nft_set_hash: try later when GC hits EAGAIN on iteration

commit b079155faae94e9b3ab9337e82100a914ebb4e8d upstream.

Skip GC run if iterator rewinds to the beginning with EAGAIN, otherwise GC
might collect the same element more than once.

Fixes: f6c383b8c31a ("netfilter: nf_tables: adapt set backend to use GC transaction API")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
Gbp-Pq: Topic bugfix/all
Gbp-Pq: Name netfilter-nft_set_hash-try-later-when-gc-hits-eagain.patch

netfilter: nft_set_pipapo: stop GC iteration if GC transaction allocation fails

commit 6d365eabce3c018a80f6e0379b17df2abb17405e upstream.

nft_trans_gc_queue_sync() enqueues the GC transaction and it allocates a
new one. If this allocation fails, then stop this GC sync run and retry
later.

Fixes: 5f68718b34a5 ("netfilter: nf_tables: GC transaction API to avoid race with control plane")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
Gbp-Pq: Topic bugfix/all
Gbp-Pq: Name netfilter-nft_set_pipapo-stop-gc-iteration-if-gc-tra.patch

netfilter: nft_set_pipapo: call nft_trans_gc_queue_sync() in catchall GC

commit 4a9e12ea7e70223555ec010bec9f711089ce96f6 upstream.

pipapo needs to enqueue GC transactions for catchall elements through
nft_trans_gc_queue_sync(). Add nft_trans_gc_catchall_sync() and
nft_trans_gc_catchall_async() to handle GC transaction queueing
accordingly.

Fixes: 5f68718b34a5 ("netfilter: nf_tables: GC transaction API to avoid race with control plane")
Fixes: f6c383b8c31a ("netfilter: nf_tables: adapt set backend to use GC transaction API")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
Gbp-Pq: Topic bugfix/all
Gbp-Pq: Name netfilter-nft_set_pipapo-call-nft_trans_gc_queue_syn.patch

netfilter: nft_set_rbtree: use read spinlock to avoid datapath contention

commit 96b33300fba880ec0eafcf3d82486f3463b4b6da upstream.

rbtree GC does not modify the datastructure, instead it collects expired
elements and it enqueues a GC transaction. Use a read spinlock instead
to avoid data contention while GC worker is running.

Fixes: f6c383b8c31a ("netfilter: nf_tables: adapt set backend to use GC transaction API")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
Gbp-Pq: Topic bugfix/all
Gbp-Pq: Name netfilter-nft_set_rbtree-use-read-spinlock-to-avoid-.patch

netfilter: nft_set_rbtree: skip sync GC for new elements in this transaction

commit 2ee52ae94baabf7ee09cf2a8d854b990dac5d0e4 upstream.

New elements in this transaction might expired before such transaction
ends. Skip sync GC for such elements otherwise commit path might walk
over an already released object. Once transaction is finished, async GC
will collect such expired element.

Fixes: f6c383b8c31a ("netfilter: nf_tables: adapt set backend to use GC transaction API")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
Gbp-Pq: Topic bugfix/all
Gbp-Pq: Name netfilter-nft_set_rbtree-skip-sync-gc-for-new-elemen.patch

netfilter: nf_tables: defer gc run if previous batch is still pending

commit 8e51830e29e12670b4c10df070a4ea4c9593e961 upstream.

Don't queue more gc work, else we may queue the same elements multiple
times.

If an element is flagged as dead, this can mean that either the previous
gc request was invalidated/discarded by a transaction or that the previous
request is still pending in the system work queue.

The latter will happen if the gc interval is set to a very low value,
e.g. 1ms, and system work queue is backlogged.

The sets refcount is 1 if no previous gc requeusts are queued, so add
a helper for this and skip gc run if old requests are pending.

Add a helper for this and skip the gc run in this case.

Fixes: f6c383b8c31a ("netfilter: nf_tables: adapt set backend to use GC transaction API")
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
Gbp-Pq: Topic bugfix/all
Gbp-Pq: Name netfilter-nf_tables-defer-gc-run-if-previous-batch-i.patch

netfilter: nf_tables: use correct lock to protect gc_list

commit 8357bc946a2abc2a10ca40e5a2105d2b4c57515e upstream.

Use nf_tables_gc_list_lock spinlock, not nf_tables_destroy_list_lock to
protect the gc_list.

Fixes: 5f68718b34a5 ("netfilter: nf_tables: GC transaction API to avoid race with control plane")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
Gbp-Pq: Topic bugfix/all
Gbp-Pq: Name netfilter-nf_tables-use-correct-lock-to-protect-gc_l.patch

netfilter: nf_tables: GC transaction race with abort path

commit 720344340fb9be2765bbaab7b292ece0a4570eae upstream.

Abort path is missing a synchronization point with GC transactions. Add
GC sequence number hence any GC transaction losing race will be
discarded.

