platform/kernel/linux-rpi.git
3 years agonet/smc: add missing error check in smc_clc_prfx_set()
Karsten Graul [Mon, 20 Sep 2021 19:18:14 +0000 (21:18 +0200)]
net/smc: add missing error check in smc_clc_prfx_set()

Coverity stumbled over a missing error check in smc_clc_prfx_set():

*** CID 1475954:  Error handling issues  (CHECKED_RETURN)
/net/smc/smc_clc.c: 233 in smc_clc_prfx_set()
>>>     CID 1475954:  Error handling issues  (CHECKED_RETURN)
>>>     Calling "kernel_getsockname" without checking return value (as is done elsewhere 8 out of 10 times).
233      kernel_getsockname(clcsock, (struct sockaddr *)&addrs);

Add the return code check in smc_clc_prfx_set().

Fixes: c246d942eabc ("net/smc: restructure netinfo for CLC proposal msgs")
Reported-by: Julian Wiedmann <jwi@linux.ibm.com>
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge branch 'hns3-fixes'
David S. Miller [Mon, 20 Sep 2021 12:28:50 +0000 (13:28 +0100)]
Merge branch 'hns3-fixes'

Guangbin Huang says:

====================
net: hns3: add some fixes for -net

This series adds some fixes for the HNS3 ethernet driver.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: hns3: fix a return value error in hclge_get_reset_status()
Yufeng Mo [Wed, 15 Sep 2021 13:52:11 +0000 (21:52 +0800)]
net: hns3: fix a return value error in hclge_get_reset_status()

hclge_get_reset_status() should return the tqp reset status.
However, if the CMDQ fails, the caller will take it as tqp reset
success status by mistake. Therefore, uses a parameters to get
the tqp reset status instead.

Fixes: 46a3df9f9718 ("net: hns3: Add HNS3 Acceleration Engine & Compatibility Layer Support")
Signed-off-by: Yufeng Mo <moyufeng@huawei.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: hns3: check vlan id before using it
liaoguojia [Wed, 15 Sep 2021 13:52:10 +0000 (21:52 +0800)]
net: hns3: check vlan id before using it

The input parameters may not be reliable, so check the vlan id before
using it, otherwise may set wrong vlan id into hardware.

Fixes: dc8131d846d4 ("net: hns3: Fix for packet loss due wrong filter config in VLAN tbls")
Signed-off-by: liaoguojia <liaoguojia@huawei.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: hns3: check queue id range before using
Yufeng Mo [Wed, 15 Sep 2021 13:52:09 +0000 (21:52 +0800)]
net: hns3: check queue id range before using

The input parameters may not be reliable. Before using the
queue id, we should check this parameter. Otherwise, memory
overwriting may occur.

Fixes: d34100184685 ("net: hns3: refactor the mailbox message between PF and VF")
Signed-off-by: Yufeng Mo <moyufeng@huawei.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: hns3: fix misuse vf id and vport id in some logs
Jiaran Zhang [Wed, 15 Sep 2021 13:52:08 +0000 (21:52 +0800)]
net: hns3: fix misuse vf id and vport id in some logs

vport_id include PF and VFs, vport_id = 0 means PF, other values mean VFs.
So the actual vf id is equal to vport_id minus 1.

Some VF print logs are actually vport, and logs of vf id actually use
vport id, so this patch fixes them.

Fixes: ac887be5b0fe ("net: hns3: change print level of RAS error log from warning to error")
Fixes: adcf738b804b ("net: hns3: cleanup some print format warning")
Signed-off-by: Jiaran Zhang <zhangjiaran@huawei.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: hns3: fix inconsistent vf id print
Jian Shen [Wed, 15 Sep 2021 13:52:07 +0000 (21:52 +0800)]
net: hns3: fix inconsistent vf id print

The vf id from ethtool is added 1 before configured to driver.
So it's necessary to minus 1 when printing it, in order to
keep consistent with user's configuration.

Fixes: dd74f815dd41 ("net: hns3: Add support for rule add/delete for flow director")
Signed-off-by: Jian Shen <shenjian15@huawei.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: hns3: fix change RSS 'hfunc' ineffective issue
Jian Shen [Wed, 15 Sep 2021 13:52:06 +0000 (21:52 +0800)]
net: hns3: fix change RSS 'hfunc' ineffective issue

When user change rss 'hfunc' without set rss 'hkey' by ethtool
-X command, the driver will ignore the 'hfunc' for the hkey is
NULL. It's unreasonable. So fix it.

Fixes: 46a3df9f9718 ("net: hns3: Add HNS3 Acceleration Engine & Compatibility Layer Support")
Fixes: 374ad291762a ("net: hns3: Add RSS general configuration support for VF")
Signed-off-by: Jian Shen <shenjian15@huawei.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoptp: ocp: add COMMON_CLK dependency
Arnd Bergmann [Mon, 20 Sep 2021 09:57:49 +0000 (11:57 +0200)]
ptp: ocp: add COMMON_CLK dependency

Without CONFIG_COMMON_CLK, this fails to link:

arm-linux-gnueabi-ld: drivers/ptp/ptp_ocp.o: in function `ptp_ocp_register_i2c':
ptp_ocp.c:(.text+0xcc0): undefined reference to `__clk_hw_register_fixed_rate'
arm-linux-gnueabi-ld: ptp_ocp.c:(.text+0xcf4): undefined reference to `devm_clk_hw_register_clkdev'
arm-linux-gnueabi-ld: drivers/ptp/ptp_ocp.o: in function `ptp_ocp_detach':
ptp_ocp.c:(.text+0x1c24): undefined reference to `clk_hw_unregister_fixed_rate'

Fixes: a7e1abad13f3 ("ptp: Add clock driver for the OpenCompute TimeCard.")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agobnxt_en: Fix TX timeout when TX ring size is set to the smallest
Michael Chan [Mon, 20 Sep 2021 06:51:52 +0000 (02:51 -0400)]
bnxt_en: Fix TX timeout when TX ring size is set to the smallest

The smallest TX ring size we support must fit a TX SKB with MAX_SKB_FRAGS
+ 1.  Because the first TX BD for a packet is always a long TX BD, we
need an extra TX BD to fit this packet.  Define BNXT_MIN_TX_DESC_CNT with
this value to make this more clear.  The current code uses a minimum
that is off by 1.  Fix it using this constant.

The tx_wake_thresh to determine when to wake up the TX queue is half the
ring size but we must have at least BNXT_MIN_TX_DESC_CNT for the next
packet which may have maximum fragments.  So the comparison of the
available TX BDs with tx_wake_thresh should be >= instead of > in the
current code.  Otherwise, at the smallest ring size, we will never wake
up the TX queue and will cause TX timeout.

Fixes: c0c050c58d84 ("bnxt_en: New Broadcom ethernet driver.")
Reviewed-by: Pavan Chebbi <pavan.chebbi@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadocm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonexthop: Fix division by zero while replacing a resilient group
Ido Schimmel [Fri, 17 Sep 2021 13:02:18 +0000 (16:02 +0300)]
nexthop: Fix division by zero while replacing a resilient group

The resilient nexthop group torture tests in fib_nexthop.sh exposed a
possible division by zero while replacing a resilient group [1]. The
division by zero occurs when the data path sees a resilient nexthop
group with zero buckets.

