Merge branch 'mlxsw-various-fixes'

Ido Schimmel says:

====================
mlxsw: Various fixes

Patch #1 fixes wrong invocation of mausezahn in a couple of selftests.
The tests started failing after Fedora updated their libnet package from
version 1.1.6 to 1.2.1. With the fix the tests pass regardless of libnet
version.

Patch #2 fixes an issue in the mirroring to CPU code that results in
policer configuration being overwritten.
====================

Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

mlxsw: spectrum_span: Do not overwrite policer configuration

The purpose of the delayed work in the SPAN module is to potentially
update the destination port and various encapsulation parameters of SPAN
agents that point to a VLAN device or a GRE tap. The destination port
can change following the insertion of a new route, for example.

SPAN agents that point to a physical port or the CPU port are static and
never change throughout the lifetime of the SPAN agent. Therefore, skip
over them in the delayed work.

This fixes an issue where the delayed work overwrites the policer
that was set on a SPAN agent pointing to the CPU. Modifying the delayed
work to inherit the original policer configuration is error-prone, as
the same will be needed for any new parameter.

Fixes: 4039504e6a0c ("mlxsw: spectrum_span: Allow setting policer on a SPAN agent")
Reviewed-by: Petr Machata <[email protected]>
Signed-off-by: Ido Schimmel <[email protected]>
Signed-off-by: Jakub Kicinski <[email protected]>

selftests: forwarding: Specify interface when invoking mausezahn

Specify the interface through which packets should be transmitted so
that the test will pass regardless of the libnet version against which
mausezahn is linked.

Fixes: cab14d1087d9 ("selftests: Add version of router_multipath.sh using nexthop objects")
Fixes: 3d578d879517 ("selftests: forwarding: Test IPv4 weighted nexthops")
Signed-off-by: Danielle Ratson <[email protected]>
Signed-off-by: Ido Schimmel <[email protected]>
Signed-off-by: Jakub Kicinski <[email protected]>

stmmac: intel: Configure EHL PSE0 GbE and PSE1 GbE to 32 bits DMA addressing

Fix an issue where dump stack is printed and Reset Adapter occurs when
PSE0 GbE or/and PSE1 GbE is/are enabled. EHL PSE0 GbE and PSE1 GbE use
32 bits DMA addressing whereas EHL PCH GbE uses 64 bits DMA addressing.

[   25.535095] ------------[ cut here ]------------
[   25.540276] NETDEV WATCHDOG: enp0s29f2 (intel-eth-pci): transmit queue 2 timed out
[   25.548749] WARNING: CPU: 2 PID: 0 at net/sched/sch_generic.c:443 dev_watchdog+0x259/0x260
[   25.558004] Modules linked in: 8021q bnep bluetooth ecryptfs snd_hda_codec_hdmi intel_gpy marvell intel_ishtp_loader intel_ishtp_hid iTCO_wdt mei_hdcp iTCO_vendor_support x86_pkg_temp_thermal kvm_intel dwmac_intel stmmac kvm igb pcs_xpcs irqbypass phylink snd_hda_intel intel_rapl_msr pcspkr dca snd_hda_codec i915 i2c_i801 i2c_smbus libphy intel_ish_ipc snd_hda_core mei_me intel_ishtp mei spi_dw_pci 8250_lpss spi_dw thermal dw_dmac_core parport_pc tpm_crb tpm_tis parport tpm_tis_core tpm intel_pmc_core sch_fq_codel uhid fuse configfs snd_sof_pci snd_sof_intel_byt snd_sof_intel_ipc snd_sof_intel_hda_common snd_sof_xtensa_dsp snd_sof snd_soc_acpi_intel_match snd_soc_acpi snd_intel_dspcfg ledtrig_audio snd_soc_core snd_compress ac97_bus snd_pcm snd_timer snd soundcore
[   25.633795] CPU: 2 PID: 0 Comm: swapper/2 Tainted: G     U            5.11.0-rc4-intel-lts-MISMAIL5+ #5
[   25.644306] Hardware name: Intel Corporation Elkhart Lake Embedded Platform/ElkhartLake LPDDR4x T4 RVP1, BIOS EHLSFWI1.R00.2434.A00.2010231402 10/23/2020
[   25.659674] RIP: 0010:dev_watchdog+0x259/0x260
[   25.664650] Code: e8 3b 6b 60 ff eb 98 4c 89 ef c6 05 ec e7 bf 00 01 e8 fb e5 fa ff 89 d9 4c 89 ee 48 c7 c7 78 31 d2 9e 48 89 c2 e8 79 1b 18 00 <0f> 0b e9 77 ff ff ff 0f 1f 44 00 00 48 c7 47 08 00 00 00 00 48 c7
[   25.685647] RSP: 0018:ffffb7ca80160eb8 EFLAGS: 00010286
[   25.691498] RAX: 0000000000000000 RBX: 0000000000000002 RCX: 0000000000000103
[   25.699483] RDX: 0000000080000103 RSI: 00000000000000f6 RDI: 00000000ffffffff
[   25.707465] RBP: ffff985709ce0440 R08: 0000000000000000 R09: c0000000ffffefff
[   25.715455] R10: ffffb7ca80160cf0 R11: ffffb7ca80160ce8 R12: ffff985709ce039c
[   25.723438] R13: ffff985709ce0000 R14: 0000000000000008 R15: ffff9857068af940
[   25.731425] FS:  0000000000000000(0000) GS:ffff985864300000(0000) knlGS:0000000000000000
[   25.740481] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[   25.746913] CR2: 00005567f8bb76b8 CR3: 00000001f8e0a000 CR4: 0000000000350ee0
[   25.754900] Call Trace:
[   25.757631]  <IRQ>
[   25.759891]  ? qdisc_put_unlocked+0x30/0x30
[   25.764565]  ? qdisc_put_unlocked+0x30/0x30
[   25.769245]  call_timer_fn+0x2e/0x140
[   25.773346]  run_timer_softirq+0x1f3/0x430
[   25.777932]  ? __hrtimer_run_queues+0x12c/0x2c0
[   25.783005]  ? ktime_get+0x3e/0xa0
[   25.786812]  __do_softirq+0xa6/0x2ef
[   25.790816]  asm_call_irq_on_stack+0xf/0x20
[   25.795501]  </IRQ>
[   25.797852]  do_softirq_own_stack+0x5d/0x80
[   25.802538]  irq_exit_rcu+0x94/0xb0
[   25.806475]  sysvec_apic_timer_interrupt+0x42/0xc0
[   25.811836]  asm_sysvec_apic_timer_interrupt+0x12/0x20
[   25.817586] RIP: 0010:cpuidle_enter_state+0xd9/0x370
[   25.823142] Code: 85 c0 0f 8f 0a 02 00 00 31 ff e8 22 d5 7e ff 45 84 ff 74 12 9c 58 f6 c4 02 0f 85 47 02 00 00 31 ff e8 7b a0 84 ff fb 45 85 f6 <0f> 88 ab 00 00 00 49 63 ce 48 2b 2c 24 48 89 c8 48 6b d1 68 48 c1
[   25.844140] RSP: 0018:ffffb7ca800f7e80 EFLAGS: 00000206
[   25.849996] RAX: ffff985864300000 RBX: 0000000000000003 RCX: 000000000000001f
[   25.857975] RDX: 00000005f2028ea8 RSI: ffffffff9ec5907f RDI: ffffffff9ec62a5d
[   25.865961] RBP: 00000005f2028ea8 R08: 0000000000000000 R09: 0000000000029d00
[   25.873947] R10: 000000137b0e0508 R11: ffff9858643294e4 R12: ffff9858643336d0
[   25.881935] R13: ffffffff9ef74b00 R14: 0000000000000003 R15: 0000000000000000
[   25.889918]  cpuidle_enter+0x29/0x40
[   25.893922]  do_idle+0x24a/0x290
[   25.897536]  cpu_startup_entry+0x19/0x20
[   25.901930]  start_secondary+0x128/0x160
[   25.906326]  secondary_startup_64_no_verify+0xb0/0xbb
[   25.911983] ---[ end trace b4c0c8195d0ba61f ]---
[   25.917193] intel-eth-pci 0000:00:1d.2 enp0s29f2: Reset adapter.

Fixes: 67c08ac4140a ("net: stmmac: add EHL PSE0 & PSE1 1Gbps PCI info and PCI ID")
Signed-off-by: Voon Weifeng <[email protected]>
Co-developed-by: Mohammad Athari Bin Ismail <[email protected]>
Signed-off-by: Mohammad Athari Bin Ismail <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

net: usb: cdc_ether: added support for Thales Cinterion PLSx3 modem family.

lsusb -v for this device:

Bus 003 Device 007: ID 1e2d:0069
Device Descriptor:
  bLength                18
  bDescriptorType         1
  bcdUSB               2.00
  bDeviceClass          239 Miscellaneous Device
  bDeviceSubClass         2 ?
  bDeviceProtocol         1 Interface Association
  bMaxPacketSize0        64
  idVendor           0x1e2d
  idProduct          0x0069
  bcdDevice            0.00
  iManufacturer           4 Cinterion Wireless Modules
  iProduct                3 PLSx3
  iSerial                 5 fa3c1419
  bNumConfigurations      1
  Configuration Descriptor:
    bLength                 9
    bDescriptorType         2
    wTotalLength          352
    bNumInterfaces         10
    bConfigurationValue     1
    iConfiguration          2 Cinterion Configuration
    bmAttributes         0xe0
      Self Powered
      Remote Wakeup
    MaxPower              500mA
    Interface Association:
      bLength                 8
      bDescriptorType        11
      bFirstInterface         0
      bInterfaceCount         2
      bFunctionClass          2 Communications
      bFunctionSubClass       2 Abstract (modem)
      bFunctionProtocol       1 AT-commands (v.25ter)
      iFunction               0
    Interface Descriptor:
      bLength                 9
      bDescriptorType         4
      bInterfaceNumber        0
      bAlternateSetting       0
      bNumEndpoints           1
      bInterfaceClass         2 Communications
      bInterfaceSubClass      2 Abstract (modem)
      bInterfaceProtocol      1 AT-commands (v.25ter)
      iInterface              0
      CDC Header:
        bcdCDC               1.10
      CDC ACM:
        bmCapabilities       0x02
          line coding and serial state
      CDC Call Management:
        bmCapabilities       0x03
          call management
          use DataInterface
        bDataInterface          1
      CDC Union:
        bMasterInterface        0
        bSlaveInterface         1
      Endpoint Descriptor:
        bLength                 7
        bDescriptorType         5
        bEndpointAddress     0x81  EP 1 IN
        bmAttributes            3
          Transfer Type            Interrupt
          Synch Type               None
          Usage Type               Data
        wMaxPacketSize     0x0040  1x 64 bytes
        bInterval               5
    Interface Descriptor:
      bLength                 9
      bDescriptorType         4
      bInterfaceNumber        1
      bAlternateSetting       0
      bNumEndpoints           2
      bInterfaceClass        10 CDC Data
      bInterfaceSubClass      0 Unused
      bInterfaceProtocol      0
      iInterface              0
      Endpoint Descriptor:
        bLength                 7
        bDescriptorType         5
        bEndpointAddress     0x82  EP 2 IN
        bmAttributes            2
          Transfer Type            Bulk
          Synch Type               None
          Usage Type               Data
        wMaxPacketSize     0x0200  1x 512 bytes
        bInterval               0
      Endpoint Descriptor:
        bLength                 7
        bDescriptorType         5
        bEndpointAddress     0x01  EP 1 OUT
        bmAttributes            2
          Transfer Type            Bulk
          Synch Type               None
          Usage Type               Data
        wMaxPacketSize     0x0200  1x 512 bytes
        bInterval               0
    Interface Association:
      bLength                 8
      bDescriptorType        11
      bFirstInterface         2
      bInterfaceCount         2
      bFunctionClass          2 Communications
      bFunctionSubClass       2 Abstract (modem)
      bFunctionProtocol       1 AT-commands (v.25ter)
      iFunction               0
    Interface Descriptor:
      bLength                 9
      bDescriptorType         4
      bInterfaceNumber        2
      bAlternateSetting       0
      bNumEndpoints           1
      bInterfaceClass         2 Communications
      bInterfaceSubClass      2 Abstract (modem)
      bInterfaceProtocol      1 AT-commands (v.25ter)
      iInterface              0
      CDC Header:
        bcdCDC               1.10
      CDC ACM:
        bmCapabilities       0x02
          line coding and serial state
      CDC Call Management:
        bmCapabilities       0x03
          call management
          use DataInterface
        bDataInterface          3
      CDC Union:
        bMasterInterface        2
        bSlaveInterface         3
      Endpoint Descriptor:
        bLength                 7
        bDescriptorType         5
        bEndpointAddress     0x83  EP 3 IN
        bmAttributes            3
          Transfer Type            Interrupt
          Synch Type               None
          Usage Type               Data
        wMaxPacketSize     0x0040  1x 64 bytes
        bInterval               5
    Interface Descriptor:
      bLength                 9
      bDescriptorType         4
      bInterfaceNumber        3
      bAlternateSetting       0
      bNumEndpoints           2
      bInterfaceClass        10 CDC Data
      bInterfaceSubClass      0 Unused
      bInterfaceProtocol      0
      iInterface              0
      Endpoint Descriptor:
        bLength                 7
        bDescriptorType         5
        bEndpointAddress     0x84  EP 4 IN
        bmAttributes            2
          Transfer Type            Bulk
          Synch Type               None
          Usage Type               Data
        wMaxPacketSize     0x0200  1x 512 bytes
        bInterval               0
      Endpoint Descriptor:
        bLength                 7
        bDescriptorType         5
        bEndpointAddress     0x02  EP 2 OUT
        bmAttributes            2
          Transfer Type            Bulk
          Synch Type               None
          Usage Type               Data
        wMaxPacketSize     0x0200  1x 512 bytes
        bInterval               0
    Interface Association:
      bLength                 8
      bDescriptorType        11
      bFirstInterface         4
      bInterfaceCount         2
      bFunctionClass          2 Communications
      bFunctionSubClass       2 Abstract (modem)
      bFunctionProtocol       1 AT-commands (v.25ter)
      iFunction               0
    Interface Descriptor:
      bLength                 9
      bDescriptorType         4
      bInterfaceNumber        4
      bAlternateSetting       0
      bNumEndpoints           1
      bInterfaceClass         2 Communications
      bInterfaceSubClass      2 Abstract (modem)
      bInterfaceProtocol      1 AT-commands (v.25ter)
      iInterface              0
      CDC Header:
        bcdCDC               1.10
      CDC ACM:
        bmCapabilities       0x02
          line coding and serial state
      CDC Call Management:
        bmCapabilities       0x03
          call management
          use DataInterface
        bDataInterface          5
      CDC Union:
        bMasterInterface        4
        bSlaveInterface         5
      Endpoint Descriptor:
        bLength                 7
        bDescriptorType         5
        bEndpointAddress     0x85  EP 5 IN
        bmAttributes            3
          Transfer Type            Interrupt
          Synch Type               None
          Usage Type               Data
        wMaxPacketSize     0x0040  1x 64 bytes
        bInterval               5
    Interface Descriptor:
      bLength                 9
      bDescriptorType         4
      bInterfaceNumber        5
      bAlternateSetting       0
      bNumEndpoints           2
      bInterfaceClass        10 CDC Data
      bInterfaceSubClass      0 Unused
      bInterfaceProtocol      0
      iInterface              0
      Endpoint Descriptor:
        bLength                 7
        bDescriptorType         5
        bEndpointAddress     0x86  EP 6 IN
        bmAttributes            2
          Transfer Type            Bulk
          Synch Type               None
          Usage Type               Data
        wMaxPacketSize     0x0200  1x 512 bytes
        bInterval               0
      Endpoint Descriptor:
        bLength                 7
        bDescriptorType         5
        bEndpointAddress     0x03  EP 3 OUT
        bmAttributes            2
          Transfer Type            Bulk
          Synch Type               None
          Usage Type               Data
        wMaxPacketSize     0x0200  1x 512 bytes
        bInterval               0
    Interface Association:
      bLength                 8
      bDescriptorType        11
      bFirstInterface         6
      bInterfaceCount         2
      bFunctionClass          2 Communications
      bFunctionSubClass       2 Abstract (modem)
      bFunctionProtocol       1 AT-commands (v.25ter)
      iFunction               0
    Interface Descriptor:
      bLength                 9
      bDescriptorType         4
      bInterfaceNumber        6
      bAlternateSetting       0
      bNumEndpoints           1
      bInterfaceClass         2 Communications
      bInterfaceSubClass      2 Abstract (modem)
      bInterfaceProtocol      1 AT-commands (v.25ter)
      iInterface              0
      CDC Header:
        bcdCDC               1.10
      CDC ACM:
        bmCapabilities       0x02
          line coding and serial state
      CDC Call Management:
        bmCapabilities       0x03
          call management
          use DataInterface
        bDataInterface          7
      CDC Union:
        bMasterInterface        6
        bSlaveInterface         7
      Endpoint Descriptor:
        bLength                 7
        bDescriptorType         5
        bEndpointAddress     0x87  EP 7 IN
        bmAttributes            3
          Transfer Type            Interrupt
          Synch Type               None
          Usage Type               Data
        wMaxPacketSize     0x0040  1x 64 bytes
        bInterval               5
    Interface Descriptor:
      bLength                 9
      bDescriptorType         4
      bInterfaceNumber        7
      bAlternateSetting       0
      bNumEndpoints           2
      bInterfaceClass        10 CDC Data
      bInterfaceSubClass      0 Unused
      bInterfaceProtocol      0
      iInterface              0
      Endpoint Descriptor:
        bLength                 7
        bDescriptorType         5
        bEndpointAddress     0x88  EP 8 IN
        bmAttributes            2
          Transfer Type            Bulk
          Synch Type               None
          Usage Type               Data
        wMaxPacketSize     0x0200  1x 512 bytes
        bInterval               0
      Endpoint Descriptor:
        bLength                 7
        bDescriptorType         5
        bEndpointAddress     0x04  EP 4 OUT
        bmAttributes            2
          Transfer Type            Bulk
          Synch Type               None
          Usage Type               Data
        wMaxPacketSize     0x0200  1x 512 bytes
        bInterval               0
    Interface Association:
      bLength                 8
      bDescriptorType        11
      bFirstInterface         8
      bInterfaceCount         2
      bFunctionClass          2 Communications
      bFunctionSubClass       0
      bFunctionProtocol       0
      iFunction               0
    Interface Descriptor:
      bLength                 9
      bDescriptorType         4
      bInterfaceNumber        8
      bAlternateSetting       0
      bNumEndpoints           1
      bInterfaceClass         2 Communications
      bInterfaceSubClass      6 Ethernet Networking
      bInterfaceProtocol      0
      iInterface              0
      CDC Header:
        bcdCDC               1.10
      CDC Ethernet:
        iMacAddress                      1 00A0C6C14190
        bmEthernetStatistics    0x00000000
        wMaxSegmentSize              16384
        wNumberMCFilters            0x0001
        bNumberPowerFilters              0
      CDC Union:
        bMasterInterface        8
        bSlaveInterface         9
      Endpoint Descriptor:
        bLength                 7
        bDescriptorType         5
        bEndpointAddress     0x89  EP 9 IN
        bmAttributes            3
          Transfer Type            Interrupt
          Synch Type               None
          Usage Type               Data
        wMaxPacketSize     0x0040  1x 64 bytes
        bInterval               5
    Interface Descriptor:
      bLength                 9
      bDescriptorType         4
      bInterfaceNumber        9
      bAlternateSetting       0
      bNumEndpoints           0
      bInterfaceClass        10 CDC Data
      bInterfaceSubClass      0 Unused
      bInterfaceProtocol      0
      iInterface              0
    Interface Descriptor:
      bLength                 9
      bDescriptorType         4
      bInterfaceNumber        9
      bAlternateSetting       1
      bNumEndpoints           2
      bInterfaceClass        10 CDC Data
      bInterfaceSubClass      0 Unused
      bInterfaceProtocol      0
      iInterface              0
      Endpoint Descriptor:
        bLength                 7
        bDescriptorType         5
        bEndpointAddress     0x8a  EP 10 IN
        bmAttributes            2
          Transfer Type            Bulk
          Synch Type               None
          Usage Type               Data
        wMaxPacketSize     0x0200  1x 512 bytes
        bInterval               0
      Endpoint Descriptor:
        bLength                 7
        bDescriptorType         5
        bEndpointAddress     0x05  EP 5 OUT
        bmAttributes            2
          Transfer Type            Bulk
          Synch Type               None
          Usage Type               Data
        wMaxPacketSize     0x0200  1x 512 bytes
        bInterval               0
Device Qualifier (for other device speed):
  bLength                10
  bDescriptorType         6
  bcdUSB               2.00
  bDeviceClass          239 Miscellaneous Device
  bDeviceSubClass         2 ?
  bDeviceProtocol         1 Interface Association
  bMaxPacketSize0        64
  bNumConfigurations      1
Device Status:     0x0000
  (Bus Powered)

