linux-stable.git/drivers/net/ethernet/intel, branch v4.9.331

ixgbe: stop resetting SYSTIME in ixgbe_ptp_start_cyclecounter

2022-09-05T08:23:56+00:00

[ Upstream commit 25d7a5f5a6bb15a2dae0a3f39ea5dda215024726 ]

The ixgbe_ptp_start_cyclecounter is intended to be called whenever the
cyclecounter parameters need to be changed.

Since commit a9763f3cb54c ("ixgbe: Update PTP to support X550EM_x
devices"), this function has cleared the SYSTIME registers and reset the
TSAUXC DISABLE_SYSTIME bit.

While these need to be cleared during ixgbe_ptp_reset, it is wrong to clear
them during ixgbe_ptp_start_cyclecounter. This function may be called
during both reset and link status change. When link changes, the SYSTIME
counter is still operating normally, but the cyclecounter should be updated
to account for the possibly changed parameters.

Clearing SYSTIME when link changes causes the timecounter to jump because
the cycle counter now reads zero.

Extract the SYSTIME initialization out to a new function and call this
during ixgbe_ptp_reset. This prevents the timecounter adjustment and avoids
an unnecessary reset of the current time.

This also restores the original SYSTIME clearing that occurred during
ixgbe_ptp_reset before the commit above.

Reported-by: Steve Payne 
Reported-by: Ilya Evenbach 
Fixes: a9763f3cb54c ("ixgbe: Update PTP to support X550EM_x devices")
Signed-off-by: Jacob Keller 
Tested-by: Gurucharan  (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen 
Signed-off-by: Sasha Levin

igb: Make DMA faster when CPU is active on the PCIe link

2022-07-02T14:17:15+00:00

[ Upstream commit 4e0effd9007ea0be31f7488611eb3824b4541554 ]

Intel I210 on some Intel Alder Lake platforms can only achieve ~750Mbps
Tx speed via iperf. The RR2DCDELAY shows around 0x2xxx DMA delay, which
will be significantly lower when 1) ASPM is disabled or 2) SoC package
c-state stays above PC3. When the RR2DCDELAY is around 0x1xxx the Tx
speed can reach to ~950Mbps.

According to the I210 datasheet "8.26.1 PCIe Misc. Register - PCIEMISC",
"DMA Idle Indication" doesn't seem to tie to DMA coalesce anymore, so
set it to 1b for "DMA is considered idle when there is no Rx or Tx AND
when there are no TLPs indicating that CPU is active detected on the
PCIe link (such as the host executes CSR or Configuration register read
or write operation)" and performing Tx should also fall under "active
CPU on PCIe link" case.

In addition to that, commit b6e0c419f040 ("igb: Move DMA Coalescing init
code to separate function.") seems to wrongly changed from enabling
E1000_PCIEMISC_LX_DECISION to disabling it, also fix that.

Fixes: b6e0c419f040 ("igb: Move DMA Coalescing init code to separate function.")
Signed-off-by: Kai-Heng Feng 
Tested-by: Gurucharan  (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen 
Link: https://lore.kernel.org/r/20220621221056.604304-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski 
Signed-off-by: Sasha Levin

igb: skip phy status check where unavailable

2022-05-25T06:39:50+00:00

[ Upstream commit 942d2ad5d2e0df758a645ddfadffde2795322728 ]

igb_read_phy_reg() will silently return, leaving phy_data untouched, if
hw->ops.read_reg isn't set. Depending on the uninitialized value of
phy_data, this led to the phy status check either succeeding immediately
or looping continuously for 2 seconds before emitting a noisy err-level
timeout. This message went out to the console even though there was no
actual problem.

Instead, first check if there is read_reg function pointer. If not,
proceed without trying to check the phy status register.

Fixes: b72f3f72005d ("igb: When GbE link up, wait for Remote receiver status condition")
Signed-off-by: Kevin Mitchell 
Tested-by: Gurucharan  (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen 
Signed-off-by: David S. Miller 
Signed-off-by: Sasha Levin

e1000e: Fix possible overflow in LTR decoding

2022-04-27T11:14:11+00:00

commit 04ebaa1cfddae5f240cc7404f009133bb0389a47 upstream.