The tests replace a resilient nexthop group in a loop while traffic is
forwarded through it. The tests do not specify the number of buckets
while performing the replacement, resulting in the kernel allocating a
stub resilient table (i.e, 'struct nh_res_table') with zero buckets.

This table should never be visible to the data path, but the old nexthop
group (i.e., 'oldg') might still be used by the data path when the stub
table is assigned to it.

Fix this by only assigning the stub table to the old nexthop group after
making sure the group is no longer used by the data path.

Tested with fib_nexthops.sh:

Tests passed: 222
Tests failed:   0

[1]
 divide error: 0000 [#1] PREEMPT SMP KASAN
 CPU: 0 PID: 1850 Comm: ping Not tainted 5.14.0-custom-10271-ga86eb53057fe #1107
 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.14.0-4.fc34 04/01/2014
 RIP: 0010:nexthop_select_path+0x2d2/0x1a80
[...]
 Call Trace:
  fib_select_multipath+0x79b/0x1530
  fib_select_path+0x8fb/0x1c10
  ip_route_output_key_hash_rcu+0x1198/0x2da0
  ip_route_output_key_hash+0x190/0x340
  ip_route_output_flow+0x21/0x120
  raw_sendmsg+0x91d/0x2e10
  inet_sendmsg+0x9e/0xe0
  __sys_sendto+0x23d/0x360
  __x64_sys_sendto+0xe1/0x1b0
  do_syscall_64+0x35/0x80
  entry_SYSCALL_64_after_hwframe+0x44/0xae

Cc: stable@vger.kernel.org
Fixes: 283a72a5599e ("nexthop: Add implementation of resilient next-hop groups")
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonapi: fix race inside napi_enable
Xuan Zhuo [Sat, 18 Sep 2021 08:52:32 +0000 (16:52 +0800)]
napi: fix race inside napi_enable

The process will cause napi.state to contain NAPI_STATE_SCHED and
not in the poll_list, which will cause napi_disable() to get stuck.

The prefix "NAPI_STATE_" is removed in the figure below, and
NAPI_STATE_HASHED is ignored in napi.state.

                      CPU0       |                   CPU1       | napi.state
===============================================================================
napi_disable()                   |                              | SCHED | NPSVC
napi_enable()                    |                              |
{                                |                              |
    smp_mb__before_atomic();     |                              |
    clear_bit(SCHED, &n->state); |                              | NPSVC
                                 | napi_schedule_prep()         | SCHED | NPSVC
                                 | napi_poll()                  |
                                 |   napi_complete_done()       |
                                 |   {                          |
                                 |      if (n->state & (NPSVC | | (1)
                                 |               _BUSY_POLL)))  |
                                 |           return false;      |
                                 |     ................         |
                                 |   }                          | SCHED | NPSVC
                                 |                              |
    clear_bit(NPSVC, &n->state); |                              | SCHED
}                                |                              |
                                 |                              |
napi_schedule_prep()             |                              | SCHED | MISSED (2)

(1) Here return direct. Because of NAPI_STATE_NPSVC exists.
(2) NAPI_STATE_SCHED exists. So not add napi.poll_list to sd->poll_list

Since NAPI_STATE_SCHED already exists and napi is not in the
sd->poll_list queue, NAPI_STATE_SCHED cannot be cleared and will always
exist.

1. This will cause this queue to no longer receive packets.
2. If you encounter napi_disable under the protection of rtnl_lock, it
   will cause the entire rtnl_lock to be locked, affecting the overall
   system.

This patch uses cmpxchg to implement napi_enable(), which ensures that
there will be no race due to the separation of clear two bits.

Fixes: 2d8bff12699abc ("netpoll: Close race condition between poll_one_napi and napi_disable")
Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Reviewed-by: Dust Li <dust.li@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoselftests: net: af_unix: Fix makefile to use TEST_GEN_PROGS
Shuah Khan [Fri, 17 Sep 2021 21:53:56 +0000 (15:53 -0600)]
selftests: net: af_unix: Fix makefile to use TEST_GEN_PROGS

Makefile uses TEST_PROGS instead of TEST_GEN_PROGS to define
executables. TEST_PROGS is for shell scripts that need to be
installed and run by the common lib.mk framework. The common
framework doesn't touch TEST_PROGS when it does build and clean.

As a result "make kselftest-clean" and "make clean" fail to remove
executables. Run and install work because the common framework runs
and installs TEST_PROGS. Build works because the Makefile defines
"all" rule which is unnecessary if TEST_GEN_PROGS is used.

Use TEST_GEN_PROGS so the common framework can handle build/run/
install/clean properly.

Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet/mlx4_en: Resolve bad operstate value
Lama Kayal [Sun, 19 Sep 2021 11:55:45 +0000 (14:55 +0300)]
net/mlx4_en: Resolve bad operstate value

Any link state change that's done prior to net device registration
isn't reflected on the state, thus the operational state is left
obsolete, with 'UNKNOWN' status.

To resolve the issue, query link state from FW upon open operations
to ensure operational state is updated.

Fixes: c27a02cd94d6 ("mlx4_en: Add driver for Mellanox ConnectX 10GbE NIC")
Signed-off-by: Lama Kayal <lkayal@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoselftests: net: af_unix: Fix incorrect args in test result msg
Shuah Khan [Fri, 17 Sep 2021 19:26:14 +0000 (13:26 -0600)]
selftests: net: af_unix: Fix incorrect args in test result msg

Fix the args to fprintf(). Splitting the message ends up passing
incorrect arg for "sigurg %d" and an extra arg overall. The test
result message ends up incorrect.

test_unix_oob.c: In function ‘main’:
test_unix_oob.c:274:43: warning: format ‘%d’ expects argument of type ‘int’, but argument 3 has type ‘char *’ [-Wformat=]
  274 |   fprintf(stderr, "Test 3 failed, sigurg %d len %d OOB %c ",
      |                                          ~^
      |                                           |
      |                                           int
      |                                          %s
  275 |   "atmark %d\n", signal_recvd, len, oob, atmark);
      |   ~~~~~~~~~~~~~
      |   |
      |   char *
test_unix_oob.c:274:19: warning: too many arguments for format [-Wformat-extra-args]
  274 |   fprintf(stderr, "Test 3 failed, sigurg %d len %d OOB %c ",

Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: bgmac-bcma: handle deferred probe error due to mac-address
Christian Lamparter [Sun, 19 Sep 2021 11:57:25 +0000 (13:57 +0200)]
net: bgmac-bcma: handle deferred probe error due to mac-address

Due to the inclusion of nvmem handling into the mac-address getter
function of_get_mac_address() by
commit d01f449c008a ("of_net: add NVMEM support to of_get_mac_address")
it is now possible to get a -EPROBE_DEFER return code. Which did cause
bgmac to assign a random ethernet address.