Signed-off-by: Giacinto Cifelli <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

Merge tag 'locking-urgent-2021-01-28' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull locking fixes from Thomas Gleixner:
 "A set of PI futex fixes:

   - Address a longstanding issue where the user space part of the PI
     futex is not writeable. The kernel returns with inconsistent state
     which can in the worst case result in a UAF of a tasks kernel
     stack.

     The solution is to establish consistent kernel state which makes
     future operations on the futex fail because user space and kernel
     space state are inconsistent. Not a problem as PI futexes
     fundamentaly require a functional RW mapping and if user space
     pulls the rug under it, then it can keep the pieces it asked for.

   - Address an issue where the return value is incorrect in case that
     the futex was acquired after a timeout/signal made the waiter drop
     out of the rtmutex wait.

     In one of the corner cases the kernel returned an error code
     despite having successfully acquired the futex"

* tag 'locking-urgent-2021-01-28' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  futex: Handle faults correctly for PI futexes
  futex: Simplify fixup_pi_state_owner()
  futex: Use pi_state_update_owner() in put_pi_state()
  rtmutex: Remove unused argument from rt_mutex_proxy_unlock()
  futex: Provide and use pi_state_update_owner()
  futex: Replace pointless printk in fixup_owner()
  futex: Ensure the correct return value from futex_lock_pi()

Merge tag 'nvme-5.11-2021-01-28' of git://git.infradead.org/nvme into block-5.11

Pull NVMe fixes from Christoph:

"nvme fixes for 5.11:

 - add another Write Zeroes quirk (Chaitanya Kulkarni)
 - handle a no path available corner case (Daniel Wagner)
 - use the proper RCU aware list_add helper (Chao Leng)"

* tag 'nvme-5.11-2021-01-28' of git://git.infradead.org/nvme:
  nvme-core: use list_add_tail_rcu instead of list_add_tail for nvme_init_ns_head
  nvme-multipath: Early exit if no path is available
  nvme-pci: add the DISABLE_WRITE_ZEROES quirk for a SPCC device

io_uring: fix sqo ownership false positive warning

WARNING: CPU: 0 PID: 21359 at fs/io_uring.c:9042
    io_uring_cancel_task_requests+0xe55/0x10c0 fs/io_uring.c:9042
Call Trace:
 io_uring_flush+0x47b/0x6e0 fs/io_uring.c:9227
 filp_close+0xb4/0x170 fs/open.c:1295
 close_files fs/file.c:403 [inline]
 put_files_struct fs/file.c:418 [inline]
 put_files_struct+0x1cc/0x350 fs/file.c:415
 exit_files+0x7e/0xa0 fs/file.c:435
 do_exit+0xc22/0x2ae0 kernel/exit.c:820
 do_group_exit+0x125/0x310 kernel/exit.c:922
 get_signal+0x427/0x20f0 kernel/signal.c:2773
 arch_do_signal_or_restart+0x2a8/0x1eb0 arch/x86/kernel/signal.c:811
 handle_signal_work kernel/entry/common.c:147 [inline]
 exit_to_user_mode_loop kernel/entry/common.c:171 [inline]
 exit_to_user_mode_prepare+0x148/0x250 kernel/entry/common.c:201
 __syscall_exit_to_user_mode_work kernel/entry/common.c:291 [inline]
 syscall_exit_to_user_mode+0x19/0x50 kernel/entry/common.c:302
 entry_SYSCALL_64_after_hwframe+0x44/0xa9

Now io_uring_cancel_task_requests() can be called not through file
notes but directly, remove a WARN_ONCE() there that give us false
positives. That check is not very important and we catch it in other
places.

Fixes: 84965ff8a84f0 ("io_uring: if we see flush on exit, cancel related tasks")
Cc: [email protected] # 5.9+
Reported-by: [email protected]
Signed-off-by: Pavel Begunkov <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>

io_uring: fix list corruption for splice file_get

kernel BUG at lib/list_debug.c:29!
Call Trace:
 __list_add include/linux/list.h:67 [inline]
 list_add include/linux/list.h:86 [inline]
 io_file_get+0x8cc/0xdb0 fs/io_uring.c:6466
 __io_splice_prep+0x1bc/0x530 fs/io_uring.c:3866
 io_splice_prep fs/io_uring.c:3920 [inline]
 io_req_prep+0x3546/0x4e80 fs/io_uring.c:6081
 io_queue_sqe+0x609/0x10d0 fs/io_uring.c:6628
 io_submit_sqe fs/io_uring.c:6705 [inline]
 io_submit_sqes+0x1495/0x2720 fs/io_uring.c:6953
 __do_sys_io_uring_enter+0x107d/0x1f30 fs/io_uring.c:9353
 do_syscall_64+0x2d/0x70 arch/x86/entry/common.c:46
 entry_SYSCALL_64_after_hwframe+0x44/0xa9

io_file_get() may be called from splice, and so REQ_F_INFLIGHT may
already be set.

Fixes: 02a13674fa0e8 ("io_uring: account io_uring internal files as REQ_F_INFLIGHT")
Cc: [email protected] # 5.9+
Reported-by: [email protected]
Signed-off-by: Pavel Begunkov <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>

amdgpu: fix clang build warning

clang warns about the -mhard-float command line arguments
on architectures that do not support this:

clang: error: argument unused during compilation: '-mhard-float' [-Werror,-Wunused-command-line-argument]

Move this into the gcc-specific arguments.

Fixes: e77165bf7b02 ("drm/amd/display: Add DCN3 blocks to Makefile")
Signed-off-by: Arnd Bergmann <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

Revert "drm/amdgpu/swsmu: drop set_fan_speed_percent (v2)"

On some boards the rpm interface apparently does not work at all
leading to the fan not spinning or spinning at strange speeds.
Revert this for now to fix 5.10, 5.11.  The follow on patch
fixes this properly for 5.12.

This reverts commit 8d6e65adc25e23fabbc5293b6cd320195c708dca.

Bug: https://gitlab.freedesktop.org/drm/amd/-/issues/1408
Signed-off-by: Alex Deucher <[email protected]>
Cc: [email protected]

nvme-core: use list_add_tail_rcu instead of list_add_tail for nvme_init_ns_head

The "list" of nvme_ns_head is used as rcu list, now in nvme_init_ns_head
list_add_tail is used to add ns->siblings to the rcu list. It is not safe.
Should use list_add_tail_rcu instead of list_add_tail.

Signed-off-by: Chao Leng <[email protected]>
Signed-off-by: Christoph Hellwig <[email protected]>

nvme-multipath: Early exit if no path is available

nvme_round_robin_path() should test if the return ns pointer is valid.
nvme_next_ns() will return a NULL pointer if there is no path left.

Fixes: 75c10e732724 ("nvme-multipath: round-robin I/O policy")
Signed-off-by: Daniel Wagner <[email protected]>
Reviewed-by: Hannes Reinecke <[email protected]>
Signed-off-by: Christoph Hellwig <[email protected]>

nvme-pci: add the DISABLE_WRITE_ZEROES quirk for a SPCC device

This adds a quirk for SPCC 256GB NVMe 1.3 drive which fixes timeouts and
I/O errors due to the fact that the controller does not properly
handle the Write Zeroes command:

[ 2745.659527] CPU: 2 PID: 0 Comm: swapper/2 Tainted: G            E 5.10.6-BET #1
[ 2745.659528] Hardware name: System manufacturer System Product Name/PRIME X570-P, BIOS 3001 12/04/2020
[ 2776.138874] nvme nvme1: I/O 414 QID 3 timeout, aborting
[ 2776.138886] nvme nvme1: I/O 415 QID 3 timeout, aborting
[ 2776.138891] nvme nvme1: I/O 416 QID 3 timeout, aborting
[ 2776.138895] nvme nvme1: I/O 417 QID 3 timeout, aborting
[ 2776.138912] nvme nvme1: Abort status: 0x0
[ 2776.138921] nvme nvme1: I/O 428 QID 3 timeout, aborting
[ 2776.138922] nvme nvme1: Abort status: 0x0
[ 2776.138925] nvme nvme1: Abort status: 0x0
[ 2776.138974] nvme nvme1: Abort status: 0x0
[ 2776.138977] nvme nvme1: Abort status: 0x0
[ 2806.346792] nvme nvme1: I/O 414 QID 3 timeout, reset controller
[ 2806.363566] nvme nvme1: 15/0/0 default/read/poll queues
[ 2836.554298] nvme nvme1: I/O 415 QID 3 timeout, disable controller
[ 2836.672064] blk_update_request: I/O error, dev nvme1n1, sector 16350 op 0x9:(WRITE_ZEROES) flags 0x0 phys_seg 0 prio class 0
[ 2836.672072] blk_update_request: I/O error, dev nvme1n1, sector 16093 op 0x9:(WRITE_ZEROES) flags 0x0 phys_seg 0 prio class 0
[ 2836.672074] blk_update_request: I/O error, dev nvme1n1, sector 15836 op 0x9:(WRITE_ZEROES) flags 0x0 phys_seg 0 prio class 0
[ 2836.672076] blk_update_request: I/O error, dev nvme1n1, sector 15579 op 0x9:(WRITE_ZEROES) flags 0x0 phys_seg 0 prio class 0
[ 2836.672078] blk_update_request: I/O error, dev nvme1n1, sector 15322 op 0x9:(WRITE_ZEROES) flags 0x0 phys_seg 0 prio class 0
[ 2836.672080] blk_update_request: I/O error, dev nvme1n1, sector 15065 op 0x9:(WRITE_ZEROES) flags 0x0 phys_seg 0 prio class 0
[ 2836.672082] blk_update_request: I/O error, dev nvme1n1, sector 14808 op 0x9:(WRITE_ZEROES) flags 0x0 phys_seg 0 prio class 0
[ 2836.672083] blk_update_request: I/O error, dev nvme1n1, sector 14551 op 0x9:(WRITE_ZEROES) flags 0x0 phys_seg 0 prio class 0
[ 2836.672085] blk_update_request: I/O error, dev nvme1n1, sector 14294 op 0x9:(WRITE_ZEROES) flags 0x0 phys_seg 0 prio class 0
[ 2836.672087] blk_update_request: I/O error, dev nvme1n1, sector 14037 op 0x9:(WRITE_ZEROES) flags 0x0 phys_seg 0 prio class 0
[ 2836.672121] nvme nvme1: failed to mark controller live state
[ 2836.672123] nvme nvme1: Removing after probe failure status: -19
[ 2836.689016] Aborting journal on device dm-0-8.
[ 2836.689024] Buffer I/O error on dev dm-0, logical block 25198592, lost sync page write
[ 2836.689027] JBD2: Error -5 detected when updating journal superblock for dm-0-8.