When we decode the latency and the max_latency, u16 value may not fit
the required size and could lead to the wrong LTR representation.

Scaling is represented as:
scale 0 - 1         (2^(5*0)) = 2^0
scale 1 - 32        (2^(5 *1))= 2^5
scale 2 - 1024      (2^(5 *2)) =2^10
scale 3 - 32768     (2^(5 *3)) =2^15
scale 4 - 1048576   (2^(5 *4)) = 2^20
scale 5 - 33554432  (2^(5 *4)) = 2^25
scale 4 and scale 5 required 20 and 25 bits respectively.
scale 6 reserved.

Replace the u16 type with the u32 type and allow corrected LTR
representation.

Cc: stable@vger.kernel.org
Fixes: 44a13a5d99c7 ("e1000e: Fix the max snoop/no-snoop latency for 10M")
Reported-by: James Hutchinson 
Link: https://bugzilla.kernel.org/show_bug.cgi?id=215689
Suggested-by: Dima Ruinskiy 
Signed-off-by: Sasha Neftin 
Tested-by: Naama Meir 
Tested-by: James Hutchinson 
Signed-off-by: Tony Nguyen 
Signed-off-by: Greg Kroah-Hartman

i40e: Fix incorrect netdev's real number of RX/TX queues

2022-01-11T12:38:12+00:00

commit e738451d78b2f8a9635d66c6a87f304b4d965f7a upstream.

There was a wrong queues representation in sysfs during
driver's reinitialization in case of online cpus number is
less than combined queues. It was caused by stopped
NetworkManager, which is responsible for calling vsi_open
function during driver's initialization.
In specific situation (ex. 12 cpus online) there were 16 queues
in /sys/class/net//queues. In case of modifying queues with
value higher, than number of online cpus, then it caused write
errors and other errors.
Add updating of sysfs's queues representation during driver
initialization.

Fixes: 41c445ff0f48 ("i40e: main driver core")
Signed-off-by: Lukasz Cieplicki 
Signed-off-by: Jedrzej Jagielski 
Tested-by: Gurucharan G 
Signed-off-by: Tony Nguyen 
Signed-off-by: Greg Kroah-Hartman

ixgbe: set X550 MDIO speed before talking to PHY

2021-12-22T08:05:14+00:00

[ Upstream commit bf0a375055bd1afbbf02a0ef45f7655da7b71317 ]

The MDIO bus speed must be initialized before talking to the PHY the first
time in order to avoid talking to it using a speed that the PHY doesn't
support.

This fixes HW initialization error -17 (IXGBE_ERR_PHY_ADDR_INVALID) on
Denverton CPUs (a.k.a. the Atom C3000 family) on ports with a 10Gb network
plugged in. On those devices, HLREG0[MDCSPD] resets to 1, which combined
with the 10Gb network results in a 24MHz MDIO speed, which is apparently
too fast for the connected PHY. PHY register reads over MDIO bus return
garbage, leading to initialization failure.

Reproduced with Linux kernel 4.19 and 5.15-rc7. Can be reproduced using
the following setup:

* Use an Atom C3000 family system with at least one X552 LAN on the SoC
* Disable PXE or other BIOS network initialization if possible
  (the interface must not be initialized before Linux boots)
* Connect a live 10Gb Ethernet cable to an X550 port
* Power cycle (not reset, doesn't always work) the system and boot Linux
* Observe: ixgbe interfaces w/ 10GbE cables plugged in fail with error -17

Fixes: e84db7272798 ("ixgbe: Introduce function to control MDIO speed")
Signed-off-by: Cyril Novikov 
Reviewed-by: Andrew Lunn 
Signed-off-by: Tony Nguyen 
Signed-off-by: Sasha Levin

igbvf: fix double free in `igbvf_probe`

2021-12-22T08:05:14+00:00

[ Upstream commit b6d335a60dc624c0d279333b22c737faa765b028 ]

In `igbvf_probe`, if register_netdev() fails, the program will go to
label err_hw_init, and then to label err_ioremap. In free_netdev() which
is just below label err_ioremap, there is `list_for_each_entry_safe` and
`netif_napi_del` which aims to delete all entries in `dev->napi_list`.
The program has added an entry `adapter->rx_ring->napi` which is added by
`netif_napi_add` in igbvf_alloc_queues(). However, adapter->rx_ring has
been freed below label err_hw_init. So this a UAF.