This exact issue happened on my Meraki MR32. The nvmem provider is
an EEPROM (at24c64) which gets instantiated once the module
driver is loaded... This happens once the filesystem becomes available.

With this patch, bgmac_probe() will propagate the -EPROBE_DEFER error.
Then the driver subsystem will reschedule the probe at a later time.

Cc: Petr Štetiar <ynezz@true.cz>
Cc: Michael Walle <michael@walle.cc>
Fixes: d01f449c008a ("of_net: add NVMEM support to of_get_mac_address")
Signed-off-by: Christian Lamparter <chunkeey@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: dsa: tear down devlink port regions when tearing down the devlink port on error
Vladimir Oltean [Fri, 17 Sep 2021 14:29:16 +0000 (17:29 +0300)]
net: dsa: tear down devlink port regions when tearing down the devlink port on error

Commit 86f8b1c01a0a ("net: dsa: Do not make user port errors fatal")
decided it was fine to ignore errors on certain ports that fail to
probe, and go on with the ports that do probe fine.

Commit fb6ec87f7229 ("net: dsa: Fix type was not set for devlink port")
noticed that devlink_port_type_eth_set(dlp, dp->slave); does not get
called, and devlink notices after a timeout of 3600 seconds and prints a
WARN_ON. So it went ahead to unregister the devlink port. And because
there exists an UNUSED port flavour, we actually re-register the devlink
port as UNUSED.

Commit 08156ba430b4 ("net: dsa: Add devlink port regions support to
DSA") added devlink port regions, which are set up by the driver and not
by DSA.

When we trigger the devlink port deregistration and reregistration as
unused, devlink now prints another WARN_ON, from here:

devlink_port_unregister:
WARN_ON(!list_empty(&devlink_port->region_list));

So the port still has regions, which makes sense, because they were set
up by the driver, and the driver doesn't know we're unregistering the
devlink port.

Somebody needs to tear them down, and optionally (actually it would be
nice, to be consistent) set them up again for the new devlink port.

But DSA's layering stays in our way quite badly here.

The options I've considered are:

1. Introduce a function in devlink to just change a port's type and
   flavour. No dice, devlink keeps a lot of state, it really wants the
   port to not be registered when you set its parameters, so changing
   anything can only be done by destroying what we currently have and
   recreating it.

2. Make DSA cache the parameters passed to dsa_devlink_port_region_create,
   and the region returned, keep those in a list, then when the devlink
   port unregister needs to take place, the existing devlink regions are
   destroyed by DSA, and we replay the creation of new regions using the
   cached parameters. Problem: mv88e6xxx keeps the region pointers in
   chip->ports[port].region, and these will remain stale after DSA frees
   them. There are many things DSA can do, but updating mv88e6xxx's
   private pointers is not one of them.

3. Just let the driver do it (i.e. introduce a very specific method
   called ds->ops->port_reinit_as_unused, which unregisters its devlink
   port devlink regions, then the old devlink port, then registers the
   new one, then the devlink port regions for it). While it does work,
   as opposed to the others, it's pretty horrible from an API
   perspective and we can do better.

4. Introduce a new pair of methods, ->port_setup and ->port_teardown,
   which in the case of mv88e6xxx must register and unregister the
   devlink port regions. Call these 2 methods when the port must be
   reinitialized as unused.

Naturally, I went for the 4th approach.

Fixes: 08156ba430b4 ("net: dsa: Add devlink port regions support to DSA")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: freescale: drop unneeded MODULE_ALIAS
Krzysztof Kozlowski [Fri, 17 Sep 2021 09:20:58 +0000 (11:20 +0200)]
net: freescale: drop unneeded MODULE_ALIAS

The MODULE_DEVICE_TABLE already creates proper alias for platform
driver.  Having another MODULE_ALIAS causes the alias to be duplicated.

Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge branch 'ocelot-phylink-fixes'
David S. Miller [Sun, 19 Sep 2021 11:59:52 +0000 (12:59 +0100)]
Merge branch 'ocelot-phylink-fixes'

Colin Foster says:

====================
ocelot phylink fixes

When the ocelot driver was migrated to phylink, e6e12df625f2 ("net:
mscc: ocelot: convert to phylink") there were two additional writes to
registers that became stale. One write was to DEV_CLOCK_CFG and one was
to ANA_PFC_PCF_CFG.

Both of these writes referenced the variable "speed" which originally
was set to OCELOT_SPEED_{10,100,1000,2500}. These macros expand to
values of 3, 2, 1, or 0, respectively. After the update, the variable
speed is set to SPEED_{10,100,1000,2500} which expand to 10, 100, 1000,
and 2500. So invalid values were getting written to the two registers,
which would lead to either a lack of functionality or undefined
funcationality.

Fixing these values was the intent of v1 of this patch set - submitted
as "[PATCH v1 net] net: ethernet: mscc: ocelot: bug fix when writing MAC
speed"

During that review it was determined that both writes were actually
unnecessary. DEV_CLOCK_CFG is a duplicate write, so can be removed
entirely. This was accidentally submitted as as a new, lone patch titled
"[PATCH v1 net] net: mscc: ocelot: remove buggy duplicate write to
DEV_CLOCK_CFG". This is part of what is considered v2 of this patch set.

Additionally, the write to ANA_PFC_PFC_CFG is also unnecessary. Priority
flow contol is disabled, so configuring it is useless and should be
removed. This was also submitted as a new, lone patch titled "[PATCH v1
net] net: mscc: ocelot: remove buggy and useless write to ANA_PFC_PFC_CFG".
This is the rest of what is considered v2 of this patch set.

v3
Identical to v2, but fixes the patch numbering to v3 and submitting the
two changes as a patch set.

v2
Note: I misunderstood and submitted two new "v1" patches instead of a
single "v2" patch set.
- Remove the buggy writes altogher
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: mscc: ocelot: remove buggy duplicate write to DEV_CLOCK_CFG
Colin Foster [Fri, 17 Sep 2021 15:39:05 +0000 (08:39 -0700)]
net: mscc: ocelot: remove buggy duplicate write to DEV_CLOCK_CFG

When updating ocelot to use phylink, a second write to DEV_CLOCK_CFG was
mistakenly left in. It used the variable "speed" which, previously, would
would have been assigned a value of OCELOT_SPEED_1000. In phylink the
variable is be SPEED_1000, which is invalid for the
DEV_CLOCK_LINK_SPEED macro. Removing it as unnecessary and buggy.

Fixes: e6e12df625f2 ("net: mscc: ocelot: convert to phylink")
Signed-off-by: Colin Foster <colin.foster@in-advantage.com>
Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: mscc: ocelot: remove buggy and useless write to ANA_PFC_PFC_CFG
Colin Foster [Fri, 17 Sep 2021 15:39:04 +0000 (08:39 -0700)]
net: mscc: ocelot: remove buggy and useless write to ANA_PFC_PFC_CFG

A useless write to ANA_PFC_PFC_CFG was left in while refactoring ocelot to
phylink. Since priority flow control is disabled, writing the speed has no
effect.