Reported-by: Bradley Chapman <[email protected]>
Signed-off-by: Chaitanya Kulkarni <[email protected]>
Tested-by: Bradley Chapman <[email protected]>
Reviewed-by: Keith Busch <[email protected]>
Signed-off-by: Christoph Hellwig <[email protected]>

Merge tag 'for-linus-5.11-rc6-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip

Pull xen fixes from Juergen Gross:

 - A fix for a regression introduced in 5.11 resulting in Xen dom0
   having problems to correctly initialize Xenstore.

 - A fix for avoiding WARN splats when booting as Xen dom0 with
   CONFIG_AMD_MEM_ENCRYPT enabled due to a missing trap handler for the
   #VC exception (even if the handler should never be called).

 - A fix for the Xen bklfront driver adapting to the correct but
   unexpected behavior of new qemu.

* tag 'for-linus-5.11-rc6-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip:
  x86/xen: avoid warning in Xen pv guest with CONFIG_AMD_MEM_ENCRYPT enabled
  xen: Fix XenStore initialisation for XS_LOCAL
  xen-blkfront: allow discard-* nodes to be optional

Fix unsynchronized access to sev members through svm_register_enc_region

Grab kvm->lock before pinning memory when registering an encrypted
region; sev_pin_memory() relies on kvm->lock being held to ensure
correctness when checking and updating the number of pinned pages.

Add a lockdep assertion to help prevent future regressions.

Cc: Thomas Gleixner <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: "H. Peter Anvin" <[email protected]>
Cc: Paolo Bonzini <[email protected]>
Cc: Joerg Roedel <[email protected]>
Cc: Tom Lendacky <[email protected]>
Cc: Brijesh Singh <[email protected]>
Cc: Sean Christopherson <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Fixes: 1e80fdc09d12 ("KVM: SVM: Pin guest memory when SEV is active")
Signed-off-by: Peter Gonda <[email protected]>
V2
 - Fix up patch description
 - Correct file paths svm.c -> sev.c
 - Add unlock of kvm->lock on sev_pin_memory error

V1
 - https://lore.kernel.org/kvm/20210126185431.1824530[email protected]/

Message-Id: <20210127161524.2832400[email protected]>
Signed-off-by: Paolo Bonzini <[email protected]>

KVM: Documentation: Fix documentation for nested.

Nested VMX was enabled by default in commit 1e58e5e59148 ("KVM:
VMX: enable nested virtualization by default"), which was merged
in Linux 4.20. This patch is to fix the documentation accordingly.

Signed-off-by: Yu Zhang <[email protected]>
Message-Id: <20210128154747[email protected]>
Signed-off-by: Paolo Bonzini <[email protected]>

Merge tag 'kvmarm-fixes-5.11-3' of git://git.kernel.org/pub/scm/linux/kernel/git/kvmarm/kvmarm into HEAD

KVM/arm64 fixes for 5.11, take #3

- Avoid clobbering extra registers on initialisation

Merge tag 'asm-generic-fixes-v5.11' of git://git.kernel.org/pub/scm/linux/kernel/git/arnd/asm-generic

Pull ia64 fixes from Arnd Bergmann:
 "asm-generic/ia64 fixes, and mark as orphaned

  Commit 2b49ddcef297 ("ia64: convert to legacy_timer_tick") from my
  timer series I merged through the asm-generic tree caused a regression
  on all ia64 machines, as bisected by Adrian Glaubitz.

  Tony Luck is no longer really working on ia64, so instead of merging
  the fix through his tree, we ended up deciding that I'd merge the fix
  myself along a patch to mark the architecture as Orphaned and a
  compile time warning fix I made while working on the regression"

[ HPE no longer accepts orders for new Itanium hardware, and Intel
  stopped accepting orders a year ago. While intel is still officially
  shipping chips until July 29, 2021, it's unlikely that any such orders
  actually exist.

  It's dead, Jim.
                     - Linus ]

* tag 'asm-generic-fixes-v5.11' of git://git.kernel.org/pub/scm/linux/kernel/git/arnd/asm-generic:
  ia64: Mark architecture as orphaned
  ia64: fix xchg() warning
  ia64: fix timer cleanup regression

Merge tag 'arm-soc-fixes-v5.11-2' of git://git.kernel.org/pub/scm/linux/kernel/git/soc/soc

Pull ARM SoC fixes from Arnd Bergmann:
 "These are the current arm-soc bug fixes for linux-5.11. I already
  merged a larger set that just came in during the past three days but
  has not had much exposure in linux-next, but this is the subset I
  merged last week.

  Most of these are for the NXP i.MX platform (descriptions from their
  pull request):

   - Fix pcf2127 reset for imx7d-flex-concentrator board.
   - Fix i.MX6 suspend with Thumb-2 kernel.
   - Fix ethernet-phy address issue on imx6qdl-sr-som board.
   - Fix GPIO3 `gpio-ranges` on i.MX8MP.
   - Select SOC_BUS for IMX_SCU driver to fix build issue.
   - Fix backlight pwm on imx6qdl-kontron-samx6i which is lost from
     #pwm-cells conversion.
   - Fix duplicated bus node name for i.MX8MN SoC.
   - Fix reset register offset on LS1028A SoC.
   - Rename MMC node aliases for imx6q-tbs2910 to keep the MMC device
     index consistent with previous kernel version.
   - Selecting ARM_GIC_V3 on non-CP15 processors to fix one build
     failure with i.MX8M SoC driver.
   - Fix typos with status property on imx6qdl-kontron-samx6i board.
   - Fix duplicated regulator-name on imx6qdl-gw52xx board.

  Aside from i.MX, the bugfixes are all over the place:

   - Coccinelle found a refcount imbalance on integrator
   - defconfig fix for TI K3
   - A boot regression fix for ST ux500
   - A code preemption fix for the optee driver
   - USB DMA regression on Broadcom Stingray
   - A bogus boot time warning fix for at91 code"

* tag 'arm-soc-fixes-v5.11-2' of git://git.kernel.org/pub/scm/linux/kernel/git/soc/soc:
  MAINTAINERS: Include bcm2835 subsequents into search
  arm64: dts: broadcom: Fix USB DMA address translation for Stingray
  drivers: soc: atmel: add null entry at the end of at91_soc_allowed_list[]
  drivers: soc: atmel: Avoid calling at91_soc_init on non AT91 SoCs
  tee: optee: replace might_sleep with cond_resched
  firmware: imx: select SOC_BUS to fix firmware build
  arm64: dts: imx8mp: Correct the gpio ranges of gpio3
  ARM: dts: imx6qdl-sr-som: fix some cubox-i platforms
  ARM: imx: build suspend-imx6.S with arm instruction set
  ARM: dts: imx7d-flex-concentrator: fix pcf2127 reset
  ARM: dts: ux500: Reserve memory carveouts
  arm64: defconfig: Drop unused K3 SoC specific options
  bus: arm-integrator-lm: Add of_node_put() before return statement
  ARM: dts: imx6qdl-gw52xx: fix duplicate regulator naming
  ARM: dts: imx6qdl-kontron-samx6i: fix i2c_lcd/cam default status
  ARM: imx: fix imx8m dependencies
  ARM: dts: tbs2910: rename MMC node aliases
  arm64: dts: ls1028a: fix the offset of the reset register
  arm64: dts: imx8mn: Fix duplicate node name
  ARM: dts: imx6qdl-kontron-samx6i: fix pwms for lcd-backlight

Merge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma

Pull rdma fixes from Jason Gunthorpe:
 "Several recent regressions and some bug fixes:

   - Typo corrupting the max_recv_sge for cxgb4

   - Regression from re-using kernel enums as a HW AbI in vmw_pvrdma

   - Sleeping inside a spinlock in hns

   - Revert the attempt to fix devlink deadlocks as the fix is more buggy

   - Typo in sysfs_emit_at conversions

   - Revert the removal of VLAN support in rxe"

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma:
  Revert "RDMA/rxe: Remove VLAN code leftovers from RXE"
  RDMA/usnic: Fix misuse of sysfs_emit_at
  Revert "RDMA/mlx5: Fix devlink deadlock on net namespace deletion"
  RDMA/hns: Use mutex instead of spinlock for ida allocation
  RDMA/vmw_pvrdma: Fix network_hdr_type reported in WC
  RDMA/cxgb4: Fix the reported max_recv_sge value

Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/hid/hid

Pull HID fixes from Jiri Kosina:

 - NULL pointer dereference regression fix for Wacom driver (Jason
   Gerecke)

 - functional regression fix for pam handling on some Elan and Synaptics
   touchpads (Kai-Heng Feng)

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/hid/hid:
  HID: wacom: Correct NULL dereference on AES pen proximity
  HID: multitouch: Apply MT_QUIRK_CONFIDENCE quirk for multi-input devices

cifs: fix mounts to subdirectories of target

The "prefixpath" mount option needs to be ignored
which was missed in the recent conversion to the
new mount API (prefixpath would be set by the mount
helper if mounting a subdirectory of the root of a
share e.g. //server/share/subdir)

Fixes: 24e0a1eff9e2 ("cifs: switch to new mount api")
Suggested-by: Ronnie Sahlberg <[email protected]>
Signed-off-by: Steve French <[email protected]>
Reviewed-by: Ronnie Sahlberg <[email protected]>

Merge tag 'media/v5.11-2' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-media

Pull media fixes from Mauro Carvalho Chehab:

 - a V4L2 core regression at videobuf2 when checking for single-plane
   dmabuf

 - a change at uAPI header v4l2-subdev.h, fixing a breakage as BIT()
   macro is not available in userspace

 - fix some regressions at RC core due to the usage of microseconds
   everywhere on it

 - a fix for a race condition at RC core

 - a rename on a newly-introduced kAPI symbol (v4l2_get_link_rate),
   currently used only by a single driver

 - Regression fixes for rcar-vin, cedrus, ite-cir, hantro, css, venus,
   and cec drivers.

* tag 'media/v5.11-2' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-media:
  media: hantro: Fix reset_raw_fmt initialization
  media: cec: add stm32 driver
  media: cedrus: Fix H264 decoding
  media: v4l2-subdev.h: BIT() is not available in userspace
  media: Revert "media: videobuf2: Fix length check for single plane dmabuf queueing"
  media: rc: ite-cir: fix min_timeout calculation
  media: venus: core: Fix platform driver shutdown
  media: rc: fix timeout handling after switch to microsecond durations
  media: v4l: common: Fix naming of v4l2_get_link_rate
  media: rcar-vin: fix return, use ret instead of zero
  media: ccs: Get static data version minor correctly
  media: ccs-pll: Fix link frequency for C-PHY
  media: rc: ensure that uevent can be read directly after rc device register

Merge tag 'clk-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/clk/linux

Pull clk fixes from Stephen Boyd:
 "A handful of clk driver fixes:

   - Build fix for CONFIG_PM=n in the mmp2 driver

   - Kconfig warning for unmet dependencies in the i.MX driver

   - Make the camera AHB clk always be enabled on qcom sc7180

   - Use rate round down semantics for qcom sm8250 SD clks"

* tag 'clk-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/clk/linux:
  clk: mmp2: fix build without CONFIG_PM
  clk: qcom: gcc-sm250: Use floor ops for sdcc clks
  clk: imx: fix Kconfig warning for i.MX SCU clk
  clk: qcom: gcc-sc7180: Mark the camera abh clock always ON

Merge tag 'sound-5.11-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound

Pull sound fixes from Takashi Iwai:
 "Although the incoming fixes haven't settled down yet, all changes here
  are small and mostly device-specific fixes, so nothing look worrisome.

   - Yet another USB-audio regression fixes

   - HD-audio ID fix and device-specific quirks

   - SOF Intel / SoundWire fixes including topology

   - ASoC Qualcomm and Mediatek fixes"

* tag 'sound-5.11-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound: (24 commits)
  ALSA: hda/via: Apply the workaround generically for Clevo machines
  ASoC: Intel: sof_sdw: set proper flags for Dell TGL-H SKU 0A5E
  ASoC: qcom: lpass: Fix out-of-bounds DAI ID lookup
  ASoC: mediatek: mt8192-mt6359: add format constraints for RT5682
  ASoC: ak4458: correct reset polarity
  ASoC: SOF: SND_INTEL_DSP_CONFIG dependency
  ASoC: SOF: Intel: soundwire: fix select/depend unmet dependencies
  ALSA: hda: intel-dsp-config: add PCI id for TGL-H
  ALSA: usb-audio: workaround for iface reset issue
  ALSA: pcm: One more dependency for hw constraints
  ALSA: hda/realtek: Enable headset of ASUS B1400CEPE with ALC256
  ASoC: Intel: Skylake: Zero snd_ctl_elem_value
  ASoC: Intel: Skylake: skl-topology: Fix OOPs ib skl_tplg_complete
  ASoC: qcom: Fix number of HDMI RDMA channels on sc7180
  ASoC: mediatek: mt8183-da7219: ignore TDM DAI link by default
  ASoC: mediatek: mt8183-mt6358: ignore TDM DAI link by default
  ASoC: topology: Properly unregister DAI on removal
  ASoC: topology: Fix memory corruption in soc_tplg_denum_create_values()
  ASoC: qcom: lpass-ipq806x: fix bitwidth regmap field
  ASoC: AMD Renoir - refine DMI entries for some Lenovo products
  ...

Revert "mm/slub: fix a memory leak in sysfs_slab_add()"

This reverts commit dde3c6b72a16c2db826f54b2d49bdea26c3534a2.

syzbot report a double-free bug. The following case can cause this bug.

 - mm/slab_common.c: create_cache(): if the __kmem_cache_create() fails,
   it does:

out_free_cache:
kmem_cache_free(kmem_cache, s);

 - but __kmem_cache_create() - at least for slub() - will have done

sysfs_slab_add(s)
-> sysfs_create_group() .. fails ..
-> kobject_del(&s->kobj); .. which frees s ...

We can't remove the kmem_cache_free() in create_cache(), because other
error cases of __kmem_cache_create() do not free this.