In terms of how to patch the problem, we can refer to igbvf_remove() and
delete the entry before `adapter->rx_ring`.

The KASAN logs are as follows:

[   35.126075] BUG: KASAN: use-after-free in free_netdev+0x1fd/0x450
[   35.127170] Read of size 8 at addr ffff88810126d990 by task modprobe/366
[   35.128360]
[   35.128643] CPU: 1 PID: 366 Comm: modprobe Not tainted 5.15.0-rc2+ #14
[   35.129789] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.12.0-59-gc9ba5276e321-prebuilt.qemu.org 04/01/2014
[   35.131749] Call Trace:
[   35.132199]  dump_stack_lvl+0x59/0x7b
[   35.132865]  print_address_description+0x7c/0x3b0
[   35.133707]  ? free_netdev+0x1fd/0x450
[   35.134378]  __kasan_report+0x160/0x1c0
[   35.135063]  ? free_netdev+0x1fd/0x450
[   35.135738]  kasan_report+0x4b/0x70
[   35.136367]  free_netdev+0x1fd/0x450
[   35.137006]  igbvf_probe+0x121d/0x1a10 [igbvf]
[   35.137808]  ? igbvf_vlan_rx_add_vid+0x100/0x100 [igbvf]
[   35.138751]  local_pci_probe+0x13c/0x1f0
[   35.139461]  pci_device_probe+0x37e/0x6c0
[   35.165526]
[   35.165806] Allocated by task 366:
[   35.166414]  ____kasan_kmalloc+0xc4/0xf0
[   35.167117]  foo_kmem_cache_alloc_trace+0x3c/0x50 [igbvf]
[   35.168078]  igbvf_probe+0x9c5/0x1a10 [igbvf]
[   35.168866]  local_pci_probe+0x13c/0x1f0
[   35.169565]  pci_device_probe+0x37e/0x6c0
[   35.179713]
[   35.179993] Freed by task 366:
[   35.180539]  kasan_set_track+0x4c/0x80
[   35.181211]  kasan_set_free_info+0x1f/0x40
[   35.181942]  ____kasan_slab_free+0x103/0x140
[   35.182703]  kfree+0xe3/0x250
[   35.183239]  igbvf_probe+0x1173/0x1a10 [igbvf]
[   35.184040]  local_pci_probe+0x13c/0x1f0

Fixes: d4e0fe01a38a0 (igbvf: add new driver to support 82576 virtual functions)
Reported-by: Zheyu Ma 
Signed-off-by: Letu Ren 
Tested-by: Konrad Jankowski 
Signed-off-by: Tony Nguyen 
Signed-off-by: Sasha Levin

iavf: Fix for the false positive ASQ/ARQ errors while issuing VF reset

2021-11-26T10:48:42+00:00

[ Upstream commit 321421b57a12e933f92b228e0e6d0b2c6541f41d ]

While issuing VF Reset from the guest OS, the VF driver prints
logs about critical / Overflow error detection. This is not an
actual error since the VF_MBX_ARQLEN register is set to all FF's
for a short period of time and the VF would catch the bits set if
it was reading the register during that spike of time.
This patch introduces an additional check to ignore this condition
since the VF is in reset.