Further, it was using ethtool.h SPEED_ instead of OCELOT_SPEED_ macros,
which are incorrectly offset for GENMASK.

Lastly, for priority flow control to properly function, some scenarios
would rely on the rate adaptation from the PCS while the MAC speed would
be fixed. So it isn't used, and even if it was, neither "speed" nor
"mac_speed" are necessarily the correct values to be used.

Fixes: e6e12df625f2 ("net: mscc: ocelot: convert to phylink")
Signed-off-by: Colin Foster <colin.foster@in-advantage.com>
Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: core: Correct the sock::sk_lock.owned lockdep annotations
Thomas Gleixner [Sat, 18 Sep 2021 12:42:35 +0000 (14:42 +0200)]
net: core: Correct the sock::sk_lock.owned lockdep annotations

lock_sock_fast() and lock_sock_nested() contain lockdep annotations for the
sock::sk_lock.owned 'mutex'. sock::sk_lock.owned is not a regular mutex. It
is just lockdep wise equivalent. In fact it's an open coded trivial mutex
implementation with some interesting features.

sock::sk_lock.slock is a regular spinlock protecting the 'mutex'
representation sock::sk_lock.owned which is a plain boolean. If 'owned' is
true, then some other task holds the 'mutex', otherwise it is uncontended.
As this locking construct is obviously endangered by lock ordering issues as
any other locking primitive it got lockdep annotated via a dedicated
dependency map sock::sk_lock.dep_map which has to be updated at the lock
and unlock sites.

lock_sock_nested() is a straight forward 'mutex' lock operation:

  might_sleep();
  spin_lock_bh(sock::sk_lock.slock)
  while (!try_lock(sock::sk_lock.owned)) {
      spin_unlock_bh(sock::sk_lock.slock);
      wait_for_release();
      spin_lock_bh(sock::sk_lock.slock);
  }

The lockdep annotation for sock::sk_lock.owned is for unknown reasons
_after_ the lock has been acquired, i.e. after the code block above and
after releasing sock::sk_lock.slock, but inside the bottom halves disabled
region:

  spin_unlock(sock::sk_lock.slock);
  mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
  local_bh_enable();

The placement after the unlock is obvious because otherwise the
mutex_acquire() would nest into the spin lock held region.

But that's from the lockdep perspective still the wrong place:

 1) The mutex_acquire() is issued _after_ the successful acquisition which
    is pointless because in a dead lock scenario this point is never
    reached which means that if the deadlock is the first instance of
    exposing the wrong lock order lockdep does not have a chance to detect
    it.

 2) It only works because lockdep is rather lax on the context from which
    the mutex_acquire() is issued. Acquiring a mutex inside a bottom halves
    and therefore non-preemptible region is obviously invalid, except for a
    trylock which is clearly not the case here.

    This 'works' stops working on RT enabled kernels where the bottom halves
    serialization is done via a local lock, which exposes this misplacement
    because the 'mutex' and the local lock nest the wrong way around and
    lockdep complains rightfully about a lock inversion.

The placement is wrong since the initial commit a5b5bb9a053a ("[PATCH]
lockdep: annotate sk_locks") which introduced this.

Fix it by moving the mutex_acquire() in front of the actual lock
acquisition, which is what the regular mutex_lock() operation does as well.

lock_sock_fast() is not that straight forward. It looks at the first glance
like a convoluted trylock operation:

  spin_lock_bh(sock::sk_lock.slock)
  if (!sock::sk_lock.owned)
      return false;
  while (!try_lock(sock::sk_lock.owned)) {
      spin_unlock_bh(sock::sk_lock.slock);
      wait_for_release();
      spin_lock_bh(sock::sk_lock.slock);
  }
  spin_unlock(sock::sk_lock.slock);
  mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
  local_bh_enable();
  return true;

But that's not the case: lock_sock_fast() is an interesting optimization
for short critical sections which can run with bottom halves disabled and
sock::sk_lock.slock held. This allows to shortcut the 'mutex' operation in
the non contended case by preventing other lockers to acquire
sock::sk_lock.owned because they are blocked on sock::sk_lock.slock, which
in turn avoids the overhead of doing the heavy processing in release_sock()
including waking up wait queue waiters.

In the contended case, i.e. when sock::sk_lock.owned == true the behavior
is the same as lock_sock_nested().

Semantically this shortcut means, that the task acquired the 'mutex' even
if it does not touch the sock::sk_lock.owned field in the non-contended
case. Not telling lockdep about this shortcut acquisition is hiding
potential lock ordering violations in the fast path.

As a consequence the same reasoning as for the above lock_sock_nested()
case vs. the placement of the lockdep annotation applies.

The current placement of the lockdep annotation was just copied from
the original lock_sock(), now renamed to lock_sock_nested(),
implementation.

Fix this by moving the mutex_acquire() in front of the actual lock
acquisition and adding the corresponding mutex_release() into
unlock_sock_fast(). Also document the fast path return case with a comment.

Reported-by: Sebastian Siewior <bigeasy@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: netdev@vger.kernel.org
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Jakub Kicinski <kuba@kernel.org>
Cc: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agodocs: net: dsa: sja1105: fix reference to sja1105.txt
Alejandro Concepcion-Rodriguez [Fri, 17 Sep 2021 18:49:59 +0000 (18:49 +0000)]
docs: net: dsa: sja1105: fix reference to sja1105.txt

The file sja1105.txt was converted to nxp,sja1105.yaml.

Signed-off-by: Alejandro Concepcion-Rodriguez <asconcepcion@acoro.eu>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoigc: fix build errors for PTP
Randy Dunlap [Fri, 17 Sep 2021 21:05:47 +0000 (14:05 -0700)]
igc: fix build errors for PTP

When IGC=y and PTP_1588_CLOCK=m, the ptp_*() interface family is
not available to the igc driver. Make this driver depend on
PTP_1588_CLOCK_OPTIONAL so that it will build without errors.

Various igc commits have used ptp_*() functions without checking
that PTP_1588_CLOCK is enabled. Fix all of these here.