Fixes: 5f68718b34a5 ("netfilter: nf_tables: GC transaction API to avoid race with control plane")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
Gbp-Pq: Topic bugfix/all
Gbp-Pq: Name netfilter-nf_tables-gc-transaction-race-with-abort-p.patch

netfilter: nf_tables: GC transaction race with netns dismantle

commit 02c6c24402bf1c1e986899c14ba22a10b510916b upstream.

Use maybe_get_net() since GC workqueue might race with netns exit path.

Fixes: 5f68718b34a5 ("netfilter: nf_tables: GC transaction API to avoid race with control plane")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
Gbp-Pq: Topic bugfix/all
Gbp-Pq: Name netfilter-nf_tables-gc-transaction-race-with-netns-d.patch

netfilter: nf_tables: fix GC transaction races with netns and netlink event exit path

commit 6a33d8b73dfac0a41f3877894b38082bd0c9a5bc upstream.

Netlink event path is missing a synchronization point with GC
transactions. Add GC sequence number update to netns release path and
netlink event path, any GC transaction losing race will be discarded.

Fixes: 5f68718b34a5 ("netfilter: nf_tables: GC transaction API to avoid race with control plane")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
Gbp-Pq: Topic bugfix/all
Gbp-Pq: Name netfilter-nf_tables-fix-gc-transaction-races-with-ne.patch

netfilter: nf_tables: don't fail inserts if duplicate has expired

commit 7845914f45f066497ac75b30c50dbc735e84e884 upstream.

nftables selftests fail:
run-tests.sh testcases/sets/0044interval_overlap_0
Expected: 0-2 . 0-3, got:
W: [FAILED]     ./testcases/sets/0044interval_overlap_0: got 1

Insertion must ignore duplicate but expired entries.

Moreover, there is a strange asymmetry in nft_pipapo_activate:

It refetches the current element, whereas the other ->activate callbacks
(bitmap, hash, rhash, rbtree) use elem->priv.
Same for .remove: other set implementations take elem->priv,
nft_pipapo_remove fetches elem->priv, then does a relookup,
remove this.

I suspect this was the reason for the change that prompted the
removal of the expired check in pipapo_get() in the first place,
but skipping exired elements there makes no sense to me, this helper
is used for normal get requests, insertions (duplicate check)
and deactivate callback.

In first two cases expired elements must be skipped.

For ->deactivate(), this gets called for DELSETELEM, so it
seems to me that expired elements should be skipped as well, i.e.
delete request should fail with -ENOENT error.

Fixes: 24138933b97b ("netfilter: nf_tables: don't skip expired elements during walk")
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
Gbp-Pq: Topic bugfix/all
Gbp-Pq: Name netfilter-nf_tables-don-t-fail-inserts-if-duplicate-.patch

netfilter: nf_tables: remove busy mark and gc batch API

commit a2dd0233cbc4d8a0abb5f64487487ffc9265beb5 upstream.

Ditch it, it has been replace it by the GC transaction API and it has no
clients anymore.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
Gbp-Pq: Topic bugfix/all
Gbp-Pq: Name netfilter-nf_tables-remove-busy-mark-and-gc-batch-ap.patch

netfilter: nft_set_hash: mark set element as dead when deleting from packet path

upstream c92db3030492b8ad1d0faace7a93bbcf53850d0c commit.

Set on the NFT_SET_ELEM_DEAD_BIT flag on this element, instead of
performing element removal which might race with an ongoing transaction.
Enable gc when dynamic flag is set on since dynset deletion requires
garbage collection after this patch.

Fixes: d0a8d877da97 ("netfilter: nft_dynset: support for element deletion")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
Gbp-Pq: Topic bugfix/all
Gbp-Pq: Name netfilter-nft_set_hash-mark-set-element-as-dead-when.patch

netfilter: nf_tables: adapt set backend to use GC transaction API

commit f6c383b8c31a93752a52697f8430a71dcbc46adf upstream.

Use the GC transaction API to replace the old and buggy gc API and the
busy mark approach.

No set elements are removed from async garbage collection anymore,
instead the _DEAD bit is set on so the set element is not visible from
lookup path anymore. Async GC enqueues transaction work that might be
aborted and retried later.

rbtree and pipapo set backends does not set on the _DEAD bit from the
sync GC path since this runs in control plane path where mutex is held.
In this case, set elements are deactivated, removed and then released
via RCU callback, sync GC never fails.

Fixes: 3c4287f62044 ("nf_tables: Add set type for arbitrary concatenation of ranges")
Fixes: 8d8540c4f5e0 ("netfilter: nft_set_rbtree: add timeout support")
Fixes: 9d0982927e79 ("netfilter: nft_hash: add support for timeouts")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
Gbp-Pq: Topic bugfix/all
Gbp-Pq: Name netfilter-nf_tables-adapt-set-backend-to-use-gc-tran.patch

netfilter: nf_tables: GC transaction API to avoid race with control plane

commit 5f68718b34a531a556f2f50300ead2862278da26 upstream.

The set types rhashtable and rbtree use a GC worker to reclaim memory.
From system work queue, in periodic intervals, a scan of the table is
done.