So, revert the commit dde3c6b72a16 ("mm/slub: fix a memory leak in
sysfs_slab_add()") to fix this.

Reported-by: [email protected]
Fixes: dde3c6b72a16 ("mm/slub: fix a memory leak in sysfs_slab_add()")
Acked-by: Vlastimil Babka <[email protected]>
Signed-off-by: Wang Hai <[email protected]>
Cc: <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

cifs: ignore auto and noauto options if given

In 24e0a1eff9e2, the noauto and auto options were missed when migrating
to the new mount API. As a result, users with noauto in their fstab
mount options are now unable to mount cifs filesystems, as they'll
receive an "Unknown parameter" error.

This restores the old behaviour of ignoring noauto and auto if they're
given.

Fixes: 24e0a1eff9e2 ("cifs: switch to new mount api")
Signed-off-by: Adam Harvey <[email protected]>
Reviewed-by: Ronnie Sahlberg <[email protected]>
Signed-off-by: Steve French <[email protected]>

bcache: only check feature sets when sb->version >= BCACHE_SB_VERSION_CDEV_WITH_FEATURES

For super block version < BCACHE_SB_VERSION_CDEV_WITH_FEATURES, it
doesn't make sense to check the feature sets. This patch checks
super block version in bch_has_feature_* routines, if the version
doesn't have feature sets yet, returns 0 (false) to the caller.

Fixes: 5342fd425502 ("bcache: set bcache device into read-only mode for BCH_FEATURE_INCOMPAT_OBSO_LARGE_BUCKET")
Fixes: ffa470327572 ("bcache: add bucket_size_hi into struct cache_sb_disk for large bucket")
Cc: [email protected] # 5.9+
Reported-and-tested-by: Bockholdt Arne <[email protected]>
Signed-off-by: Coly Li <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>

block: fix bd_size_lock use

Some block device drivers, e.g. the skd driver, call set_capacity() with
IRQ disabled. This results in lockdep ito complain about inconsistent
lock states ("inconsistent {HARDIRQ-ON-W} -> {IN-HARDIRQ-W} usage")
because set_capacity takes a block device bd_size_lock using the
functions spin_lock() and spin_unlock(). Ensure a consistent locking
state by replacing these calls with spin_lock_irqsave() and
spin_lock_irqrestore(). The same applies to bdev_set_nr_sectors().
With this fix, all lockdep complaints are resolved.

Signed-off-by: Damien Le Moal <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>

blk-cgroup: Use cond_resched() when destroy blkgs

On !PREEMPT kernel, we can get below softlockup when doing stress
testing with creating and destroying block cgroup repeatly. The
reason is it may take a long time to acquire the queue's lock in
the loop of blkcg_destroy_blkgs(), or the system can accumulate a
huge number of blkgs in pathological cases. We can add a need_resched()
check on each loop and release locks and do cond_resched() if true
to avoid this issue, since the blkcg_destroy_blkgs() is not called
from atomic contexts.

[ 4757.010308] watchdog: BUG: soft lockup - CPU#11 stuck for 94s!
[ 4757.010698] Call trace:
[ 4757.010700]  blkcg_destroy_blkgs+0x68/0x150
[ 4757.010701]  cgwb_release_workfn+0x104/0x158
[ 4757.010702]  process_one_work+0x1bc/0x3f0
[ 4757.010704]  worker_thread+0x164/0x468
[ 4757.010705]  kthread+0x108/0x138

Suggested-by: Tejun Heo <[email protected]>
Signed-off-by: Baolin Wang <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>

iommu/vt-d: Do not use flush-queue when caching-mode is on

When an Intel IOMMU is virtualized, and a physical device is
passed-through to the VM, changes of the virtual IOMMU need to be
propagated to the physical IOMMU. The hypervisor therefore needs to
monitor PTE mappings in the IOMMU page-tables. Intel specifications
provide "caching-mode" capability that a virtual IOMMU uses to report
that the IOMMU is virtualized and a TLB flush is needed after mapping to
allow the hypervisor to propagate virtual IOMMU mappings to the physical
IOMMU. To the best of my knowledge no real physical IOMMU reports
"caching-mode" as turned on.

Synchronizing the virtual and the physical IOMMU tables is expensive if
the hypervisor is unaware which PTEs have changed, as the hypervisor is
required to walk all the virtualized tables and look for changes.
Consequently, domain flushes are much more expensive than page-specific
flushes on virtualized IOMMUs with passthrough devices. The kernel
therefore exploited the "caching-mode" indication to avoid domain
flushing and use page-specific flushing in virtualized environments. See
commit 78d5f0f500e6 ("intel-iommu: Avoid global flushes with caching
mode.")

This behavior changed after commit 13cf01744608 ("iommu/vt-d: Make use
of iova deferred flushing"). Now, when batched TLB flushing is used (the
default), full TLB domain flushes are performed frequently, requiring
the hypervisor to perform expensive synchronization between the virtual
TLB and the physical one.

Getting batched TLB flushes to use page-specific invalidations again in
such circumstances is not easy, since the TLB invalidation scheme
assumes that "full" domain TLB flushes are performed for scalability.

Disable batched TLB flushes when caching-mode is on, as the performance
benefit from using batched TLB invalidations is likely to be much
smaller than the overhead of the virtual-to-physical IOMMU page-tables
synchronization.

Fixes: 13cf01744608 ("iommu/vt-d: Make use of iova deferred flushing")
Signed-off-by: Nadav Amit <[email protected]>
Cc: David Woodhouse <[email protected]>
Cc: Lu Baolu <[email protected]>
Cc: Joerg Roedel <[email protected]>
Cc: Will Deacon <[email protected]>
Cc: [email protected]
Acked-by: Lu Baolu <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Joerg Roedel <[email protected]>

Merge tag 'amlogic-fixes-3' of git://git.kernel.org/pub/scm/linux/kernel/git/khilman/linux-amlogic into arm/fixes

arm64: dts: amlogic fixes for v5.11-rc (round 3)
- fix reboot issue on odroid C4

* tag 'amlogic-fixes-3' of git://git.kernel.org/pub/scm/linux/kernel/git/khilman/linux-amlogic:
  arm64: dts: meson: switch TFLASH_VDD_EN pin to open drain on Odroid-C4

Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnd Bergmann <[email protected]>

Merge tag 'amlogic-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/khilman/linux-amlogic into arm/fixes

arm64: dts: amlogic fixes for v5.11-rc (round 2)
- fix ethernet resets

* tag 'amlogic-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/khilman/linux-amlogic:
  Revert "arm64: dts: amlogic: add missing ethernet reset ID"

Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnd Bergmann <[email protected]>

Merge tag 'v5.11-rockchip-dtsfixes1' of git://git.kernel.org/pub/scm/linux/kernel/git/mmind/linux-rockchip into arm/fixes

Wrong irq number on px30 and cleanups of stuff on rk3399 regarding
wrongly used dt properties and parts that shouldn't be enabled.

* tag 'v5.11-rockchip-dtsfixes1' of git://git.kernel.org/pub/scm/linux/kernel/git/mmind/linux-rockchip:
  arm64: dts: rockchip: Disable display for NanoPi R2S
  arm64: dts: rockchip: remove interrupt-names property from rk3399 vdec node
  arm64: dts: rockchip: Fix PCIe DT properties on rk3399
  arm64: dts: rockchip: Use only supported PCIe link speed on Pinebook Pro
  arm64: dts: rockchip: fix vopl iommu irq on px30

Link: https://lore.kernel.org/r/5429065.DvuYhMxLoT@phil
Signed-off-by: Arnd Bergmann <[email protected]>

Merge tag 'omap-for-v5.11/fixes-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/tmlind/linux-omap into arm/fixes

Fixes for omaps for v5.11-rc cycle

The recent changes to switch SoCs to boot with ti-sysc interconnect
target module driver and genpd caused few regressions:

- The omap_prm reset driver needs to clear any reset bits deasserted by
  the bootloader or kexec boot for the three reset bit cases. Otherwise
  we can have an oops with accelerators starting to boot with potentially
  unconfigured MMU for example

- Custom kernel configs are not automatically selecting simple-pm-bus
  driver that we now need to probe interconnects so we need to select it
  always

- We are not passing legacy platform data in auxdata with simple-pm-bus
  like we do for simple-bus. We need to pass auxdata to simple-pm-bus so
  it can pass it to of_platform_populate()

Then recent RCU changes started causing splats for cpuidle44xx that now
need RCU_NONIDLE added to the calls in several places

And then we have few device specific fixes:

- We need to remove legacy spi-cs-hig for gta04 display to work, and
  set the gpio to active low

- Omap1 specific ohci-omap needs to call gpio_free()

- Droid4 needs to use padconf interrupt for the slider as the edge
  gpio interrupts may be lost for deeper idle states

* tag 'omap-for-v5.11/fixes-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/tmlind/linux-omap:
  ARM: dts: omap4-droid4: Fix lost keypad slide interrupts for droid4
  drivers: bus: simple-pm-bus: Fix compatibility with simple-bus for auxdata
  ARM: OMAP2+: Fix booting for am335x after moving to simple-pm-bus
  ARM: OMAP2+: Fix suspcious RCU usage splats for omap_enter_idle_coupled
  ARM: dts; gta04: SPI panel chip select is active low
  soc: ti: omap-prm: Fix boot time errors for rst_map_012 bits 0 and 1
  ARM: OMAP1: OSK: fix ohci-omap breakage
  DTS: ARM: gta04: remove legacy spi-cs-high to make display work again

Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnd Bergmann <[email protected]>

entry: Unbreak single step reporting behaviour

The move of TIF_SYSCALL_EMU to SYSCALL_WORK_SYSCALL_EMU broke single step
reporting. The original code reported the single step when TIF_SINGLESTEP
was set and TIF_SYSCALL_EMU was not set. The SYSCALL_WORK conversion got
the logic wrong and now the reporting only happens when both bits are set.

Restore the original behaviour.

[ tglx: Massaged changelog and dropped the pointless double negation ]

Fixes: 64eb35f701f0 ("ptrace: Migrate TIF_SYSCALL_EMU to use SYSCALL_WORK flag")
Signed-off-by: Yuxuan Shui <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
Reviewed-by: Gabriel Krisman Bertazi <[email protected]>
Link: https://lore.kernel.org/r/877do3gaq9.fsf@m5Zedd9JOGzJrf0

thunderbolt: Fix possible NULL pointer dereference in tb_acpi_add_link()

When we walk up the device hierarchy in tb_acpi_add_link() make sure we
break the loop if the device has no parent. Otherwise we may crash the
kernel by dereferencing a NULL pointer.

Fixes: b2be2b05cf3b ("thunderbolt: Create device links from ACPI description")
Cc: [email protected]
Signed-off-by: Mario Limonciello <[email protected]>
Acked-by: Yehezkel Bernat <[email protected]>
Signed-off-by: Mika Westerberg <[email protected]>

iommu/vt-d: Correctly check addr alignment in qi_flush_dev_iotlb_pasid()

An incorrect address mask is being used in the qi_flush_dev_iotlb_pasid()
to check the address alignment. This leads to a lot of spurious kernel
warnings:

[  485.837093] DMAR: Invalidate non-aligned address 7f76f47f9000, order 0
[  485.837098] DMAR: Invalidate non-aligned address 7f76f47f9000, order 0
[  492.494145] qi_flush_dev_iotlb_pasid: 5734 callbacks suppressed
[  492.494147] DMAR: Invalidate non-aligned address 7f7728800000, order 11
[  492.508965] DMAR: Invalidate non-aligned address 7f7728800000, order 11

Fix it by checking the alignment in right way.

Fixes: 288d08e780088 ("iommu/vt-d: Handle non-page aligned address")
Reported-and-tested-by: Guo Kaijie <[email protected]>
Signed-off-by: Lu Baolu <[email protected]>
Cc: Liu Yi L <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Joerg Roedel <[email protected]>

drm/ttm: Use __GFP_NOWARN for huge pages in ttm_pool_alloc_page

Without __GFP_NOWARN, attempts at allocating huge pages can trigger
dmesg splats like below (which are essentially noise, since TTM falls
back to normal pages if it can't get a huge one).

[ 9556.710241] clinfo: page allocation failure: order:9, mode:0x194dc2(GFP_HIGHUSER|__GFP_RETRY_MAYFAIL|__GFP_NORETRY|__GFP_ZERO|__GFP_NOMEMALLOC), nodemask=(null),cpuset=user.slice,mems_allowed=0
[ 9556.710259] CPU: 1 PID: 470821 Comm: clinfo Tainted: G            E     5.10.10+ #4
[ 9556.710264] Hardware name: Micro-Star International Co., Ltd. MS-7A34/B350 TOMAHAWK (MS-7A34), BIOS 1.OR 11/29/2019
[ 9556.710268] Call Trace:
[ 9556.710281]  dump_stack+0x6b/0x83
[ 9556.710288]  warn_alloc.cold+0x7b/0xdf
[ 9556.710297]  ? __alloc_pages_direct_compact+0x137/0x150
[ 9556.710303]  __alloc_pages_slowpath.constprop.0+0xc1b/0xc50
[ 9556.710312]  __alloc_pages_nodemask+0x2ec/0x320
[ 9556.710325]  ttm_pool_alloc+0x2e4/0x5e0 [ttm]
[ 9556.710332]  ? kvmalloc_node+0x46/0x80
[ 9556.710341]  ttm_tt_populate+0x37/0xe0 [ttm]
[ 9556.710350]  ttm_bo_handle_move_mem+0x142/0x180 [ttm]
[ 9556.710359]  ttm_bo_validate+0x11d/0x190 [ttm]
[ 9556.710391]  ? drm_vma_offset_add+0x2f/0x60 [drm]
[ 9556.710399]  ttm_bo_init_reserved+0x2a7/0x320 [ttm]
[ 9556.710529]  amdgpu_bo_do_create+0x1b8/0x500 [amdgpu]
[ 9556.710657]  ? amdgpu_bo_subtract_pin_size+0x60/0x60 [amdgpu]
[ 9556.710663]  ? get_page_from_freelist+0x11f9/0x1450
[ 9556.710789]  amdgpu_bo_create+0x40/0x270 [amdgpu]
[ 9556.710797]  ? _raw_spin_unlock+0x16/0x30
[ 9556.710927]  amdgpu_gem_create_ioctl+0x123/0x310 [amdgpu]
[ 9556.711062]  ? amdgpu_gem_force_release+0x150/0x150 [amdgpu]
[ 9556.711098]  drm_ioctl_kernel+0xaa/0xf0 [drm]
[ 9556.711133]  drm_ioctl+0x20f/0x3a0 [drm]
[ 9556.711267]  ? amdgpu_gem_force_release+0x150/0x150 [amdgpu]
[ 9556.711276]  ? preempt_count_sub+0x9b/0xd0
[ 9556.711404]  amdgpu_drm_ioctl+0x49/0x80 [amdgpu]
[ 9556.711411]  __x64_sys_ioctl+0x83/0xb0
[ 9556.711417]  do_syscall_64+0x33/0x80
[ 9556.711421]  entry_SYSCALL_64_after_hwframe+0x44/0xa9

Fixes: bf9eee249ac2 ("drm/ttm: stop using GFP_TRANSHUGE_LIGHT")
Signed-off-by: Michel Dänzer <[email protected]>
Reviewed-by: Christian König <[email protected]>
Signed-off-by: Christian König <[email protected]>
Link: https://patchwork.freedesktop.org/patch/416353/

iommu/amd: Use IVHD EFR for early initialization of IOMMU features

IOMMU Extended Feature Register (EFR) is used to communicate
the supported features for each IOMMU to the IOMMU driver.
This is normally read from the PCI MMIO register offset 0x30,
and used by the iommu_feature() helper function.