Fixes these build errors:

ld: drivers/net/ethernet/intel/igc/igc_main.o: in function `igc_msix_other':
igc_main.c:(.text+0x6494): undefined reference to `ptp_clock_event'
ld: igc_main.c:(.text+0x64ef): undefined reference to `ptp_clock_event'
ld: igc_main.c:(.text+0x6559): undefined reference to `ptp_clock_event'
ld: drivers/net/ethernet/intel/igc/igc_ethtool.o: in function `igc_ethtool_get_ts_info':
igc_ethtool.c:(.text+0xc7a): undefined reference to `ptp_clock_index'
ld: drivers/net/ethernet/intel/igc/igc_ptp.o: in function `igc_ptp_feature_enable_i225':
igc_ptp.c:(.text+0x330): undefined reference to `ptp_find_pin'
ld: igc_ptp.c:(.text+0x36f): undefined reference to `ptp_find_pin'
ld: drivers/net/ethernet/intel/igc/igc_ptp.o: in function `igc_ptp_init':
igc_ptp.c:(.text+0x11cd): undefined reference to `ptp_clock_register'
ld: drivers/net/ethernet/intel/igc/igc_ptp.o: in function `igc_ptp_stop':
igc_ptp.c:(.text+0x12dd): undefined reference to `ptp_clock_unregister'
ld: drivers/platform/x86/dell/dell-wmi-privacy.o: in function `dell_privacy_wmi_probe':

Fixes: 64433e5bf40ab ("igc: Enable internal i225 PPS")
Fixes: 60dbede0c4f3d ("igc: Add support for ethtool GET_TS_INFO command")
Fixes: 87938851b6efb ("igc: enable auxiliary PHC functions for the i225")
Fixes: 5f2958052c582 ("igc: Add basic skeleton for PTP")
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Ederson de Souza <ederson.desouza@intel.com>
Cc: Tony Nguyen <anthony.l.nguyen@intel.com>
Cc: Vinicius Costa Gomes <vinicius.gomes@intel.com>
Cc: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Jakub Kicinski <kuba@kernel.org>
Cc: Jesse Brandeburg <jesse.brandeburg@intel.com>
Cc: intel-wired-lan@lists.osuosl.org
Acked-by: Vinicius Costa Gomes <vinicius.gomes@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoenetc: Fix uninitialized struct dim_sample field usage
Claudiu Manoil [Fri, 17 Sep 2021 10:22:06 +0000 (13:22 +0300)]
enetc: Fix uninitialized struct dim_sample field usage

The only struct dim_sample member that does not get
initialized by dim_update_sample() is comp_ctr. (There
is special API to initialize comp_ctr:
dim_update_sample_with_comps(), and it is currently used
only for RDMA.) comp_ctr is used to compute curr_stats->cmps
and curr_stats->cpe_ratio (see dim_calc_stats()) which in
turn are consumed by the rdma_dim_*() API.  Therefore,
functionally, the net_dim*() API consumers are not affected.
Nevertheless, fix the computation of statistics based
on an uninitialized variable, even if the mentioned statistics
are not used at the moment.

Fixes: ae0e6a5d1627 ("enetc: Add adaptive interrupt coalescing")
Signed-off-by: Claudiu Manoil <claudiu.manoil@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoenetc: Fix illegal access when reading affinity_hint
Claudiu Manoil [Fri, 17 Sep 2021 10:22:05 +0000 (13:22 +0300)]
enetc: Fix illegal access when reading affinity_hint

irq_set_affinity_hit() stores a reference to the cpumask_t
parameter in the irq descriptor, and that reference can be
accessed later from irq_affinity_hint_proc_show(). Since
the cpu_mask parameter passed to irq_set_affinity_hit() has
only temporary storage (it's on the stack memory), later
accesses to it are illegal. Thus reads from the corresponding
procfs affinity_hint file can result in paging request oops.

The issue is fixed by the get_cpu_mask() helper, which provides
a permanent storage for the cpumask_t parameter.

Fixes: d4fd0404c1c9 ("enetc: Introduce basic PF and VF ENETC ethernet drivers")
Signed-off-by: Claudiu Manoil <claudiu.manoil@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agovirtio-net: fix pages leaking when building skb in big mode
Jason Wang [Fri, 17 Sep 2021 08:34:06 +0000 (16:34 +0800)]
virtio-net: fix pages leaking when building skb in big mode

We try to use build_skb() if we had sufficient tailroom. But we forget
to release the unused pages chained via private in big mode which will
leak pages. Fixing this by release the pages after building the skb in
big mode.

Cc: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Fixes: fb32856b16ad ("virtio-net: page_to_skb() use build_skb when there's sufficient tailroom")
Signed-off-by: Jason Wang <jasowang@redhat.com>
Reviewed-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoxen-netback: correct success/error reporting for the SKB-with-fraglist case
Jan Beulich [Fri, 17 Sep 2021 06:27:10 +0000 (08:27 +0200)]
xen-netback: correct success/error reporting for the SKB-with-fraglist case

When re-entering the main loop of xenvif_tx_check_gop() a 2nd time, the
special considerations for the head of the SKB no longer apply. Don't
mistakenly report ERROR to the frontend for the first entry in the list,
even if - from all I can tell - this shouldn't matter much as the overall
transmit will need to be considered failed anyway.

Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Paul Durrant <paul@xen.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge branch 'dsa-shutdown'
David S. Miller [Sun, 19 Sep 2021 11:08:37 +0000 (12:08 +0100)]
Merge branch 'dsa-shutdown'

Vladimir Oltean says:

====================
Make DSA switch drivers compatible with masters which unregister on shutdown

Changes in v2:
- fix build for b53_mmap
- use unregister_netdevice_many

It was reported by Lino here:

https://lore.kernel.org/netdev/20210909095324.12978-1-LinoSanfilippo@gmx.de/

that when the DSA master attempts to unregister its net_device on
shutdown, DSA should prevent that operation from succeeding because it
holds a reference to it. This hangs the shutdown process.

This issue was essentially introduced in commit 2f1e8ea726e9 ("net: dsa:
link interfaces with the DSA master to get rid of lockdep warnings").
The present series patches all DSA drivers to handle that case,
depending on whether those drivers were introduced before or after the
offending commit, a different Fixes: tag is specified for them.

The approach taken by this series solves the issue in essentially the
same way as Lino's patches, except for three key differences:

- this series takes a more minimal approach in what is done on shutdown,
  we do not attempt a full tree teardown as that is not strictly
  necessary. I might revisit this if there are compelling reasons to do
  otherwise

- this series fixes the issues for all DSA drivers, not just KSZ9897

- this series works even if the ->remove driver method gets called for
  the same device too, not just ->shutdown. This is really possible to
  happen for SPI device drivers, and potentially possible for other bus
  device drivers too.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: dsa: xrs700x: be compatible with masters which unregister on shutdown
Vladimir Oltean [Fri, 17 Sep 2021 13:34:36 +0000 (16:34 +0300)]
net: dsa: xrs700x: be compatible with masters which unregister on shutdown

Since commit 2f1e8ea726e9 ("net: dsa: link interfaces with the DSA
master to get rid of lockdep warnings"), DSA gained a requirement which
it did not fulfill, which is to unlink itself from the DSA master at
shutdown time.

Since the Arrow SpeedChips XRS700x driver was introduced after the bad
commit, it has never worked with DSA masters which decide to unregister
their net_device on shutdown, effectively hanging the reboot process.
To fix that, we need to call dsa_switch_shutdown.

These devices can be connected by I2C or by MDIO, and if I search for
I2C or MDIO bus drivers that implement their ->shutdown by redirecting
it to ->remove I don't see any, however this does not mean it would not
be possible. To be compatible with that pattern, it is necessary to
implement an "if this then not that" scheme, to avoid ->remove and
->shutdown from being called both for the same struct device.