The major caveat here is that the nft transaction mutex is not held.
This causes a race between control plane and GC when they attempt to
delete the same element.

We cannot grab the netlink mutex from the work queue, because the
control plane has to wait for the GC work queue in case the set is to be
removed, so we get following deadlock:

   cpu 1                                cpu2
     GC work                            transaction comes in , lock nft mutex
       `acquire nft mutex // BLOCKS
                                        transaction asks to remove the set
                                        set destruction calls cancel_work_sync()

cancel_work_sync will now block forever, because it is waiting for the
mutex the caller already owns.

This patch adds a new API that deals with garbage collection in two
steps:

1) Lockless GC of expired elements sets on the NFT_SET_ELEM_DEAD_BIT
   so they are not visible via lookup. Annotate current GC sequence in
   the GC transaction. Enqueue GC transaction work as soon as it is
   full. If ruleset is updated, then GC transaction is aborted and
   retried later.

2) GC work grabs the mutex. If GC sequence has changed then this GC
   transaction lost race with control plane, abort it as it contains
   stale references to objects and let GC try again later. If the
   ruleset is intact, then this GC transaction deactivates and removes
   the elements and it uses call_rcu() to destroy elements.

Note that no elements are removed from GC lockless path, the _DEAD bit
is set and pointers are collected. GC catchall does not remove the
elements anymore too. There is a new set->dead flag that is set on to
abort the GC transaction to deal with set->ops->destroy() path which
removes the remaining elements in the set from commit_release, where no
mutex is held.

To deal with GC when mutex is held, which allows safe deactivate and
removal, add sync GC API which releases the set element object via
call_rcu(). This is used by rbtree and pipapo backends which also
perform garbage collection from control plane path.

Since element removal from sets can happen from control plane and
element garbage collection/timeout, it is necessary to keep the set
structure alive until all elements have been deactivated and destroyed.

We cannot do a cancel_work_sync or flush_work in nft_set_destroy because
its called with the transaction mutex held, but the aforementioned async
work queue might be blocked on the very mutex that nft_set_destroy()
callchain is sitting on.

This gives us the choice of ABBA deadlock or UaF.

To avoid both, add set->refs refcount_t member. The GC API can then
increment the set refcount and release it once the elements have been
free'd.

Set backends are adapted to use the GC transaction API in a follow up
patch entitled:

  ("netfilter: nf_tables: use gc transaction API in set backends")

This is joint work with Florian Westphal.

Fixes: cfed7e1b1f8e ("netfilter: nf_tables: add set garbage collection helpers")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
Gbp-Pq: Topic bugfix/all
Gbp-Pq: Name netfilter-nf_tables-gc-transaction-api-to-avoid-race.patch

netfilter: nf_tables: don't skip expired elements during walk

commit 24138933b97b055d486e8064b4a1721702442a9b upstream.

There is an asymmetry between commit/abort and preparation phase if the
following conditions are met:

1. set is a verdict map ("1.2.3.4 : jump foo")
2. timeouts are enabled

In this case, following sequence is problematic:

1. element E in set S refers to chain C
2. userspace requests removal of set S
3. kernel does a set walk to decrement chain->use count for all elements
   from preparation phase
4. kernel does another set walk to remove elements from the commit phase
   (or another walk to do a chain->use increment for all elements from
    abort phase)

If E has already expired in 1), it will be ignored during list walk, so its use count
won't have been changed.

Then, when set is culled, ->destroy callback will zap the element via
nf_tables_set_elem_destroy(), but this function is only safe for
elements that have been deactivated earlier from the preparation phase:
lack of earlier deactivate removes the element but leaks the chain use
count, which results in a WARN splat when the chain gets removed later,
plus a leak of the nft_chain structure.

Update pipapo_get() not to skip expired elements, otherwise flush
command reports bogus ENOENT errors.

Fixes: 3c4287f62044 ("nf_tables: Add set type for arbitrary concatenation of ranges")
Fixes: 8d8540c4f5e0 ("netfilter: nft_set_rbtree: add timeout support")
Fixes: 9d0982927e79 ("netfilter: nft_hash: add support for timeouts")
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
Gbp-Pq: Topic bugfix/all
Gbp-Pq: Name netfilter-nf_tables-don-t-skip-expired-elements-duri.patch

ntfs: mark it as broken

NTFS has unfixed issues CVE-2018-12929, CVE-2018-12930, and
CVE-2018-12931.  ntfs-3g is a better supported alternative.

Make sure it can't be enabled even in custom kernels.

Gbp-Pq: Topic debian
Gbp-Pq: Name ntfs-mark-it-as-broken.patch

[i386/686-pae] PCI: Set pci=nobios by default

Forwarded: not-needed

CONFIG_PCI_GOBIOS results in physical addresses 640KB-1MB being mapped
W+X, which is undesirable for security reasons and will result in a
warning at boot now that we enable CONFIG_DEBUG_WX.

This can be overridden using the kernel parameter "pci=nobios", but we
want to disable W+X by default.  Disable PCI BIOS probing by default;
it can still be enabled using "pci=bios".