However, there are certain scenarios where the information is needed
prior to PCI initialization, and the iommu_feature() function is used
prematurely w/o warning. This has caused incorrect initialization of IOMMU.
This is the case for the commit 6d39bdee238f ("iommu/amd: Enforce 4k
mapping for certain IOMMU data structures")

Since, the EFR is also available in the IVHD header, and is available to
the driver prior to PCI initialization. Therefore, default to using
the IVHD EFR instead.

Fixes: 6d39bdee238f ("iommu/amd: Enforce 4k mapping for certain IOMMU data structures")
Signed-off-by: Suravee Suthikulpanit <[email protected]>
Tested-by: Brijesh Singh <[email protected]>
Reviewed-by: Robert Richter <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Joerg Roedel <[email protected]>

drm/bridge/lontium-lt9611uxc: move HPD notification out of IRQ handler

drm hotplug handling code (drm_client_dev_hotplug()) can wait on mutex,
thus delaying further lt9611uxc IRQ events processing.  It was observed
occasionally during bootups, when drm_client_modeset_probe() was waiting
for EDID ready event, which was delayed because IRQ handler was stuck
trying to deliver hotplug event.
Move hotplug notifications from IRQ handler to separate work to be able
to process IRQ events without delays.

Signed-off-by: Dmitry Baryshkov <[email protected]>
Fixes: 0cbbd5b1a012 ("drm: bridge: add support for lontium LT9611UXC bridge")
Reviewed-by: Andrzej Hajda <[email protected]>
Signed-off-by: Andrzej Hajda <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/bridge/lontium-lt9611uxc: fix get_edid return code

Return NULL pointer from get_edid() callback rather than ERR_PTR()
pointer, as DRM code does NULL checks rather than IS_ERR(). Also while
we are at it, return NULL if getting EDID timed out.

Signed-off-by: Dmitry Baryshkov <[email protected]>
Fixes: 0cbbd5b1a012 ("drm: bridge: add support for lontium LT9611UXC bridge")
Reviewed-by: Bjorn Andersson <[email protected]>
Reviewed-by: Andrzej Hajda <[email protected]>
Signed-off-by: Andrzej Hajda <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

drm/bridge/lontium-lt9611uxc: fix waiting for EDID to become available

- Call wake_up() when EDID ready event is received to wake
  wait_event_interruptible_timeout()

- Increase waiting timeout, reading EDID can take longer than 100ms, so
  let's be on a safe side.

Signed-off-by: Dmitry Baryshkov <[email protected]>
Fixes: 0cbbd5b1a012 ("drm: bridge: add support for lontium LT9611UXC bridge")
Reviewed-by: Bjorn Andersson <[email protected]>
Reviewed-by: Andrzej Hajda <[email protected]>
Signed-off-by: Andrzej Hajda <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

media: rockchip: rkisp1: extend uapi array sizes

Later variants of the rkisp1 block use more entries in some arrays:

RKISP1_CIF_ISP_AE_MEAN_MAX                 25 -> 81
RKISP1_CIF_ISP_HIST_BIN_N_MAX              16 -> 32
RKISP1_CIF_ISP_GAMMA_OUT_MAX_SAMPLES       17 -> 34
RKISP1_CIF_ISP_HISTOGRAM_WEIGHT_GRIDS_SIZE 25 -> 81

and we can still extend the uapi during the 5.11-rc cycle, so do that
now to be on the safe side.

V10 and V11 only need the smaller sizes, while V12 and V13 needed
the larger sizes.

When adding the bigger sizes make sure, values filled from hardware
values and transmitted to userspace don't leak kernel data by zeroing
them beforehand.

Signed-off-by: Heiko Stuebner <[email protected]>
Signed-off-by: Dafna Hirschfeld <[email protected]>
Reviewed-by: Ezequiel Garcia <[email protected]>
Acked-by: Helen Koike <[email protected]>
Signed-off-by: Hans Verkuil <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>

media: rockchip: rkisp1: carry ip version information

The IP block evolved from its rk3288/rk3399 base and the vendor
designates them with a numerical version. rk3399 for example
is designated V10 probably meaning V1.0.

There doesn't seem to be an actual version register we could read that
information from, so allow the match_data to carry that information
for future differentiation.

Also carry that information in the hw_revision field of the media-
controller API, so that userspace also has access to that.

The added versions are:
- V10: at least rk3288 + rk3399
- V11: seemingly unused as of now, but probably appeared in some soc
- V12: at least rk3326 + px30
- V13: at least rk1808

[fix checkpatch warning don't use multiple blank lines]

Signed-off-by: Heiko Stuebner <[email protected]>
Signed-off-by: Dafna Hirschfeld <[email protected]>
Reviewed-by: Ezequiel Garcia <[email protected]>
Acked-by: Helen Koike <[email protected]>
Signed-off-by: Hans Verkuil <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>

media: rockchip: rkisp1: reduce number of histogram grid elements in uapi

The uapi right now specifies an array size of 28 but the actual number
of elements is only 25 with the last 3 being unused.

Reduce the array size to the correct number of elements and change
the params code to iterate the array 25 times.

Signed-off-by: Heiko Stuebner <[email protected]>
Signed-off-by: Dafna Hirschfeld <[email protected]>
Acked-by: Helen Koike <[email protected]>
Signed-off-by: Hans Verkuil <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>

media: rkisp1: stats: mask the hist_bins values

hist_bins is an array of type __u32. Each entry represents
a 20 bit value. So mask out the unused bits.

Signed-off-by: Dafna Hirschfeld <[email protected]>
Acked-by: Helen Koike <[email protected]>
Signed-off-by: Hans Verkuil <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>

media: rkisp1: stats: remove a wrong cast to u8

hist_bins is an array of type __u32. Each entry represent
a 20 bit fixed point value as documented inline.
The cast to u8 when setting the values is wrong. Remove it.

Signed-off-by: Dafna Hirschfeld <[email protected]>
Reviewed-by: Heiko Stuebner <[email protected]>
Acked-by: Helen Koike <[email protected]>
Signed-off-by: Hans Verkuil <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>

media: rkisp1: uapi: change hist_bins array type from __u16 to __u32

Each entry in the array is a 20 bits value composed of 16 bits unsigned
integer and 4 bits fractional part. So the type should change to __u32.
In addition add a documentation of how the measurements are done.

Signed-off-by: Dafna Hirschfeld <[email protected]>
Acked-by: Helen Koike <[email protected]>
Signed-off-by: Hans Verkuil <[email protected]>
Signed-off-by: Mauro Carvalho Chehab <[email protected]>

soc: sunxi: mbus: Remove DE2 display engine compatibles

The DE2 display engine hardware takes physical addresses that do not
need PHYS_BASE subtracted. As a result, they should not be present
on the mbus driver match list. Remove them.

This was tested on the A83T, along with the patch allowing the DMA
range map to be non-NULL and restores a working display.

Fixes: b4bdc4fbf8d0 ("soc: sunxi: Deal with the MBUS DMA offsets in a central place")
Signed-off-by: Paul Kocialkowski <[email protected]>
Signed-off-by: Maxime Ripard <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

i2c: mediatek: Move suspend and resume handling to NOIRQ phase

Some i2c device driver indirectly uses I2C driver when it is now
being suspended. The i2c devices driver is suspended during the
NOIRQ phase and this cannot be changed due to other dependencies.
Therefore, we also need to move the suspend handling for the I2C
controller driver to the NOIRQ phase as well.

Signed-off-by: Qii Wang <[email protected]>
Signed-off-by: Wolfram Sang <[email protected]>

ovl: implement volatile-specific fsync error behaviour

Overlayfs's volatile option allows the user to bypass all forced sync calls
to the upperdir filesystem. This comes at the cost of safety. We can never
ensure that the user's data is intact, but we can make a best effort to
expose whether or not the data is likely to be in a bad state.

The best way to handle this in the time being is that if an overlayfs's
upperdir experiences an error after a volatile mount occurs, that error
will be returned on fsync, fdatasync, sync, and syncfs. This is
contradictory to the traditional behaviour of VFS which fails the call
once, and only raises an error if a subsequent fsync error has occurred,
and been raised by the filesystem.

One awkward aspect of the patch is that we have to manually set the
superblock's errseq_t after the sync_fs callback as opposed to just
returning an error from syncfs. This is because the call chain looks
something like this:

sys_syncfs ->
sync_filesystem ->
__sync_filesystem ->
/* The return value is ignored here
sb->s_op->sync_fs(sb)
_sync_blockdev
/* Where the VFS fetches the error to raise to userspace */
errseq_check_and_advance

Because of this we call errseq_set every time the sync_fs callback occurs.
Due to the nature of this seen / unseen dichotomy, if the upperdir is an
inconsistent state at the initial mount time, overlayfs will refuse to
mount, as overlayfs cannot get a snapshot of the upperdir's errseq that
will increment on error until the user calls syncfs.

Signed-off-by: Sargun Dhillon <[email protected]>
Suggested-by: Amir Goldstein <[email protected]>
Reviewed-by: Amir Goldstein <[email protected]>
Fixes: c86243b090bc ("ovl: provide a mount option "volatile"")
Cc: [email protected]
Reviewed-by: Vivek Goyal <[email protected]>
Reviewed-by: Jeff Layton <[email protected]>
Signed-off-by: Miklos Szeredi <[email protected]>

ovl: skip getxattr of security labels

When inode has no listxattr op of its own (e.g. squashfs) vfs_listxattr
calls the LSM inode_listsecurity hooks to list the xattrs that LSMs will
intercept in inode_getxattr hooks.

When selinux LSM is installed but not initialized, it will list the
security.selinux xattr in inode_listsecurity, but will not intercept it
in inode_getxattr.  This results in -ENODATA for a getxattr call for an
xattr returned by listxattr.

This situation was manifested as overlayfs failure to copy up lower
files from squashfs when selinux is built-in but not initialized,
because ovl_copy_xattr() iterates the lower inode xattrs by
vfs_listxattr() and vfs_getxattr().

ovl_copy_xattr() skips copy up of security labels that are indentified by
inode_copy_up_xattr LSM hooks, but it does that after vfs_getxattr().
Since we are not going to copy them, skip vfs_getxattr() of the security
labels.

Reported-by: Michael Labriola <[email protected]>
Tested-by: Michael Labriola <[email protected]>
Link: https://lore.kernel.org/linux-unionfs/[email protected]/
Signed-off-by: Amir Goldstein <[email protected]>
Signed-off-by: Miklos Szeredi <[email protected]>

ovl: fix dentry leak in ovl_get_redirect

We need to lock d_parent->d_lock before dget_dlock, or this may
have d_lockref updated parallelly like calltrace below which will
cause dentry->d_lockref leak and risk a crash.

     CPU 0                                CPU 1
ovl_set_redirect                       lookup_fast
  ovl_get_redirect                       __d_lookup
    dget_dlock
      //no lock protection here            spin_lock(&dentry->d_lock)
      dentry->d_lockref.count++            dentry->d_lockref.count++

[   49.799059] PGD 800000061fed7067 P4D 800000061fed7067 PUD 61fec5067 PMD 0
[   49.799689] Oops: 0002 [#1] SMP PTI
[   49.800019] CPU: 2 PID: 2332 Comm: node Not tainted 4.19.24-7.20.al7.x86_64 #1
[   49.800678] Hardware name: Alibaba Cloud Alibaba Cloud ECS, BIOS 8a46cfe 04/01/2014
[   49.801380] RIP: 0010:_raw_spin_lock+0xc/0x20
[   49.803470] RSP: 0018:ffffac6fc5417e98 EFLAGS: 00010246
[   49.803949] RAX: 0000000000000000 RBX: ffff93b8da3446c0 RCX: 0000000a00000000
[   49.804600] RDX: 0000000000000001 RSI: 000000000000000a RDI: 0000000000000088
[   49.805252] RBP: 0000000000000000 R08: 0000000000000000 R09: ffffffff993cf040
[   49.805898] R10: ffff93b92292e580 R11: ffffd27f188a4b80 R12: 0000000000000000
[   49.806548] R13: 00000000ffffff9c R14: 00000000fffffffe R15: ffff93b8da3446c0
[   49.807200] FS:  00007ffbedffb700(0000) GS:ffff93b927880000(0000) knlGS:0000000000000000
[   49.807935] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[   49.808461] CR2: 0000000000000088 CR3: 00000005e3f74006 CR4: 00000000003606a0
[   49.809113] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[   49.809758] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[   49.810410] Call Trace:
[   49.810653]  d_delete+0x2c/0xb0
[   49.810951]  vfs_rmdir+0xfd/0x120
[   49.811264]  do_rmdir+0x14f/0x1a0
[   49.811573]  do_syscall_64+0x5b/0x190
[   49.811917]  entry_SYSCALL_64_after_hwframe+0x44/0xa9
[   49.812385] RIP: 0033:0x7ffbf505ffd7
[   49.814404] RSP: 002b:00007ffbedffada8 EFLAGS: 00000297 ORIG_RAX: 0000000000000054
[   49.815098] RAX: ffffffffffffffda RBX: 00007ffbedffb640 RCX: 00007ffbf505ffd7
[   49.815744] RDX: 0000000004449700 RSI: 0000000000000000 RDI: 0000000006c8cd50
[   49.816394] RBP: 00007ffbedffaea0 R08: 0000000000000000 R09: 0000000000017d0b
[   49.817038] R10: 0000000000000000 R11: 0000000000000297 R12: 0000000000000012
[   49.817687] R13: 00000000072823d8 R14: 00007ffbedffb700 R15: 00000000072823d8
[   49.818338] Modules linked in: pvpanic cirrusfb button qemu_fw_cfg atkbd libps2 i8042
[   49.819052] CR2: 0000000000000088
[   49.819368] ---[ end trace 4e652b8aa299aa2d ]---
[   49.819796] RIP: 0010:_raw_spin_lock+0xc/0x20
[   49.821880] RSP: 0018:ffffac6fc5417e98 EFLAGS: 00010246
[   49.822363] RAX: 0000000000000000 RBX: ffff93b8da3446c0 RCX: 0000000a00000000
[   49.823008] RDX: 0000000000000001 RSI: 000000000000000a RDI: 0000000000000088
[   49.823658] RBP: 0000000000000000 R08: 0000000000000000 R09: ffffffff993cf040
[   49.825404] R10: ffff93b92292e580 R11: ffffd27f188a4b80 R12: 0000000000000000
[   49.827147] R13: 00000000ffffff9c R14: 00000000fffffffe R15: ffff93b8da3446c0
[   49.828890] FS:  00007ffbedffb700(0000) GS:ffff93b927880000(0000) knlGS:0000000000000000
[   49.830725] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[   49.832359] CR2: 0000000000000088 CR3: 00000005e3f74006 CR4: 00000000003606a0
[   49.834085] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[   49.835792] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400

Cc: <[email protected]>
Fixes: a6c606551141 ("ovl: redirect on rename-dir")
Signed-off-by: Liangyan <[email protected]>
Reviewed-by: Joseph Qi <[email protected]>
Suggested-by: Al Viro <[email protected]>
Signed-off-by: Miklos Szeredi <[email protected]>

ovl: avoid deadlock on directory ioctl

The function ovl_dir_real_file() currently uses the inode lock to serialize
writes to the od->upperfile field.