Fixes: ee00b24f32eb ("net: dsa: add Arrow SpeedChips XRS700x driver")
Link: https://lore.kernel.org/netdev/20210909095324.12978-1-LinoSanfilippo@gmx.de/
Reported-by: Lino Sanfilippo <LinoSanfilippo@gmx.de>
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: George McCollister <george.mccollister@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: dsa: microchip: ksz8863: be compatible with masters which unregister on shutdown
Vladimir Oltean [Fri, 17 Sep 2021 13:34:35 +0000 (16:34 +0300)]
net: dsa: microchip: ksz8863: be compatible with masters which unregister on shutdown

Since commit 2f1e8ea726e9 ("net: dsa: link interfaces with the DSA
master to get rid of lockdep warnings"), DSA gained a requirement which
it did not fulfill, which is to unlink itself from the DSA master at
shutdown time.

Since the Microchip sub-driver for KSZ8863 was introduced after the bad
commit, it has never worked with DSA masters which decide to unregister
their net_device on shutdown, effectively hanging the reboot process.
To fix that, we need to call dsa_switch_shutdown.

Since this driver expects the MDIO bus to be backed by mdio_bitbang, I
don't think there is currently any MDIO bus driver which implements its
->shutdown by redirecting it to ->remove, but in any case, to be
compatible with that pattern, it is necessary to implement an "if this
then not that" scheme, to avoid ->remove and ->shutdown from being
called both for the same struct device.

Fixes: 60a364760002 ("net: dsa: microchip: Add Microchip KSZ8863 SMI based driver support")
Link: https://lore.kernel.org/netdev/20210909095324.12978-1-LinoSanfilippo@gmx.de/
Reported-by: Lino Sanfilippo <LinoSanfilippo@gmx.de>
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: dsa: hellcreek: be compatible with masters which unregister on shutdown
Vladimir Oltean [Fri, 17 Sep 2021 13:34:34 +0000 (16:34 +0300)]
net: dsa: hellcreek: be compatible with masters which unregister on shutdown

Since commit 2f1e8ea726e9 ("net: dsa: link interfaces with the DSA
master to get rid of lockdep warnings"), DSA gained a requirement which
it did not fulfill, which is to unlink itself from the DSA master at
shutdown time.

Since the hellcreek driver was introduced after the bad commit, it has
never worked with DSA masters which decide to unregister their
net_device on shutdown, effectively hanging the reboot process.

Hellcreek is a platform device driver, so we probably cannot have the
oddities of ->shutdown and ->remove getting both called for the exact
same struct device. But to be in line with the pattern from the other
device drivers which are on slow buses, implement the same "if this then
not that" pattern of either running the ->shutdown or the ->remove hook.
The driver's current ->remove implementation makes that very easy
because it already zeroes out its device_drvdata on ->remove.

Fixes: e4b27ebc780f ("net: dsa: Add DSA driver for Hirschmann Hellcreek switches")
Link: https://lore.kernel.org/netdev/20210909095324.12978-1-LinoSanfilippo@gmx.de/
Reported-by: Lino Sanfilippo <LinoSanfilippo@gmx.de>
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Acked-by: Kurt Kanzenbach <kurt@linutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: dsa: be compatible with masters which unregister on shutdown
Vladimir Oltean [Fri, 17 Sep 2021 13:34:33 +0000 (16:34 +0300)]
net: dsa: be compatible with masters which unregister on shutdown

Lino reports that on his system with bcmgenet as DSA master and KSZ9897
as a switch, rebooting or shutting down never works properly.

What does the bcmgenet driver have special to trigger this, that other
DSA masters do not? It has an implementation of ->shutdown which simply
calls its ->remove implementation. Otherwise said, it unregisters its
network interface on shutdown.

This message can be seen in a loop, and it hangs the reboot process there:

unregister_netdevice: waiting for eth0 to become free. Usage count = 3

So why 3?

A usage count of 1 is normal for a registered network interface, and any
virtual interface which links itself as an upper of that will increment
it via dev_hold. In the case of DSA, this is the call path:

dsa_slave_create
-> netdev_upper_dev_link
   -> __netdev_upper_dev_link
      -> __netdev_adjacent_dev_insert
         -> dev_hold

So a DSA switch with 3 interfaces will result in a usage count elevated
by two, and netdev_wait_allrefs will wait until they have gone away.

Other stacked interfaces, like VLAN, watch NETDEV_UNREGISTER events and
delete themselves, but DSA cannot just vanish and go poof, at most it
can unbind itself from the switch devices, but that must happen strictly
earlier compared to when the DSA master unregisters its net_device, so
reacting on the NETDEV_UNREGISTER event is way too late.

It seems that it is a pretty established pattern to have a driver's
->shutdown hook redirect to its ->remove hook, so the same code is
executed regardless of whether the driver is unbound from the device, or
the system is just shutting down. As Florian puts it, it is quite a big
hammer for bcmgenet to unregister its net_device during shutdown, but
having a common code path with the driver unbind helps ensure it is well
tested.

So DSA, for better or for worse, has to live with that and engage in an
arms race of implementing the ->shutdown hook too, from all individual
drivers, and do something sane when paired with masters that unregister
their net_device there. The only sane thing to do, of course, is to
unlink from the master.

However, complications arise really quickly.

The pattern of redirecting ->shutdown to ->remove is not unique to
bcmgenet or even to net_device drivers. In fact, SPI controllers do it
too (see dspi_shutdown -> dspi_remove), and presumably, I2C controllers
and MDIO controllers do it too (this is something I have not researched
too deeply, but even if this is not the case today, it is certainly
plausible to happen in the future, and must be taken into consideration).

Since DSA switches might be SPI devices, I2C devices, MDIO devices, the
insane implication is that for the exact same DSA switch device, we
might have both ->shutdown and ->remove getting called.

So we need to do something with that insane environment. The pattern
I've come up with is "if this, then not that", so if either ->shutdown
or ->remove gets called, we set the device's drvdata to NULL, and in the
other hook, we check whether the drvdata is NULL and just do nothing.
This is probably not necessary for platform devices, just for devices on
buses, but I would really insist for consistency among drivers, because
when code is copy-pasted, it is not always copy-pasted from the best
sources.

So depending on whether the DSA switch's ->remove or ->shutdown will get
called first, we cannot really guarantee even for the same driver if
rebooting will result in the same code path on all platforms. But
nonetheless, we need to do something minimally reasonable on ->shutdown
too to fix the bug. Of course, the ->remove will do more (a full
teardown of the tree, with all data structures freed, and this is why
the bug was not caught for so long). The new ->shutdown method is kept
separate from dsa_unregister_switch not because we couldn't have
unregistered the switch, but simply in the interest of doing something
quick and to the point.

The big question is: does the DSA switch's ->shutdown get called earlier
than the DSA master's ->shutdown? If not, there is still a risk that we
might still trigger the WARN_ON in unregister_netdevice that says we are
attempting to unregister a net_device which has uppers. That's no good.
Although the reference to the master net_device won't physically go away
even if DSA's ->shutdown comes afterwards, remember we have a dev_hold
on it.