Gbp-Pq: Topic debian
Gbp-Pq: Name i386-686-pae-pci-set-pci-nobios-by-default.patch

trust machine keyring (MoK) by default

Debian always trusted keys in MoK by default. Upstream made it conditional on
a new EFI variable being set. To keep backward compatibility skip this check.

Gbp-Pq: Topic features/all/db-mok-keyring
Gbp-Pq: Name trust-machine-keyring-by-default.patch

[PATCH] KEYS: Make use of platform keyring for module signature verify

Bug-Debian: https://bugs.debian.org/935945
Bug-Debian: https://bugs.debian.org/1030200
Origin: https://src.fedoraproject.org/rpms/kernel/raw/master/f/KEYS-Make-use-of-platform-keyring-for-module-signature.patch
Forwarded: https://lore.kernel.org/linux-modules/qvgp2il2co4iyxkzxvcs4p2bpyilqsbfgcprtpfrsajwae2etc@3z2s2o52i3xg/t/#u

This allows a cert in DB to be used to sign modules,
in addition to certs in the MoK and built-in keyrings.

Signed-off-by: Robert Holmes <robeholmes@gmail.com>
Signed-off-by: Jeremy Cline <jcline@redhat.com>
[bwh: Forward-ported to 5.19: adjust filename]
[наб: reinstate for 6.1, re-write description]

Gbp-Pq: Topic features/all/db-mok-keyring
Gbp-Pq: Name KEYS-Make-use-of-platform-keyring-for-module-signature.patch

[PATCH 3/4] MODSIGN: checking the blacklisted hash before loading a kernel module

Origin: https://lore.kernel.org/patchwork/patch/933175/

This patch adds the logic for checking the kernel module's hash
base on blacklist. The hash must be generated by sha256 and enrolled
to dbx/mokx.

For example:
sha256sum sample.ko
mokutil --mokx --import-hash $HASH_RESULT

Whether the signature on ko file is stripped or not, the hash can be
compared by kernel.

Cc: David Howells <dhowells@redhat.com>
Cc: Josh Boyer <jwboyer@fedoraproject.org>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Signed-off-by: "Lee, Chun-Yi" <jlee@suse.com>
[Rebased by Luca Boccassi]
[bwh: Forward-ported to 5.19:
 - The type parameter to is_hash_blacklisted() is now an enumeration
   rather than a string
 - Adjust filename, context]

Gbp-Pq: Topic features/all/db-mok-keyring
Gbp-Pq: Name 0003-MODSIGN-checking-the-blacklisted-hash-before-loading-a-kernel-module.patch

arm64: add kernel config option to lock down when in Secure Boot mode

Bug-Debian: https://bugs.debian.org/831827
Forwarded: no

Add a kernel configuration option to lock down the kernel, to restrict
userspace's ability to modify the running kernel when UEFI Secure Boot is
enabled. Based on the x86 patch by Matthew Garrett.

Determine the state of Secure Boot in the EFI stub and pass this to the
kernel using the FDT.

Signed-off-by: Linn Crosetto <linn@hpe.com>
[bwh: Forward-ported to 4.10: adjust context]
[Lukas Wunner: Forward-ported to 4.11: drop parts applied upstream]
[bwh: Forward-ported to 4.15 and lockdown patch set:
 - Pass result of efi_get_secureboot() in stub through to
   efi_set_secure_boot() in main kernel
 - Use lockdown API and naming]
[bwh: Forward-ported to 4.19.3: adjust context in update_fdt()]
[dannf: Moved init_lockdown() call after uefi_init(), fixing SB detection]
[bwh: Drop call to init_lockdown(), as efi_set_secure_boot() now calls this]
[bwh: Forward-ported to 5.6: efi_get_secureboot() no longer takes a
 sys_table parameter]
[bwh: Forward-ported to 5.7: EFI initialisation from FDT was rewritten, so:
 - Add Secure Boot mode to the parameter enumeration in fdtparams.c
 - Add a parameter to efi_get_fdt_params() to return the Secure Boot mode
 - Since Xen does not have a property name defined for Secure Boot mode,
   change efi_get_fdt_prop() to handle a missing property name by clearing
   the output variable]
[Salvatore Bonaccorso: Forward-ported to 5.10: f30f242fb131 ("efi: Rename
arm-init to efi-init common for all arch") renamed arm-init.c to efi-init.c]

Gbp-Pq: Topic features/all/lockdown
Gbp-Pq: Name arm64-add-kernel-config-option-to-lock-down-when.patch

mtd: phram,slram: Disable when the kernel is locked down

Forwarded: https://lore.kernel.org/linux-security-module/20190830154720.eekfjt6c4jzvlbfz@decadent.org.uk/

These drivers allow mapping arbitrary memory ranges as MTD devices.
This should be disabled to preserve the kernel's integrity when it is
locked down.