However, this function will get called by ovl_ioctl_set_flags(), which
utilizes the inode lock too.  In this case ovl_dir_real_file() will try to
claim a lock that is owned by a function in its call stack, which won't get
released before ovl_dir_real_file() returns.

Fix by replacing the open coded compare and exchange by an explicit atomic
op.

Fixes: 61536bed2149 ("ovl: support [S|G]ETFLAGS and FS[S|G]ETXATTR ioctls for directories")
Cc: [email protected] # v5.10
Reported-by: Icenowy Zheng <[email protected]>
Tested-by: Icenowy Zheng <[email protected]>
Signed-off-by: Miklos Szeredi <[email protected]>

cap: fix conversions on getxattr

If a capability is stored on disk in v2 format cap_inode_getsecurity() will
currently return in v2 format unconditionally.

This is wrong: v2 cap should be equivalent to a v3 cap with zero rootid,
and so the same conversions performed on it.

If the rootid cannot be mapped, v3 is returned unconverted.  Fix this so
that both v2 and v3 return -EOVERFLOW if the rootid (or the owner of the fs
user namespace in case of v2) cannot be mapped into the current user
namespace.

Signed-off-by: Miklos Szeredi <[email protected]>
Acked-by: "Eric W. Biederman" <[email protected]>

ovl: perform vfs_getxattr() with mounter creds

The vfs_getxattr() in ovl_xattr_set() is used to check whether an xattr
exist on a lower layer file that is to be removed.  If the xattr does not
exist, then no need to copy up the file.

This call of vfs_getxattr() wasn't wrapped in credential override, and this
is probably okay.  But for consitency wrap this instance as well.

Reported-by: "Eric W. Biederman" <[email protected]>
Signed-off-by: Miklos Szeredi <[email protected]>

ovl: add warning on user_ns mismatch

Currently there's no way to create an overlay filesystem outside of the
current user namespace.  Make sure that if this assumption changes it
doesn't go unnoticed.

Reported-by: "Eric W. Biederman" <[email protected]>
Signed-off-by: Miklos Szeredi <[email protected]>

ndtest: Add papr health related flags

sysfs attibutes to show health related flags are added.

Signed-off-by: Santosh Sivaraj <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Dan Williams <[email protected]>

ndtest: Add nvdimm control functions

Add functions to support ND_CMD_GET_CONFIG_SIZE, ND_CMD_SET_CONFIG_DATA and
ND_CMD_GET_CONFIG_DATA.

Signed-off-by: Santosh Sivaraj <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Dan Williams <[email protected]>

ndtest: Add regions and mappings to the test buses

The bus config array is used to hold the regions and the respective
mappings. This config based interface enables to change the
dimm/region/namespace layouts easily.

Signed-off-by: Santosh Sivaraj <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Dan Williams <[email protected]>

ndtest: Add dimm attributes

This patch adds sysfs attributes for nvdimm and the dimm device.

Signed-off-by: Santosh Sivaraj <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Dan Williams <[email protected]>

ndtest: Add dimms to the two buses

A config array is used to hold the dimms for each bus. These dimms are
registered with nvdimm, and new nvdimms are created on the buses.

Signed-off-by: Santosh Sivaraj <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Dan Williams <[email protected]>

ndtest: Add compatability string to treat it as PAPR family

Since this module is written to be platform agnostic, the module is made
part of the PAPR_FAMILY. ndctl identifies the family using the compatible
string inside of_node dir-entry.

Signed-off-by: Santosh Sivaraj <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Dan Williams <[email protected]>

testing/nvdimm: Add test module for non-nfit platforms

The current test module cannot be used for testing platforms (make check)
that do not have support for NFIT. In order to get the ndctl tests working,
we need a module which can emulate NVDIMM devices without relying on
ACPI/NFIT.

The aim of this proposed module is to implement a similar functionality to
the existing module but without the ACPI dependencies.

This RFC series is split into reviewable and compilable chunks.

This patch adds a new driver and registers two nvdimm bus needed for ndctl
make check.

Signed-off-by: Santosh Sivaraj <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Dan Williams <[email protected]>

libnvdimm/namespace: Fix visibility of namespace resource attribute

Legacy pmem namespaces lost support for the "resource" attribute when
the code was cleaned up to put the permission visibility in the
declaration. Restore this by listing 'resource' in the default
attributes.

A new ndctl regression test for pfn_to_online_page() corner cases builds
on this fix.

Fixes: bfd2e9140656 ("libnvdimm: Simplify root read-only definition for the 'resource' attribute")
Cc: Vishal Verma <[email protected]>
Cc: Dave Jiang <[email protected]>
Cc: Ira Weiny <[email protected]>
Cc: <[email protected]>
Link: https://lore.kernel.org/r/161052334995.1805594.12054873528154362921.stgit@dwillia2-desk3.amr.corp.intel.com
Signed-off-by: Dan Williams <[email protected]>

USB: serial: option: Adding support for Cinterion MV31

Adding support for Cinterion device MV31 for enumeration with
PID 0x00B3 and 0x00B7.

usb-devices output for 0x00B3
T:  Bus=04 Lev=01 Prnt=01 Port=00 Cnt=01 Dev#=  6 Spd=5000 MxCh= 0
D:  Ver= 3.20 Cls=ef(misc ) Sub=02 Prot=01 MxPS= 9 #Cfgs=  1
P:  Vendor=1e2d ProdID=00b3 Rev=04.14
S:  Manufacturer=Cinterion
S:  Product=Cinterion PID 0x00B3 USB Mobile Broadband
S:  SerialNumber=b3246eed
C:  #Ifs= 6 Cfg#= 1 Atr=a0 MxPwr=896mA
I:  If#=0x0 Alt= 0 #EPs= 1 Cls=02(commc) Sub=0e Prot=00 Driver=cdc_mbim
I:  If#=0x1 Alt= 1 #EPs= 2 Cls=0a(data ) Sub=00 Prot=02 Driver=cdc_mbim
I:  If#=0x2 Alt= 0 #EPs= 3 Cls=ff(vend.) Sub=00 Prot=00 Driver=option
I:  If#=0x3 Alt= 0 #EPs= 1 Cls=ff(vend.) Sub=ff Prot=ff Driver=cdc_wdm
I:  If#=0x4 Alt= 0 #EPs= 3 Cls=ff(vend.) Sub=00 Prot=00 Driver=option
I:  If#=0x5 Alt= 0 #EPs= 2 Cls=ff(vend.) Sub=ff Prot=30 Driver=option

usb-devices output for 0x00B7
T:  Bus=04 Lev=01 Prnt=01 Port=00 Cnt=01 Dev#=  5 Spd=5000 MxCh= 0
D:  Ver= 3.20 Cls=ef(misc ) Sub=02 Prot=01 MxPS= 9 #Cfgs=  1
P:  Vendor=1e2d ProdID=00b7 Rev=04.14
S:  Manufacturer=Cinterion
S:  Product=Cinterion PID 0x00B3 USB Mobile Broadband
S:  SerialNumber=b3246eed
C:  #Ifs= 4 Cfg#= 1 Atr=a0 MxPwr=896mA
I:  If#=0x0 Alt= 0 #EPs= 3 Cls=ff(vend.) Sub=ff Prot=ff Driver=qmi_wwan
I:  If#=0x1 Alt= 0 #EPs= 3 Cls=ff(vend.) Sub=00 Prot=00 Driver=option
I:  If#=0x2 Alt= 0 #EPs= 3 Cls=ff(vend.) Sub=00 Prot=00 Driver=option
I:  If#=0x3 Alt= 0 #EPs= 2 Cls=ff(vend.) Sub=ff Prot=30 Driver=option

Signed-off-by: Christoph Schemmel <[email protected]>
Cc: [email protected]
Signed-off-by: Johan Hovold <[email protected]>

KVM: x86: fix CPUID entries returned by KVM_GET_CPUID2 ioctl

Recent commit 255cbecfe0 modified struct kvm_vcpu_arch to make
'cpuid_entries' a pointer to an array of kvm_cpuid_entry2 entries
rather than embedding the array in the struct. KVM_SET_CPUID and
KVM_SET_CPUID2 were updated accordingly, but KVM_GET_CPUID2 was missed.

As a result, KVM_GET_CPUID2 currently returns random fields from struct
kvm_vcpu_arch to userspace rather than the expected CPUID values. Fix
this by treating 'cpuid_entries' as a pointer when copying its
contents to userspace buffer.

Fixes: 255cbecfe0c9 ("KVM: x86: allocate vcpu->arch.cpuid_entries dynamically")
Cc: Vitaly Kuznetsov <[email protected]>
Signed-off-by: Michael Roth <[email protected]>
Message-Id: <20210128024451.1816770[email protected]>
Cc: [email protected]
Signed-off-by: Paolo Bonzini <[email protected]>

Revert "checkpatch: add check for keyword 'boolean' in Kconfig definitions"

This reverts commit 327953e9af6c59ad111b28359e59e3ec0cbd71b6.

You cannot use 'boolean' since commit b92d804a5179 ("kconfig: drop
'boolean' keyword").

This check is no longer needed.

Signed-off-by: Masahiro Yamada <[email protected]>
Acked-by: Joe Perches <[email protected]>

scripts: use pkg-config to locate libcrypto

Otherwise build fails if the headers are not in the default location. While at
it also ask pkg-config for the libs, with fallback to the existing value.

Signed-off-by: Rolf Eike Beer <[email protected]>
Cc: [email protected] # 5.6.x
Signed-off-by: Masahiro Yamada <[email protected]>

Merge tag 'mlx5-fixes-2021-01-26' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux

Saeed Mahameed says:

====================
mlx5 fixes 2021-01-26

This series introduces some fixes to mlx5 driver.

* tag 'mlx5-fixes-2021-01-26' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux:
  net/mlx5: CT: Fix incorrect removal of tuple_nat_node from nat rhashtable
  net/mlx5e: Revert parameters on errors when changing MTU and LRO state without reset
  net/mlx5e: Revert parameters on errors when changing trust state without reset
  net/mlx5e: Correctly handle changing the number of queues when the interface is down
  net/mlx5e: Fix CT rule + encap slow path offload and deletion
  net/mlx5e: Disable hw-tc-offload when MLX5_CLS_ACT config is disabled
  net/mlx5: Maintain separate page trees for ECPF and PF functions
  net/mlx5e: Fix IPSEC stats
  net/mlx5e: Reduce tc unsupported key print level
  net/mlx5e: free page before return
  net/mlx5e: E-switch, Fix rate calculation for overflow
  net/mlx5: Fix memory leak on flow table creation error flow
====================

Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

ibmvnic: Ensure that CRQ entry read are correctly ordered

Ensure that received Command-Response Queue (CRQ) entries are
properly read in order by the driver. dma_rmb barrier has
been added before accessing the CRQ descriptor to ensure
the entire descriptor is read before processing.

Fixes: 032c5e82847a ("Driver for IBM System i/p VNIC protocol")
Signed-off-by: Lijun Pan <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

Merge git://git.kernel.org/pub/scm/linux/kernel/git/pablo/nf

Pablo Neira Ayuso says:

====================
Netfilter fixes for net

1) Honor stateful expressions defined in the set from the dynset
   extension. The set definition provides a stateful expression
   that must be used by the dynset expression in case it is specified.

2) Missing timeout extension in the set element in the dynset
   extension leads to inconsistent ruleset listing, not allowing
   the user to restore timeout and expiration on ruleset reload.

3) Do not dump the stateful expression from the dynset extension
   if it coming from the set definition.

* git://git.kernel.org/pub/scm/linux/kernel/git/pablo/nf:
  netfilter: nft_dynset: dump expressions when set definition contains no expressions
  netfilter: nft_dynset: add timeout extension to template
  netfilter: nft_dynset: honor stateful expressions in set definition
====================

Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

Merge tag 'linux-can-fixes-for-5.11-20210127' of git://git.kernel.org/pub/scm/linux/kernel/git/mkl/linux-can

Marc Kleine-Budde says:

====================
pull-request: can 2021-01-27

The patch is by Dan Carpenter and fixes a potential information leak in
can_fill_info().

* tag 'linux-can-fixes-for-5.11-20210127' of git://git.kernel.org/pub/scm/linux/kernel/git/mkl/linux-can:
  can: dev: prevent potential information leak in can_fill_info()
====================

Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

MAINTAINERS: add missing header for bonding

include/net/bonding.h is missing from bonding entry.

Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

Merge branch '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue

Anthony Nguyen says:

====================
Intel Wired LAN Driver Updates 2021-01-26

This series contains updates to the ice, i40e, and igc driver.

Henry corrects setting an unspecified protocol to IPPROTO_NONE instead of
0 for IPv6 flexbytes filters for ice.

Nick fixes the IPv6 extension header being processed incorrectly and
updates the netdev->dev_addr if it exists in hardware as it may have been
modified outside the ice driver.

Brett ensures a user cannot request more channels than available LAN MSI-X
and fixes the minimum allocation logic as it was incorrectly trying to use
more MSI-X than allocated for ice.

Stefan Assmann minimizes the delay between getting and using the VSI
pointer to prevent a possible crash for i40e.

Corinna Vinschen fixes link speed advertising for igc.

* '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue:
  igc: fix link speed advertising
  i40e: acquire VSI pointer only after VF is initialized
  ice: Fix MSI-X vector fallback logic
  ice: Don't allow more channels than LAN MSI-X available
  ice: update dev_addr in ice_set_mac_address even if HW filter exists
  ice: Implement flow for IPv6 next header (extension header)
  ice: fix FDir IPv6 flexbyte
====================

Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

net: decnet: fix netdev refcount leaking on error path

On building the route there is an assumption that the destination
could be local. In this case loopback_dev is used to get the address.
If the address is still cannot be retrieved dn_route_output_slow
returns EADDRNOTAVAIL with loopback_dev reference taken.

Cannot find hash for the fixes tag because this code was introduced
long time ago. I don't think that this bug has ever fired but the
patch is done just to have a consistent code base.

Signed-off-by: Vadim Fedorenko <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

net: switchdev: don't set port_obj_info->handled true when -EOPNOTSUPP

It's not true that switchdev_port_obj_notify() only inspects the
->handled field of "struct switchdev_notifier_port_obj_info" if
call_switchdev_blocking_notifiers() returns 0 - there's a WARN_ON()
triggering for a non-zero return combined with ->handled not being
true. But the real problem here is that -EOPNOTSUPP is not being
properly handled.

The wrapper functions switchdev_handle_port_obj_add() et al change a
return value of -EOPNOTSUPP to 0, and the treatment of ->handled in
switchdev_port_obj_notify() seems to be designed to change that back
to -EOPNOTSUPP in case nobody actually acted on the notifier (i.e.,
everybody returned -EOPNOTSUPP).