The answer to that question lies in this comment above device_link_add:

 * A side effect of the link creation is re-ordering of dpm_list and the
 * devices_kset list by moving the consumer device and all devices depending
 * on it to the ends of these lists (that does not happen to devices that have
 * not been registered when this function is called).

so the fact that DSA uses device_link_add towards its master is not
exactly for nothing. device_shutdown() walks devices_kset from the back,
so this is our guarantee that DSA's shutdown happens before the master's
shutdown.

Fixes: 2f1e8ea726e9 ("net: dsa: link interfaces with the DSA master to get rid of lockdep warnings")
Link: https://lore.kernel.org/netdev/20210909095324.12978-1-LinoSanfilippo@gmx.de/
Reported-by: Lino Sanfilippo <LinoSanfilippo@gmx.de>
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Tested-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: mdio: introduce a shutdown method to mdio device drivers
Vladimir Oltean [Fri, 17 Sep 2021 13:34:32 +0000 (16:34 +0300)]
net: mdio: introduce a shutdown method to mdio device drivers

MDIO-attached devices might have interrupts and other things that might
need quiesced when we kexec into a new kernel. Things are even more
creepy when those interrupt lines are shared, and in that case it is
absolutely mandatory to disable all interrupt sources.

Moreover, MDIO devices might be DSA switches, and DSA needs its own
shutdown method to unlink from the DSA master, which is a new
requirement that appeared after commit 2f1e8ea726e9 ("net: dsa: link
interfaces with the DSA master to get rid of lockdep warnings").

So introduce a ->shutdown method in the MDIO device driver structure.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: dsa: bcm_sf2: Fix array overrun in bcm_sf2_num_active_ports()
Florian Fainelli [Thu, 16 Sep 2021 21:33:35 +0000 (14:33 -0700)]
net: dsa: bcm_sf2: Fix array overrun in bcm_sf2_num_active_ports()

After d12e1c464988 ("net: dsa: b53: Set correct number of ports in the
DSA struct") we stopped setting dsa_switch::num_ports to DSA_MAX_PORTS,
which created an off by one error between the statically allocated
bcm_sf2_priv::port_sts array (of size DSA_MAX_PORTS). When
dsa_is_cpu_port() is used, we end-up accessing an out of bounds member
and causing a NPD.

Fix this by iterating with the appropriate port count using
ds->num_ports.

Fixes: d12e1c464988 ("net: dsa: b53: Set correct number of ports in the DSA struct")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: update NXP copyright text
Vladimir Oltean [Fri, 17 Sep 2021 11:17:35 +0000 (14:17 +0300)]
net: update NXP copyright text

NXP Legal insists that the following are not fine:

- Saying "NXP Semiconductors" instead of "NXP", since the company's
  registered name is "NXP"

- Putting a "(c)" sign in the copyright string

- Putting a comma in the copyright string

The only accepted copyright string format is "Copyright <year-range> NXP".

This patch changes the copyright headers in the networking files that
were sent by me, or derived from code sent by me.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: hso: fix muxed tty registration
Johan Hovold [Fri, 17 Sep 2021 10:12:04 +0000 (12:12 +0200)]
net: hso: fix muxed tty registration

If resource allocation and registration fail for a muxed tty device
(e.g. if there are no more minor numbers) the driver should not try to
deregister the never-registered (or already-deregistered) tty.

Fix up the error handling to avoid dereferencing a NULL pointer when
attempting to remove the character device.

Fixes: 72dc1c096c70 ("HSO: add option hso driver")
Cc: stable@vger.kernel.org # 2.6.27
Signed-off-by: Johan Hovold <johan@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge tag 'net-5.15-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
Linus Torvalds [Thu, 16 Sep 2021 20:05:42 +0000 (13:05 -0700)]
Merge tag 'net-5.15-rc2' of git://git./linux/kernel/git/netdev/net

Pull networking fixes from Jakub Kicinski:
 "Including fixes from bpf.

  Current release - regressions:

   - vhost_net: fix OoB on sendmsg() failure

   - mlx5: bridge, fix uninitialized variable usage

   - bnxt_en: fix error recovery regression

  Current release - new code bugs:

   - bpf, mm: fix lockdep warning triggered by stack_map_get_build_id_offset()

  Previous releases - regressions:

   - r6040: restore MDIO clock frequency after MAC reset

   - tcp: fix tp->undo_retrans accounting in tcp_sacktag_one()

   - dsa: flush switchdev workqueue before tearing down CPU/DSA ports

  Previous releases - always broken:

   - ptp: dp83640: don't define PAGE0, avoid compiler warning

   - igc: fix tunnel segmentation offloads

   - phylink: update SFP selected interface on advertising changes

   - stmmac: fix system hang caused by eee_ctrl_timer during suspend/resume

   - mlx5e: fix mutual exclusion between CQE compression and HW TS

  Misc:

   - bpf, cgroups: fix cgroup v2 fallback on v1/v2 mixed mode

   - sfc: fallback for lack of xdp tx queues

   - hns3: add option to turn off page pool feature"

* tag 'net-5.15-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (67 commits)
  mlxbf_gige: clear valid_polarity upon open
  igc: fix tunnel offloading
  net/{mlx5|nfp|bnxt}: Remove unnecessary RTNL lock assert
  net: wan: wanxl: define CROSS_COMPILE_M68K
  selftests: nci: replace unsigned int with int
  net: dsa: flush switchdev workqueue before tearing down CPU/DSA ports
  Revert "net: phy: Uniform PHY driver access"
  net: dsa: destroy the phylink instance on any error in dsa_slave_phy_setup
  ptp: dp83640: don't define PAGE0
  bnx2x: Fix enabling network interfaces without VFs
  Revert "Revert "ipv4: fix memory leaks in ip_cmsg_send() callers""
  tcp: fix tp->undo_retrans accounting in tcp_sacktag_one()
  net-caif: avoid user-triggerable WARN_ON(1)
  bpf, selftests: Add test case for mixed cgroup v1/v2
  bpf, selftests: Add cgroup v1 net_cls classid helpers
  bpf, cgroups: Fix cgroup v2 fallback on v1/v2 mixed mode
  bpf: Add oversize check before call kvcalloc()
  net: hns3: fix the timing issue of VF clearing interrupt sources
  net: hns3: fix the exception when query imp info
  net: hns3: disable mac in flr process
  ...

3 years agonet: 6pack: Fix tx timeout and slot time
Guenter Roeck [Thu, 9 Sep 2021 03:57:43 +0000 (20:57 -0700)]
net: 6pack: Fix tx timeout and slot time

tx timeout and slot time are currently specified in units of HZ.  On
Alpha, HZ is defined as 1024.  When building alpha:allmodconfig, this
results in the following error message.

  drivers/net/hamradio/6pack.c: In function 'sixpack_open':
  drivers/net/hamradio/6pack.c:71:41: error:
   unsigned conversion from 'int' to 'unsigned char'
   changes value from '256' to '0'

In the 6PACK protocol, tx timeout is specified in units of 10 ms and
transmitted over the wire:

    https://www.linux-ax25.org/wiki/6PACK

Defining a value dependent on HZ doesn't really make sense, and
presumably comes from the (very historical) situation where HZ was
originally 100.