* Add the HWPARAM flag to the module parameters
* When slram is built-in, it uses __setup() to read kernel parameters,
  so add an explicit check security_locked_down() check

Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Cc: Matthew Garrett <mjg59@google.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Joern Engel <joern@lazybastard.org>
Cc: linux-mtd@lists.infradead.org
Gbp-Pq: Topic features/all/lockdown
Gbp-Pq: Name mtd-disable-slram-and-phram-when-locked-down.patch

efi: Lock down the kernel if booted in secure boot mode

Based on an earlier patch by David Howells, who wrote the following
description:

> UEFI Secure Boot provides a mechanism for ensuring that the firmware will
> only load signed bootloaders and kernels.  Certain use cases may also
> require that all kernel modules also be signed.  Add a configuration option
> that to lock down the kernel - which includes requiring validly signed
> modules - if the kernel is secure-booted.

Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
[Salvatore Bonaccorso: After fixing https://bugs.debian.org/956197 the
help text for LOCK_DOWN_IN_EFI_SECURE_BOOT needs to be adjusted to
mention that lockdown is triggered in integrity mode]
Signed-off-by: Salvatore Bonaccorso <carnil@debian.org>
Gbp-Pq: Topic features/all/lockdown
Gbp-Pq: Name efi-lock-down-the-kernel-if-booted-in-secure-boot-mo.patch

[28/30] efi: Add an EFI_SECURE_BOOT flag to indicate secure boot mode

Origin: https://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-fs.git/commit?id=a5d70c55c603233c192b375f72116a395909da28

UEFI machines can be booted in Secure Boot mode.  Add an EFI_SECURE_BOOT
flag that can be passed to efi_enabled() to find out whether secure boot is
enabled.

Move the switch-statement in x86's setup_arch() that inteprets the
secure_boot boot parameter to generic code and set the bit there.

Suggested-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: David Howells <dhowells@redhat.com>
Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
cc: linux-efi@vger.kernel.org
[rperier: Forward-ported to 5.5:
 - Use pr_warn()
 - Adjust context]
[bwh: Forward-ported to 5.6: adjust context]
[bwh: Forward-ported to 5.7:
 - Use the next available bit in efi.flags
 - Adjust context]

Gbp-Pq: Topic features/all/lockdown
Gbp-Pq: Name efi-add-an-efi_secure_boot-flag-to-indicate-secure-b.patch

misc: rtsx: Fix some platforms can not boot and move the l1ss judgment to probe

Origin: https://lore.kernel.org/all/37b1afb997f14946a8784c73d1f9a4f5@realtek.com/
Bug-Debian: https://bugs.debian.org/1052063

commit 101bd907b424 ("misc: rtsx: judge ASPM Mode to set PETXCFG Reg")
some readers no longer force #CLKREQ to low
when the system need to enter ASPM.
But some platform maybe not implement complete ASPM?
it causes some platforms can not boot

Like in the past only the platform support L1ss we release the #CLKREQ.
Move the judgment (L1ss) to probe,
we think read config space one time when the driver start is enough

Fixes: 101bd907b424 ("misc: rtsx: judge ASPM Mode to set PETXCFG Reg")
Signed-off-by: Ricky Wu <ricky_wu@realtek.com>
Tested-By: Jade Lovelace <lists@jade.fyi>
Gbp-Pq: Topic bugfix/all
Gbp-Pq: Name misc-rtsx-Fix-some-platforms-can-not-boot-and-move-t.patch

fs: Add MODULE_SOFTDEP declarations for hard-coded crypto drivers

Bug-Debian: https://bugs.debian.org/819725
Forwarded: http://mid.gmane.org/20160517133631.GF7555@decadent.org.uk

This helps initramfs builders and other tools to find the full
dependencies of a module.

Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
[Lukas Wunner: Forward-ported to 4.11: drop parts applied upstream]

Gbp-Pq: Topic bugfix/all
Gbp-Pq: Name fs-add-module_softdep-declarations-for-hard-coded-cr.patch

phy/marvell: disable 4-port phys

Bug-Debian: https://bugs.debian.org/723177
Forwarded: http://thread.gmane.org/gmane.linux.debian.devel.bugs.general/1107774/

The Marvell PHY was originally disabled because it can cause networking
failures on some systems. According to Lennert Buytenhek this is because some
of the variants added did not share the same register layout. Since the known
cases are all 4-ports disable those variants (indicated by a 4 in the
penultimate position of the model name) until they can be audited for
correctness.

[bwh: Also #if-out the init functions for these PHYs to avoid
 compiler warnings]

Gbp-Pq: Topic bugfix/all
Gbp-Pq: Name disable-some-marvell-phys.patch

[4/4] arm64: dts: rockchip: Add SOQuartz Model A baseboard

Origin: https://git.kernel.org/linus/afbaed737fb45bcae91e4606025fb31da71b9dfe

This patch adds the device tree for the "Model A" baseboard for
the SOQuartz CM4 SoM, which is not to be confused with the
Quartz64 Model A, which is the same form factor and SoC, but is
not a CM4 carrier board.

The board features a PCIe 2 x1 slot, USB 2 host ports, CSI/DSI
connectors, an eDP FFC connector, gigabit ethernet, HDMI, and a
12V DC barrel jack. Also present is a microSD card slot, 40-pin
GPIO, and a power and reset button.