Currently, as soon as some device down the stack passes the check_cb()
check, ->handled gets set to true, which means that
switchdev_port_obj_notify() cannot actually ever return -EOPNOTSUPP.

This, for example, means that the detection of hardware offload
support in the MRP code is broken: switchdev_port_obj_add() used by
br_mrp_switchdev_send_ring_test() always returns 0, so since the MRP
code thinks the generation of MRP test frames has been offloaded, no
such frames are actually put on the wire. Similarly,
br_mrp_switchdev_set_ring_role() also always returns 0, causing
mrp->ring_role_offloaded to be set to 1.

To fix this, continue to set ->handled true if any callback returns
success or any error distinct from -EOPNOTSUPP. But if all the
callbacks return -EOPNOTSUPP, make sure that ->handled stays false, so
the logic in switchdev_port_obj_notify() can propagate that
information.

Fixes: 9a9f26e8f7ea ("bridge: mrp: Connect MRP API with the switchdev API")
Fixes: f30f0601eb93 ("switchdev: Add helpers to aid traversal through lower devices")
Reviewed-by: Petr Machata <[email protected]>
Signed-off-by: Rasmus Villemoes <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

of/device: Update dma_range_map only when dev has valid dma-ranges

The commit e0d072782c73 ("dma-mapping: introduce DMA range map,
supplanting dma_pfn_offset") always update dma_range_map even though it was
already set, like in the sunxi_mbus driver. the issue is reported at [1].
This patch avoid this(Updating it only when dev has valid dma-ranges).

Meanwhile, dma_range_map contains the devices' dma_ranges information,
This patch moves dma_range_map before of_iommu_configure. The iommu
driver may need to know the dma_address requirements of its iommu
consumer devices.

[1] https://lore.kernel.org/linux-arm-kernel/5c7946f3-b56e-da00-a750-be097c7ceb32@arm.com/

CC: Frank Rowand <[email protected]>
Fixes: e0d072782c73 ("dma-mapping: introduce DMA range map, supplanting dma_pfn_offset"),
Suggested-by: Robin Murphy <[email protected]>
Signed-off-by: Yong Wu <[email protected]>
Signed-off-by: Paul Kocialkowski <[email protected]>
Reviewed-by: Rob Herring <[email protected]>
Reviewed-by: Robin Murphy <[email protected]>
Signed-off-by: Rob Herring <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

Merge branch 'parisc-5.11-2' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/parisc-linux

Pull parisc fixes from Helge Deller:
 "Two small fixes:

   - Fix linking error with 64-bit kernel when modules are disabled,
     reported by kernel test robot

   - Remove leftover reference to power_tasklet, by Davidlohr Bueso"

* 'parisc-5.11-2' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/parisc-linux:
  parisc: Enable -mlong-calls gcc option by default when !CONFIG_MODULES
  parisc: Remove leftover reference to the power_tasklet

arm64: dts: meson: switch TFLASH_VDD_EN pin to open drain on Odroid-C4

For the proper reboot Odroid-C4 board requires to switch TFLASH_VDD_EN
pin to the high impedance mode, otherwise the board is stuck in the
middle of loading early stages of the bootloader from SD card.

This can be achieved by using the OPEN_DRAIN flag instead of the
ACTIVE_HIGH, what will leave the pin in input mode to achieve high state
(pin has the pull-up) and solve the issue.

Suggested-by: Neil Armstrong <[email protected]>
Fixes: 326e57518b0d ("arm64: dts: meson-sm1: add support for Hardkernel ODROID-C4")
Signed-off-by: Marek Szyprowski <[email protected]>
Acked-by: Martin Blumenstingl <[email protected]>
Signed-off-by: Kevin Hilman <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

dma-mapping: benchmark: fix kernel crash when dma_map_single fails

if dma_map_single() fails, kernel will give the below oops since
task_struct has been destroyed and we are running into the memory
corruption due to use-after-free in kthread_stop():

[   48.095310] Unable to handle kernel paging request at virtual address 000000c473548040
[   48.095736] Mem abort info:
[   48.095864]   ESR = 0x96000004
[   48.096025]   EC = 0x25: DABT (current EL), IL = 32 bits
[   48.096268]   SET = 0, FnV = 0
[   48.096401]   EA = 0, S1PTW = 0
[   48.096538] Data abort info:
[   48.096659]   ISV = 0, ISS = 0x00000004
[   48.096820]   CM = 0, WnR = 0
[   48.097079] user pgtable: 4k pages, 48-bit VAs, pgdp=0000000104639000
[   48.098099] [000000c473548040] pgd=0000000000000000, p4d=0000000000000000
[   48.098832] Internal error: Oops: 96000004 [#1] PREEMPT SMP
[   48.099232] Modules linked in:
[   48.099387] CPU: 0 PID: 2 Comm: kthreadd Tainted: G        W
[   48.099887] Hardware name: linux,dummy-virt (DT)
[   48.100078] pstate: 60000005 (nZCv daif -PAN -UAO -TCO BTYPE=--)
[   48.100516] pc : __kmalloc_node+0x214/0x368
[   48.100944] lr : __kmalloc_node+0x1f4/0x368
[   48.101458] sp : ffff800011f0bb80
[   48.101843] x29: ffff800011f0bb80 x28: ffff0000c0098ec0
[   48.102330] x27: 0000000000000000 x26: 00000000001d4600
[   48.102648] x25: ffff0000c0098ec0 x24: ffff800011b6a000
[   48.102988] x23: 00000000ffffffff x22: ffff0000c0098ec0
[   48.103333] x21: ffff8000101d7a54 x20: 0000000000000dc0
[   48.103657] x19: ffff0000c0001e00 x18: 0000000000000000
[   48.104069] x17: 0000000000000000 x16: 0000000000000000
[   48.105449] x15: 000001aa0304e7b9 x14: 00000000000003b1
[   48.106401] x13: ffff8000122d5000 x12: ffff80001228d000
[   48.107296] x11: ffff0000c0154340 x10: 0000000000000000
[   48.107862] x9 : ffff80000fffffff x8 : ffff0000c473527f
[   48.108326] x7 : ffff800011e62f58 x6 : ffff0000c01c8ed8
[   48.108778] x5 : ffff0000c0098ec0 x4 : 0000000000000000
[   48.109223] x3 : 00000000001d4600 x2 : 0000000000000040
[   48.109656] x1 : 0000000000000001 x0 : ff0000c473548000
[   48.110104] Call trace:
[   48.110287]  __kmalloc_node+0x214/0x368
[   48.110493]  __vmalloc_node_range+0xc4/0x298
[   48.110805]  copy_process+0x2c8/0x15c8
[   48.111133]  kernel_clone+0x5c/0x3c0
[   48.111373]  kernel_thread+0x64/0x90
[   48.111604]  kthreadd+0x158/0x368
[   48.111810]  ret_from_fork+0x10/0x30
[   48.112336] Code: 17ffffe9 b9402a62 b94008a1 11000421 (f8626802)
[   48.112884] ---[ end trace d4890e21e75419d5 ]---

Signed-off-by: Barry Song <[email protected]>
Signed-off-by: Christoph Hellwig <[email protected]>

io_uring: fix flush cqring overflow list while TASK_INTERRUPTIBLE

Abaci reported the follow warning:

[   27.073425] do not call blocking ops when !TASK_RUNNING; state=1 set at [] prepare_to_wait_exclusive+0x3a/0xc0
[   27.075805] WARNING: CPU: 0 PID: 951 at kernel/sched/core.c:7853 __might_sleep+0x80/0xa0
[   27.077604] Modules linked in:
[   27.078379] CPU: 0 PID: 951 Comm: a.out Not tainted 5.11.0-rc3+ #1
[   27.079637] Hardware name: Red Hat KVM, BIOS 0.5.1 01/01/2011
[   27.080852] RIP: 0010:__might_sleep+0x80/0xa0
[   27.081835] Code: 65 48 8b 04 25 80 71 01 00 48 8b 90 c0 15 00 00 48 8b 70 18 48 c7 c7 08 39 95 82 c6 05 f9 5f de 08 01 48 89 d1 e8 00 c6 fa ff  0b eb bf 41 0f b6 f5 48 c7 c7 40 23 c9 82 e8 f3 48 ec 00 eb a7
[   27.084521] RSP: 0018:ffffc90000fe3ce8 EFLAGS: 00010286
[   27.085350] RAX: 0000000000000000 RBX: ffffffff82956083 RCX: 0000000000000000
[   27.086348] RDX: ffff8881057a0000 RSI: ffffffff8118cc9e RDI: ffff88813bc28570
[   27.087598] RBP: 00000000000003a7 R08: 0000000000000001 R09: 0000000000000001
[   27.088819] R10: ffffc90000fe3e00 R11: 00000000fffef9f0 R12: 0000000000000000
[   27.089819] R13: 0000000000000000 R14: ffff88810576eb80 R15: ffff88810576e800
[   27.091058] FS:  00007f7b144cf740(0000) GS:ffff88813bc00000(0000) knlGS:0000000000000000
[   27.092775] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[   27.093796] CR2: 00000000022da7b8 CR3: 000000010b928002 CR4: 00000000003706f0
[   27.094778] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[   27.095780] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[   27.097011] Call Trace:
[   27.097685]  __mutex_lock+0x5d/0xa30
[   27.098565]  ? prepare_to_wait_exclusive+0x71/0xc0
[   27.099412]  ? io_cqring_overflow_flush.part.101+0x6d/0x70
[   27.100441]  ? lockdep_hardirqs_on_prepare+0xe9/0x1c0
[   27.101537]  ? _raw_spin_unlock_irqrestore+0x2d/0x40
[   27.102656]  ? trace_hardirqs_on+0x46/0x110
[   27.103459]  ? io_cqring_overflow_flush.part.101+0x6d/0x70
[   27.104317]  io_cqring_overflow_flush.part.101+0x6d/0x70
[   27.105113]  io_cqring_wait+0x36e/0x4d0
[   27.105770]  ? find_held_lock+0x28/0xb0
[   27.106370]  ? io_uring_remove_task_files+0xa0/0xa0
[   27.107076]  __x64_sys_io_uring_enter+0x4fb/0x640
[   27.107801]  ? rcu_read_lock_sched_held+0x59/0xa0
[   27.108562]  ? lockdep_hardirqs_on_prepare+0xe9/0x1c0
[   27.109684]  ? syscall_enter_from_user_mode+0x26/0x70
[   27.110731]  do_syscall_64+0x2d/0x40
[   27.111296]  entry_SYSCALL_64_after_hwframe+0x44/0xa9
[   27.112056] RIP: 0033:0x7f7b13dc8239
[   27.112663] Code: 01 00 48 81 c4 80 00 00 00 e9 f1 fe ff ff 0f 1f 00 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05  3d 01 f0 ff ff 73 01 c3 48 8b 0d 27 ec 2c 00 f7 d8 64 89 01 48
[   27.115113] RSP: 002b:00007ffd6d7f5c88 EFLAGS: 00000286 ORIG_RAX: 00000000000001aa
[   27.116562] RAX: ffffffffffffffda RBX: 0000000000000000 RCX: 00007f7b13dc8239
[   27.117961] RDX: 000000000000478e RSI: 0000000000000000 RDI: 0000000000000003
[   27.118925] RBP: 00007ffd6d7f5cb0 R08: 0000000020000040 R09: 0000000000000008
[   27.119773] R10: 0000000000000001 R11: 0000000000000286 R12: 0000000000400480
[   27.120614] R13: 00007ffd6d7f5d90 R14: 0000000000000000 R15: 0000000000000000
[   27.121490] irq event stamp: 5635
[   27.121946] hardirqs last  enabled at (5643): [] console_unlock+0x5c4/0x740
[   27.123476] hardirqs last disabled at (5652): [] console_unlock+0x4e7/0x740
[   27.125192] softirqs last  enabled at (5272): [] __do_softirq+0x3c5/0x5aa
[   27.126430] softirqs last disabled at (5267): [] asm_call_irq_on_stack+0xf/0x20
[   27.127634] ---[ end trace 289d7e28fa60f928 ]---

This is caused by calling io_cqring_overflow_flush() which may sleep
after calling prepare_to_wait_exclusive() which set task state to
TASK_INTERRUPTIBLE

Reported-by: Abaci <[email protected]>
Fixes: 6c503150ae33 ("io_uring: patch up IOPOLL overflow_flush sync")
Reviewed-by: Pavel Begunkov <[email protected]>
Signed-off-by: Hao Xu <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>

Revert "block: simplify set_init_blocksize" to regain lost performance

The cited commit introduced a serious regression with SATA write speed,
as found by bisecting. This patch reverts this commit, which restores
write speed back to the values observed before this commit.

The performance tests were done on a Helios4 NAS (2nd batch) with 4 HDDs
(WD8003FFBX) using dd (bs=1M count=2000). "Direct" is a test with a
single HDD, the rest are different RAID levels built over the first
partitions of 4 HDDs. Test results are in MB/s, R is read, W is write.

                | Direct | RAID0 | RAID10 f2 | RAID10 n2 | RAID6
----------------+--------+-------+-----------+-----------+--------
9011495c9466    | R:256  | R:313 | R:276     | R:313     | R:323
(before faulty) | W:254  | W:253 | W:195     | W:204     | W:117
----------------+--------+-------+-----------+-----------+--------
5ff9f19231a0    | R:257  | R:398 | R:312     | R:344     | R:391
(faulty commit) | W:154  | W:122 | W:67.7    | W:66.6    | W:67.2
----------------+--------+-------+-----------+-----------+--------
5.10.10         | R:256  | R:401 | R:312     | R:356     | R:375
unpatched       | W:149  | W:123 | W:64      | W:64.1    | W:61.5
----------------+--------+-------+-----------+-----------+--------
5.10.10         | R:255  | R:396 | R:312     | R:340     | R:393
patched         | W:247  | W:274 | W:220     | W:225     | W:121

Applying this patch doesn't hurt read performance, while improves the
write speed by 1.5x - 3.5x (more impact on RAID tests). The write speed
is restored back to the state before the faulty commit, and even a bit
higher in RAID tests (which aren't HDD-bound on this device) - that is
likely related to other optimizations done between the faulty commit and
5.10.10 which also improved the read speed.

Signed-off-by: Maxim Mikityanskiy <[email protected]>
Fixes: 5ff9f19231a0 ("block: simplify set_init_blocksize")
Cc: Christoph Hellwig <[email protected]>
Cc: Jens Axboe <[email protected]>
Acked-by: Christoph Hellwig <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>

Revert "PCI/ASPM: Save/restore L1SS Capability for suspend/resume"

This reverts commit 4257f7e008ea394fcecc050f1569c3503b8bcc15.

Kenneth reported that after 4257f7e008ea, he sees a torrent of disk I/O
errors on his NVMe device after suspend/resume until a reboot.

Link: https://lore.kernel.org/linux-pci/20201228040513.GA611645@bjorn-Precision-5520/
Reported-by: Kenneth R. Crudup <[email protected]>
Signed-off-by: Bjorn Helgaas <[email protected]>

gpiolib: cdev: clear debounce period if line set to output

When set_config changes a line from input to output debounce is
implicitly disabled, as debounce makes no sense for outputs, but the
debounce period is not being cleared and is still reported in the
line info.