Fixes: 19b73d8efaa4 ("i40evf: Add additional check for reset")
Signed-off-by: Surabhi Boob 
Tested-by: Tony Brelinski 
Signed-off-by: Tony Nguyen 
Signed-off-by: Sasha Levin

i40e: fix endless loop under rtnl

2021-10-17T08:05:40+00:00

[ Upstream commit 857b6c6f665cca9828396d9743faf37fd09e9ac3 ]

The loop in i40e_get_capabilities can never end. The problem is that
although i40e_aq_discover_capabilities returns with an error if there's
a firmware problem, the returned error is not checked. There is a check for
pf->hw.aq.asq_last_status but that value is set to I40E_AQ_RC_OK on most
firmware problems.

When i40e_aq_discover_capabilities encounters a firmware problem, it will
encounter the same problem on its next invocation. As the result, the loop
becomes endless. We hit this with I40E_ERR_ADMIN_QUEUE_TIMEOUT but looking
at the code, it can happen with a range of other firmware errors.

I don't know what the correct behavior should be: whether the firmware
should be retried a few times, or whether pf->hw.aq.asq_last_status should
be always set to the encountered firmware error (but then it would be
pointless and can be just replaced by the i40e_aq_discover_capabilities
return value). However, the current behavior with an endless loop under the
rtnl mutex(!) is unacceptable and Intel has not submitted a fix, although we
explained the bug to them 7 months ago.

This may not be the best possible fix but it's better than hanging the whole
system on a firmware bug.

Fixes: 56a62fc86895 ("i40e: init code and hardware support")
Tested-by: Stefan Assmann 
Signed-off-by: Jiri Benc 
Reviewed-by: Jesse Brandeburg 
Tested-by: Dave Switzer 
Signed-off-by: Tony Nguyen 
Signed-off-by: Sasha Levin

e100: fix buffer overrun in e100_get_regs

2021-10-06T08:23:41+00:00

[ Upstream commit 51032e6f17ce990d06123ad7307f258c50d25aa7 ]

The e100_get_regs function is used to implement a simple register dump
for the e100 device. The data is broken into a couple of MAC control
registers, and then a series of PHY registers, followed by a memory dump
buffer.

The total length of the register dump is defined as (1 + E100_PHY_REGS)
* sizeof(u32) + sizeof(nic->mem->dump_buf).

The logic for filling in the PHY registers uses a convoluted inverted
count for loop which counts from E100_PHY_REGS (0x1C) down to 0, and
assigns the slots 1 + E100_PHY_REGS - i. The first loop iteration will
fill in [1] and the final loop iteration will fill in [1 + 0x1C]. This
is actually one more than the supposed number of PHY registers.

The memory dump buffer is then filled into the space at
[2 + E100_PHY_REGS] which will cause that memcpy to assign 4 bytes past
the total size.

The end result is that we overrun the total buffer size allocated by the
kernel, which could lead to a panic or other issues due to memory
corruption.

It is difficult to determine the actual total number of registers
here. The only 8255x datasheet I could find indicates there are 28 total
MDI registers. However, we're reading 29 here, and reading them in
reverse!

In addition, the ethtool e100 register dump interface appears to read
the first PHY register to determine if the device is in MDI or MDIx
mode. This doesn't appear to be documented anywhere within the 8255x
datasheet. I can only assume it must be in register 28 (the extra
register we're reading here).

Lets not change any of the intended meaning of what we copy here. Just
extend the space by 4 bytes to account for the extra register and
continue copying the data out in the same order.

Change the E100_PHY_REGS value to be the correct total (29) so that the
total register dump size is calculated properly. Fix the offset for
where we copy the dump buffer so that it doesn't overrun the total size.

Re-write the for loop to use counting up instead of the convoluted
down-counting. Correct the mdio_read offset to use the 0-based register
offsets, but maintain the bizarre reverse ordering so that we have the
ABI expected by applications like ethtool. This requires and additional
subtraction of 1. It seems a bit odd but it makes the flow of assignment
into the register buffer easier to follow.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Reported-by: Felicitas Hetzelt 
Signed-off-by: Jacob Keller 
Tested-by: Jacob Keller 
Signed-off-by: Tony Nguyen 
Signed-off-by: Sasha Levin