Signed-off-by: Andrew Powers-Holmes <aholmes@omnom.net>
[rebase, misc fixes, reword]
Signed-off-by: Nicolas Frattaroli <frattaroli.nicolas@gmail.com>
Link: https://lore.kernel.org/r/20221116115337.541601-5-frattaroli.nicolas@gmail.com
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Gbp-Pq: Topic features/arm64/quartz64
Gbp-Pq: Name arm64-dts-rockchip-Add-SOQuartz-Model-A-baseboard.patch

[3/4] dt-bindings: arm: rockchip: Add SOQuartz Model A

Origin: https://git.kernel.org/linus/7441d8c437883581dddfb616a087b399338244f0

The SOQuartz Model A base board is a carrier board for the CM4
form factor, designed around the PINE64 SOQuartz CM4 SoM.

The board sports "Model A" dimensions like the Quartz64 Model A,
but is not to be confused with that.

As for I/O, it features USB 2 ports, Gigabit Ethernet, a PCIe 2
x1 slot, HDMI, a 40-pin GPIO header, CSI/DSI connectors, an eDP
flat-flex cable connector, a 12V DC barrel jack for power input
and power/reset buttons as well as a microSD card slot.

Signed-off-by: Nicolas Frattaroli <frattaroli.nicolas@gmail.com>
Acked-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Link: https://lore.kernel.org/r/20221116115337.541601-4-frattaroli.nicolas@gmail.com
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Gbp-Pq: Topic features/arm64/quartz64
Gbp-Pq: Name dt-bindings-arm-rockchip-Add-SOQuartz-Model-A.patch

[2/4] arm64: dts: rockchip: Add SOQuartz blade board

Origin: https://git.kernel.org/linus/a5c826ecde5222f755e7d8a0c8d795189c5c1228

This adds a device tree for the PINE64 SOQuartz blade baseboard,
a 1U rack mountable baseboard for the CM4 form factor with PoE
support designed for the SOQuartz CM4 System-on-Module.

The board takes power from either PoE or a 5V DC input, and allows
for mounting an M.2 SSD.

The board also features one USB 2.0 host port, one HDMI output,
a 3.5mm jack for UART, and the aforementioned gigabit networking
port.

Signed-off-by: Andrew Powers-Holmes <aholmes@omnom.net>
[rebase, squash, reword, misc fixes]
Signed-off-by: Nicolas Frattaroli <frattaroli.nicolas@gmail.com>
Link: https://lore.kernel.org/r/20221116115337.541601-3-frattaroli.nicolas@gmail.com
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Gbp-Pq: Topic features/arm64/quartz64
Gbp-Pq: Name arm64-dts-rockchip-Add-SOQuartz-blade-board.patch

[1/4] dt-bindings: arm: rockchip: Add SOQuartz Blade

Origin: https://git.kernel.org/linus/8c84c2e51f3ee39b40e8078ebe3ad9c01fb17aff

Add a compatible for the SOQuartz Blade base board to the rockchip
platforms binding.

The SOQuartz Blade is a PoE-capable carrier board for the CM4 SoM
form factor, designed around the SOQuartz CM4 System-on-Module.

The board features the usual connectivity (GPIO, USB, HDMI,
Ethernet) and an M.2 slot for SSDs. It may also be powered from
a 5V barrel jack input, and has a 3.5mm jack for UART debug
output.

Signed-off-by: Nicolas Frattaroli <frattaroli.nicolas@gmail.com>
Acked-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Link: https://lore.kernel.org/r/20221116115337.541601-2-frattaroli.nicolas@gmail.com
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Gbp-Pq: Topic features/arm64/quartz64
Gbp-Pq: Name dt-bindings-arm-rockchip-Add-SOQuartz-Blade.patch

arm64: dts: rockchip: Enable PCIe 2 on SOQuartz CM4IO

Origin: https://git.kernel.org/linus/3736aa7ecc4cd9b4abce30052bad00aba4f0362f

This patch enables the PCIe2 on the CM4IO board when paired with
a SOQuartz CM4 System-on-Module board. combphy2 also needs to be
enabled in this case to make the PHY work for this.

Signed-off-by: Nicolas Frattaroli <frattaroli.nicolas@gmail.com>
Link: https://lore.kernel.org/r/20221112160404.70868-5-frattaroli.nicolas@gmail.com
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Gbp-Pq: Topic features/arm64/quartz64
Gbp-Pq: Name arm64-dts-rockchip-Enable-PCIe-2-on-SOQuartz-CM4IO.patch

arm64: dts: rockchip: Enable HDMI sound on SOQuartz

Origin: https://git.kernel.org/linus/70b620c4ba919a87c607b8d98b08478b213877bd

This patch enables the i2s0 node on SOQuartz, which is responsible
for hdmi audio, and adds an hdmi-sound node to enable said audio.