So clear the debounce period when the debouncer is stopped in
edge_detector_stop().

Fixes: 65cff7046406 ("gpiolib: cdev: support setting debounce")
Cc: [email protected]
Signed-off-by: Kent Gibson <[email protected]>
Reviewed-by: Linus Walleij <[email protected]>
Signed-off-by: Bartosz Golaszewski <[email protected]>

x86/xen: avoid warning in Xen pv guest with CONFIG_AMD_MEM_ENCRYPT enabled

When booting a kernel which has been built with CONFIG_AMD_MEM_ENCRYPT
enabled as a Xen pv guest a warning is issued for each processor:

[    5.964347] ------------[ cut here ]------------
[    5.968314] WARNING: CPU: 0 PID: 1 at /home/gross/linux/head/arch/x86/xen/enlighten_pv.c:660 get_trap_addr+0x59/0x90
[    5.972321] Modules linked in:
[    5.976313] CPU: 0 PID: 1 Comm: swapper/0 Tainted: G        W         5.11.0-rc5-default #75
[    5.980313] Hardware name: Dell Inc. OptiPlex 9020/0PC5F7, BIOS A05 12/05/2013
[    5.984313] RIP: e030:get_trap_addr+0x59/0x90
[    5.988313] Code: 42 10 83 f0 01 85 f6 74 04 84 c0 75 1d b8 01 00 00 00 c3 48 3d 00 80 83 82 72 08 48 3d 20 81 83 82 72 0c b8 01 00 00 00 eb db <0f> 0b 31 c0 c3 48 2d 00 80 83 82 48 ba 72 1c c7 71 1c c7 71 1c 48
[    5.992313] RSP: e02b:ffffc90040033d38 EFLAGS: 00010202
[    5.996313] RAX: 0000000000000001 RBX: ffffffff82a141d0 RCX: ffffffff8222ec38
[    6.000312] RDX: ffffffff8222ec38 RSI: 0000000000000005 RDI: ffffc90040033d40
[    6.004313] RBP: ffff8881003984a0 R08: 0000000000000007 R09: ffff888100398000
[    6.008312] R10: 0000000000000007 R11: ffffc90040246000 R12: ffff8884082182a8
[    6.012313] R13: 0000000000000100 R14: 000000000000001d R15: ffff8881003982d0
[    6.016316] FS:  0000000000000000(0000) GS:ffff888408200000(0000) knlGS:0000000000000000
[    6.020313] CS:  e030 DS: 0000 ES: 0000 CR0: 0000000080050033
[    6.024313] CR2: ffffc900020ef000 CR3: 000000000220a000 CR4: 0000000000050660
[    6.028314] Call Trace:
[    6.032313]  cvt_gate_to_trap.part.7+0x3f/0x90
[    6.036313]  ? asm_exc_double_fault+0x30/0x30
[    6.040313]  xen_convert_trap_info+0x87/0xd0
[    6.044313]  xen_pv_cpu_up+0x17a/0x450
[    6.048313]  bringup_cpu+0x2b/0xc0
[    6.052313]  ? cpus_read_trylock+0x50/0x50
[    6.056313]  cpuhp_invoke_callback+0x80/0x4c0
[    6.060313]  _cpu_up+0xa7/0x140
[    6.064313]  cpu_up+0x98/0xd0
[    6.068313]  bringup_nonboot_cpus+0x4f/0x60
[    6.072313]  smp_init+0x26/0x79
[    6.076313]  kernel_init_freeable+0x103/0x258
[    6.080313]  ? rest_init+0xd0/0xd0
[    6.084313]  kernel_init+0xa/0x110
[    6.088313]  ret_from_fork+0x1f/0x30
[    6.092313] ---[ end trace be9ecf17dceeb4f3 ]---

Reason is that there is no Xen pv trap entry for X86_TRAP_VC.

Fix that by adding a generic trap handler for unknown traps and wire all
unknown bare metal handlers to this generic handler, which will just
crash the system in case such a trap will ever happen.

Fixes: 0786138c78e793 ("x86/sev-es: Add a Runtime #VC Exception Handler")
Cc: <[email protected]> # v5.10
Signed-off-by: Juergen Gross <[email protected]>
Reviewed-by: Andrew Cooper <[email protected]>
Signed-off-by: Juergen Gross <[email protected]>

ACPI/IORT: Do not blindly trust DMA masks from firmware

Address issue observed on real world system with suboptimal IORT table
where DMA masks of PCI devices would get set to 0 as result.

iort_dma_setup() would query the root complex'/named component IORT
entry for a DMA mask, and use that over the one the device has been
configured with earlier.

Ideally we want to use the minimum mask of what the IORT contains for
the root complex and what the device was configured with.

Fixes: 5ac65e8c8941 ("ACPI/IORT: Support address size limit for root complexes")
Signed-off-by: Moritz Fischer <[email protected]>
Reviewed-by: Robin Murphy <[email protected]>
Acked-by: Lorenzo Pieralisi <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Catalin Marinas <[email protected]>

s390: uv: Fix sysfs max number of VCPUs reporting

The number reported by the query is N-1 and I think people reading the
sysfs file would expect N instead. For users creating VMs there's no
actual difference because KVM's limit is currently below the UV's
limit.

Signed-off-by: Janosch Frank <[email protected]>
Fixes: a0f60f8431999 ("s390/protvirt: Add sysfs firmware interface for Ultravisor information")
Cc: [email protected]
Reviewed-by: Claudio Imbrenda <[email protected]>
Acked-by: Cornelia Huck <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>

s390/vfio-ap: No need to disable IRQ after queue reset

The queues assigned to a matrix mediated device are currently reset when:

* The VFIO_DEVICE_RESET ioctl is invoked
* The mdev fd is closed by userspace (QEMU)
* The mdev is removed from sysfs.

Immediately after the reset of a queue, a call is made to disable
interrupts for the queue. This is entirely unnecessary because the reset of
a queue disables interrupts, so this will be removed.

Furthermore, vfio_ap_irq_disable() does an unconditional PQAP/AQIC which
can result in a specification exception (when the corresponding facility
is not available), so this is actually a bugfix.

Signed-off-by: Tony Krowiak <[email protected]>
[[email protected]: minor rework before merging]
Signed-off-by: Halil Pasic <[email protected]>
Fixes: ec89b55e3bce ("s390: ap: implement PAPQ AQIC interception in kernel")
Cc: <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>

s390/vfio-ap: clean up vfio_ap resources when KVM pointer invalidated

The vfio_ap device driver registers a group notifier with VFIO when the
file descriptor for a VFIO mediated device for a KVM guest is opened to
receive notification that the KVM pointer is set (VFIO_GROUP_NOTIFY_SET_KVM
event). When the KVM pointer is set, the vfio_ap driver takes the
following actions:
1. Stashes the KVM pointer in the vfio_ap_mdev struct that holds the state
   of the mediated device.
2. Calls the kvm_get_kvm() function to increment its reference counter.
3. Sets the function pointer to the function that handles interception of
   the instruction that enables/disables interrupt processing.
4. Sets the masks in the KVM guest's CRYCB to pass AP resources through to
   the guest.

In order to avoid memory leaks, when the notifier is called to receive
notification that the KVM pointer has been set to NULL, the vfio_ap device
driver should reverse the actions taken when the KVM pointer was set.

Fixes: 258287c994de ("s390: vfio-ap: implement mediated device open callback")
Signed-off-by: Tony Krowiak <[email protected]>
Reviewed-by: Halil Pasic <[email protected]>
Reviewed-by: Cornelia Huck <[email protected]>
Cc: [email protected]
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Christian Borntraeger <[email protected]>
Signed-off-by: Vasily Gorbik <[email protected]>

can: dev: prevent potential information leak in can_fill_info()

The "bec" struct isn't necessarily always initialized. For example, the
mcp251xfd_get_berr_counter() function doesn't initialize anything if the
interface is down.

Fixes: 52c793f24054 ("can: netlink support for bus-error reporting and counters")
Link: https://lore.kernel.org/r/YAkaRdRJncsJO8Ve@mwanda
Signed-off-by: Dan Carpenter <[email protected]>
Signed-off-by: Marc Kleine-Budde <[email protected]>

rtc: mc146818: Detect and handle broken RTCs

The recent fix for handling the UIP bit unearthed another issue in the RTC
code. If the RTC is advertised but the readout is straight 0xFF because
it's not available, the old code just proceeded with crappy values, but the
new code hangs because it waits for the UIP bit to become low.

Add a sanity check in the RTC CMOS probe function which reads the RTC_VALID
register (Register D) which should have bit 0-6 cleared. If that's not the
case then fail to register the CMOS.

Add the same check to mc146818_get_time(), warn once when the condition
is true and invalidate the rtc_time data.

Reported-by: Mickaël Salaün <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
Tested-by: Mickaël Salaün <[email protected]>
Acked-by: Alexandre Belloni <[email protected]>
Link: https://lore.kernel.org/r/[email protected]

xen: Fix XenStore initialisation for XS_LOCAL

In commit 3499ba8198ca ("xen: Fix event channel callback via INTX/GSI")
I reworked the triggering of xenbus_probe().

I tried to simplify things by taking out the workqueue based startup
triggered from wake_waiting(); the somewhat poorly named xenbus IRQ
handler.

I missed the fact that in the XS_LOCAL case (Dom0 starting its own
xenstored or xenstore-stubdom, which happens after the kernel is booted
completely), that IRQ-based trigger is still actually needed.

So... put it back, except more cleanly. By just spawning a xenbus_probe
thread which waits on xb_waitq and runs the probe the first time it
gets woken, just as the workqueue-based hack did.

This is actually a nicer approach for *all* the back ends with different
interrupt methods, and we can switch them all over to that without the
complex conditions for when to trigger it. But not in -rc6. This is
the minimal fix for the regression, although it's a step in the right
direction instead of doing a partial revert and actually putting the
workqueue back. It's also simpler than the workqueue.

Fixes: 3499ba8198ca ("xen: Fix event channel callback via INTX/GSI")
Reported-by: Juergen Gross <[email protected]>
Signed-off-by: David Woodhouse <[email protected]>
Reviewed-by: Juergen Gross <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Juergen Gross <[email protected]>

io_uring: fix wqe->lock/completion_lock deadlock

Joseph reports following deadlock:

CPU0:
...
io_kill_linked_timeout  // &ctx->completion_lock
io_commit_cqring
__io_queue_deferred
__io_queue_async_work
io_wq_enqueue
io_wqe_enqueue  // &wqe->lock

CPU1:
...
__io_uring_files_cancel
io_wq_cancel_cb
io_wqe_cancel_pending_work  // &wqe->lock
io_cancel_task_cb  // &ctx->completion_lock

Only __io_queue_deferred() calls queue_async_work() while holding
ctx->completion_lock, enqueue drained requests via io_req_task_queue()
instead.

Cc: [email protected] # 5.9+
Reported-by: Joseph Qi <[email protected]>
Tested-by: Joseph Qi <[email protected]>
Signed-off-by: Pavel Begunkov <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>

Merge branch 'net-fec-fix-temporary-rmii-clock-reset-on-link-up'

Laurent Badel says:

====================
net: fec: Fix temporary RMII clock reset on link up

v2: fixed a compilation warning

The FEC drivers performs a "hardware reset" of the MAC module when the
link is reported to be up. This causes a short glitch in the RMII clock
due to the hardware reset clearing the receive control register which
controls the MII mode. It seems that some link partners do not tolerate
this glitch, and invalidate the link, which leads to a never-ending loop
of negotiation-link up-link down events.

This was observed with the iMX28 Soc and LAN8720/LAN8742 PHYs, with two
Intel adapters I218-LM and X722-DA2 as link partners, though a number of
other link partners do not seem to mind the clock glitch. Changing the
hardware reset to a software reset (clearing bit 1 of the ECR register)
cured the issue.

Attempts to optimize fec_restart() in order to minimize the duration of
the glitch were unsuccessful. Furthermore manually producing the glitch by
setting MII mode and then back to RMII in two consecutive instructions,
resulting in a clock glitch <10us in duration, was enough to cause the
partner to invalidate the link. This strongly suggests that the root cause
of the link being dropped is indeed the change in clock frequency.

In an effort to minimize changes to driver, the patch proposes to use
soft reset only for tested SoCs (iMX28) and only if the link is up. This
preserves hardware reset in other situations, which might be required for
proper setup of the MAC.
====================

Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

net: fec: Fix temporary RMII clock reset on link up

fec_restart() does a hard reset of the MAC module when the link status
changes to up. This temporarily resets the R_CNTRL register which controls
the MII mode of the ENET_OUT clock. In the case of RMII, the clock
frequency momentarily drops from 50MHz to 25MHz until the register is
reconfigured. Some link partners do not tolerate this glitch and
invalidate the link causing failure to establish a stable link when using
PHY polling mode. Since as per IEEE802.3 the criteria for link validity
are PHY-specific, what the partner should tolerate cannot be assumed, so
avoid resetting the MII clock by using software reset instead of hardware
reset when the link is up. This is generally relevant only if the SoC
provides the clock to an external PHY and the PHY is configured for RMII.

Signed-off-by: Laurent Badel <[email protected]>
Signed-off-by: Jakub Kicinski <[email protected]>

net: lapb: Add locking to the lapb module

In the lapb module, the timers may run concurrently with other code in
this module, and there is currently no locking to prevent the code from
racing on "struct lapb_cb". This patch adds locking to prevent racing.

1. Add "spinlock_t lock" to "struct lapb_cb"; Add "spin_lock_bh" and
"spin_unlock_bh" to APIs, timer functions and notifier functions.

2. Add "bool t1timer_stop, t2timer_stop" to "struct lapb_cb" to make us
able to ask running timers to abort; Modify "lapb_stop_t1timer" and
"lapb_stop_t2timer" to make them able to abort running timers;
Modify "lapb_t2timer_expiry" and "lapb_t1timer_expiry" to make them
abort after they are stopped by "lapb_stop_t1timer", "lapb_stop_t2timer",
and "lapb_start_t1timer", "lapb_start_t2timer".

3. Let lapb_unregister wait for other API functions and running timers
to stop.

4. The lapb_device_event function calls lapb_disconnect_request. In
order to avoid trying to hold the lock twice, add a new function named
"__lapb_disconnect_request" which assumes the lock is held, and make
it called by lapb_disconnect_request and lapb_device_event.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Cc: Martin Schiller <[email protected]>
Signed-off-by: Xie He <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

ARM: zImage: atags_to_fdt: Fix node names on added root nodes

Commit 7536c7e03e74 ("of/fdt: Remove redundant kbasename function
call") exposed a bug creating DT nodes in the ATAGS to DT fixup code.
Non-existent nodes would mistaken get created with a leading '/'. The
problem was fdt_path_offset() takes a full path while creating a node
with fdt_add_subnode() takes just the basename.

Since this we only add root child nodes, we can just skip over the '/'.

Fixes: 7536c7e03e74 ("of/fdt: Remove redundant kbasename function call")
Reported-by: Chris Packham <[email protected]>
Cc: Qi Zheng <[email protected]>
Cc: Russell King <[email protected]>
Signed-off-by: Rob Herring <[email protected]>
Tested-by: Chris Packham <[email protected]>
Link: https://lore.kernel.org/r/[email protected]