Signed-off-by: Nicolas Frattaroli <frattaroli.nicolas@gmail.com>
Link: https://lore.kernel.org/r/20221112160404.70868-4-frattaroli.nicolas@gmail.com
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Gbp-Pq: Topic features/arm64/quartz64
Gbp-Pq: Name arm64-dts-rockchip-Enable-HDMI-sound-on-SOQuartz.patch

arm64: dts: rockchip: Enable video output and HDMI on SOQuartz

Origin: https://git.kernel.org/linus/36d7a605706d9648526a0574b8e7b0e02fa70c2a

This patch adds and enables the necessary device tree nodes to
enable video output and HDMI functionality on the SOQuartz module.

Signed-off-by: Nicolas Frattaroli <frattaroli.nicolas@gmail.com>
Link: https://lore.kernel.org/r/20221112160404.70868-3-frattaroli.nicolas@gmail.com
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Gbp-Pq: Topic features/arm64/quartz64
Gbp-Pq: Name arm64-dts-rockchip-Enable-video-output-and-HDMI-on-S.patch

arm64: dts: rockchip: RK356x: Add I2S2 device node

Origin: https://git.kernel.org/linus/755f37010f3eac0bdfa41bdf2308e8380a93f10c

This patch adds I2S2 device tree node for RK3566/RK3568.

Signed-off-by: Shengyu Qu <wiagn233@outlook.com>
Link: https://lore.kernel.org/r/OS3P286MB259771C12F2B15A4DDF435FE98359@OS3P286MB2597.JPNP286.PROD.OUTLOOK.COM
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Gbp-Pq: Topic features/arm64/quartz64
Gbp-Pq: Name arm64-dts-rockchip-RK356x-Add-I2S2-device-node.patch

arm64: dts: rockchip: fix spdif@fe460000 ordering on rk356x

Origin: https://git.kernel.org/linus/d4eade428d22f2ac5f32b12ec183fdff84dc07a6

Move the node to its correct position, based on its
mmio-address.

Link: https://lore.kernel.org/all/20221030193708.1671069-1-heiko@sntech.de
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Gbp-Pq: Topic bugfix/arm64
Gbp-Pq: Name arm64-dts-rockchip-fix-spdif-fe460000-ordering-on-rk.patch

x86: Make x32 syscall support conditional on a kernel parameter

Bug-Debian: https://bugs.debian.org/708070
Forwarded: https://lore.kernel.org/lkml/1415245982.3398.53.camel@decadent.org.uk/T/#u

Enabling x32 in the standard amd64 kernel would increase its attack
surface while provide no benefit to the vast majority of its users.
No-one seems interested in regularly checking for vulnerabilities
specific to x32 (at least no-one with a white hat).

Still, adding another flavour just to turn on x32 seems wasteful.  And
the only differences on syscall entry are a few instructions that mask
out the x32 flag and compare the syscall number.

Use a static key to control whether x32 syscalls are really enabled, a
Kconfig parameter to set its default value and a kernel parameter
"syscall.x32" to change it at boot time.

Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Gbp-Pq: Topic features/x86
Gbp-Pq: Name x86-make-x32-syscall-support-conditional.patch

x86: memtest: WARN if bad RAM found

Bug-Debian: https://bugs.debian.org/613321
Forwarded: http://thread.gmane.org/gmane.linux.kernel/1286471

Since this is not a particularly thorough test, if we find any bad
bits of RAM then there is a fair chance that there are other bad bits
we fail to detect.

Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Gbp-Pq: Topic features/x86
Gbp-Pq: Name x86-memtest-WARN-if-bad-RAM-found.patch

[13/13] arm64: dts: rockchip: Add IR receiver node to ODROID-M1

Origin: https://git.kernel.org/linus/d6882992fe8182e3122be34af3f491948a8b9069

Add the infrared receiver and its associated pinctrl entry. Note that
there is an external pullup to VCC3V3_SYS.

Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Tested-by: Dan Johansen <strit@manjaro.org>
Link: https://lore.kernel.org/r/20220930051246.391614-14-aurelien@aurel32.net
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Gbp-Pq: Topic features/arm64
Gbp-Pq: Name arm64-dts-rockchip-Add-IR-receiver-node-to-ODROID-M1.patch

[12/13] arm64: dts: rockchip: Add PCIEe v3 nodes to ODROID-M1

Origin: https://git.kernel.org/linus/35b28582aa3dfd7b6861b7ebc72798b0ff50ed41

Add nodes to ODROID-M1 to support PCIe v3 on the M2 slot.

Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Tested-by: Dan Johansen <strit@manjaro.org>
Link: https://lore.kernel.org/r/20220930051246.391614-13-aurelien@aurel32.net
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Gbp-Pq: Topic features/arm64
Gbp-Pq: Name arm64-dts-rockchip-Add-PCIEe-v3-nodes-to-ODROID-M1.patch

[11/13] arm64: dts: rockchip: Add SATA support to ODROID-M1

Origin: https://git.kernel.org/linus/6a5a04d52ccc42e0e59ff69fca9c1db7e08ba44b

Enable the Combo PHY and SATA nodes in ODROID-M1.

Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Tested-by: Dan Johansen <strit@manjaro.org>
Link: https://lore.kernel.org/r/20220930051246.391614-12-aurelien@aurel32.net
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Gbp-Pq: Topic features/arm64
Gbp-Pq: Name arm64-dts-rockchip-Add-SATA-support-to-ODROID-M1.patch

[10/13] arm64: dts: rockchip: Enable the USB 3.0 ports on ODROID-M1

Origin: https://git.kernel.org/linus/9984ef562653c8d0beb51021fc286706b6ec4802

The Rockchip RK3568 has two USB XHCI controllers. The USB 2.0 signals
are connected to a PHY providing one host-only port and one OTG port.
The USB 3.0 signals are connected to two USB3.0/PCIE/SATA combo PHY.

The ODROID M1 has 2 type A USB 3.0 connectors, with the USB 3.0 signals
connected to the two combo PHYs. For the USB 2.0 signals, one connector
is connected to the host-only PHY and uses the same power switch as the
USB 2.0 ports. The other connector has its own power switch and is
connected to the OTG PHY, which is also connected to a device only
micro-USB connector. The purpose of this micro-USB connector is for
firmware update using the Rockusb vendor specific USB class. Therefore
it does not make sense to enable this port on Linux, and the PHY is
forced to host mode.

Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Tested-by: Dan Johansen <strit@manjaro.org>
Link: https://lore.kernel.org/r/20220930051246.391614-11-aurelien@aurel32.net
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Gbp-Pq: Topic features/arm64
Gbp-Pq: Name arm64-dts-rockchip-Enable-the-USB-3.0-ports-on-ODROI.patch

[09/13] arm64: dts: rockchip: Enable the USB 2.0 ports on ODROID-M1

Origin: https://git.kernel.org/linus/4685d7b68aaac199ab0d950d2047405bf551f964

The Rockchip RK3568 has two USB OHCI/EHCI controllers connected to a PHY
providing one host-only port and one OTG port. On the ODROID-M1, they
are both used in host mode. The USB ports are powered by a DC/DC
converter providing 5V and named VCC5V0_SYS on the schematics, followed
by a power switch.

Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Tested-by: Dan Johansen <strit@manjaro.org>
Link: https://lore.kernel.org/r/20220930051246.391614-10-aurelien@aurel32.net
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Gbp-Pq: Topic features/arm64
Gbp-Pq: Name arm64-dts-rockchip-Enable-the-USB-2.0-ports-on-ODROI.patch

[08/13] arm64: dts: rockchip: Enable the GPU on ODROID-M1

Origin: https://git.kernel.org/linus/cb80b3455c7cadc4c1157879930e919f607d557c

Enable the GPU core on the Rockchip RK3568 ODROID-M1.

Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Tested-by: Dan Johansen <strit@manjaro.org>
Link: https://lore.kernel.org/r/20220930051246.391614-9-aurelien@aurel32.net
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Gbp-Pq: Topic features/arm64
Gbp-Pq: Name arm64-dts-rockchip-Enable-the-GPU-on-ODROID-M1.patch

[07/13] arm64: dts: rockchip: Enable HDMI audio on ODROID-M1.

Origin: https://git.kernel.org/linus/1ca7ddddf36494f0f6afd4f35d37827323271f39

This enables the i2s0 controller and the hdmi-sound node on the
ODROID-M1.

Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Tested-by: Dan Johansen <strit@manjaro.org>
Link: https://lore.kernel.org/r/20220930051246.391614-8-aurelien@aurel32.net
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Gbp-Pq: Topic features/arm64
Gbp-Pq: Name arm64-dts-rockchip-Enable-HDMI-audio-on-ODROID-M1.patch

[06/13] arm64: dts: rockchip: Enable vop2 and hdmi tx on ODROID-M1

Origin: https://git.kernel.org/linus/913404aa2e60610f9cae375069dae97e11d726ed

Enable the RK356x Video Output Processor (VOP) 2 on ODROID M1.

Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Tested-by: Dan Johansen <strit@manjaro.org>
Link: https://lore.kernel.org/r/20220930051246.391614-7-aurelien@aurel32.net
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Gbp-Pq: Topic features/arm64
Gbp-Pq: Name arm64-dts-rockchip-Enable-vop2-and-hdmi-tx-on-ODROID.patch

[05/13] arm64: dts: rockchip: Add analog audio on ODROID-M1

Origin: https://git.kernel.org/linus/78f858447cb78cac7259093d095fb783328b835c

On the ODROID-M1, the I2S1 TDM controller is connected to the rk809
codec in I2S mode. It is used to provide a stereo headphones output and
a mono speaker output. A GPIO with an external pullup is used as an
headphone detection input.

Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Tested-by: Dan Johansen <strit@manjaro.org>
Link: https://lore.kernel.org/r/20220930051246.391614-6-aurelien@aurel32.net
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Gbp-Pq: Topic features/arm64
Gbp-Pq: Name arm64-dts-rockchip-Add-analog-audio-on-ODROID-M1.patch