Note that the SIXP_SLOTTIME use explicitly is about 10ms granularity:

        mod_timer(&sp->tx_t, jiffies + ((when + 1) * HZ) / 100);

and the SIXP_TXDELAY walue is sent as a byte over the wire.

Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agodrm/rockchip: cdn-dp-core: Make cdn_dp_core_resume __maybe_unused
Arnd Bergmann [Tue, 28 Apr 2020 21:31:24 +0000 (23:31 +0200)]
drm/rockchip: cdn-dp-core: Make cdn_dp_core_resume __maybe_unused

With the new static annotation, the compiler warns when the functions
are actually unused:

   drivers/gpu/drm/rockchip/cdn-dp-core.c:1123:12: error: 'cdn_dp_resume' defined but not used [-Werror=unused-function]
    1123 | static int cdn_dp_resume(struct device *dev)
         |            ^~~~~~~~~~~~~

Mark them __maybe_unused to suppress that warning as well.

[ Not so 'new' static annotations any more, and I removed the part of
  the patch that added __maybe_unused to cdn_dp_suspend(), because it's
  used by the shutdown/remove code.

  So only the resume function ends up possibly unused if CONFIG_PM isn't
  set     - Linus ]

Fixes: 7c49abb4c2f8 ("drm/rockchip: cdn-dp-core: Make cdn_dp_core_suspend/resume static")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Enric Balletbo i Serra <enric.balletbo@collabora.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agocpufreq: vexpress: Drop unused variable
Guenter Roeck [Thu, 9 Sep 2021 18:47:14 +0000 (11:47 -0700)]
cpufreq: vexpress: Drop unused variable

arm:allmodconfig fails to build with the following error.

  drivers/cpufreq/vexpress-spc-cpufreq.c:454:13: error:
unused variable 'cur_cluster'

Remove the unused variable.

Fixes: bb8c26d9387f ("cpufreq: vexpress: Set CPUFREQ_IS_COOLING_DEV flag")
Cc: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Reviewed-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agoalpha: Declare virt_to_phys and virt_to_bus parameter as pointer to volatile
Guenter Roeck [Thu, 9 Sep 2021 05:00:33 +0000 (22:00 -0700)]
alpha: Declare virt_to_phys and virt_to_bus parameter as pointer to volatile

Some drivers pass a pointer to volatile data to virt_to_bus() and
virt_to_phys(), and that works fine.  One exception is alpha.  This
results in a number of compile errors such as

  drivers/net/wan/lmc/lmc_main.c: In function 'lmc_softreset':
  drivers/net/wan/lmc/lmc_main.c:1782:50: error:
passing argument 1 of 'virt_to_bus' discards 'volatile'
qualifier from pointer target type

  drivers/atm/ambassador.c: In function 'do_loader_command':
  drivers/atm/ambassador.c:1747:58: error:
passing argument 1 of 'virt_to_bus' discards 'volatile'
qualifier from pointer target type

Declare the parameter of virt_to_phys and virt_to_bus as pointer to
volatile to fix the problem.

Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years ago3com 3c515: make it compile on 64-bit architectures
Linus Torvalds [Thu, 16 Sep 2021 18:14:47 +0000 (11:14 -0700)]
3com 3c515: make it compile on 64-bit architectures

This driver isn't enabled most places because of the ISA config
dependency, but alpha still has it.  And I think the 'Jensen' actually
did have an ISA slot.

However, it doesn't build cleanly, because the "Vortex bus master" code
just casts the skb->data pointer to 'int':

        outl((int) (skb->data), ioaddr + Wn7_MasterAddr);

which is all kinds of broken.  Even on a good old traditional PC/AT it
would be broken because the high bits will be random kernel address
bits, but presumably the hardware ignores those bits.  I mean, it's ISA.
We're talking 16MB dma limits. The "good old days".

Make the build happy with this kind of craziness by using the proper
isa_virt_to_bus() handling that the full bus master code uses anyway
(the Vortex bus mastering is a limited special case).

Who knows, this might even work.

Reported-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agoMerge tag 'for-5.15/parisc-4' of git://git.kernel.org/pub/scm/linux/kernel/git/deller...
Linus Torvalds [Thu, 16 Sep 2021 17:32:00 +0000 (10:32 -0700)]
Merge tag 'for-5.15/parisc-4' of git://git./linux/kernel/git/deller/parisc-linux

Pull parisc fix from Helge Deller:
 "Fix a build warning when using the PAGE0 pointer"

* tag 'for-5.15/parisc-4' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/parisc-linux:
  parisc: Use absolute_pointer() to define PAGE0

3 years agoMerge tag 'm68k-for-v5.15-tag2' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Thu, 16 Sep 2021 17:26:48 +0000 (10:26 -0700)]
Merge tag 'm68k-for-v5.15-tag2' of git://git./linux/kernel/git/geert/linux-m68k

Pull m68k fixes from Geert Uytterhoeven:

 - Warning fixes to mitigate CONFIG_WERROR=y

* tag 'm68k-for-v5.15-tag2' of git://git.kernel.org/pub/scm/linux/kernel/git/geert/linux-m68k:
  m68k: mvme: Remove overdue #warnings in RTC handling
  m68k: Double cast io functions to unsigned long

3 years agomlxbf_gige: clear valid_polarity upon open
David Thompson [Wed, 15 Sep 2021 18:08:48 +0000 (14:08 -0400)]
mlxbf_gige: clear valid_polarity upon open

The network interface managed by the mlxbf_gige driver can
get into a problem state where traffic does not flow.
In this state, the interface will be up and enabled, but
will stop processing received packets.  This problem state
will happen if three specific conditions occur:
    1) driver has received more than (N * RxRingSize) packets but
       less than (N+1 * RxRingSize) packets, where N is an odd number
       Note: the command "ethtool -g <interface>" will display the
       current receive ring size, which currently defaults to 128
    2) the driver's interface was disabled via "ifconfig oob_net0 down"
       during the window described in #1.
    3) the driver's interface is re-enabled via "ifconfig oob_net0 up"

This patch ensures that the driver's "valid_polarity" field is
cleared during the open() method so that it always matches the
receive polarity used by hardware.  Without this fix, the driver
needs to be unloaded and reloaded to correct this problem state.

Fixes: f92e1869d74e ("Add Mellanox BlueField Gigabit Ethernet driver")
Reviewed-by: Asmaa Mnebhi <asmaa@nvidia.com>
Signed-off-by: David Thompson <davthompson@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoigc: fix tunnel offloading
Paolo Abeni [Wed, 15 Sep 2021 17:19:07 +0000 (10:19 -0700)]
igc: fix tunnel offloading

Checking tunnel offloading, it turns out that offloading doesn't work
as expected.  The following script allows to reproduce the issue.
Call it as `testscript DEVICE LOCALIP REMOTEIP NETMASK'

=== SNIP ===
if [ $# -ne 4 ]
then
  echo "Usage $0 DEVICE LOCALIP REMOTEIP NETMASK"
  exit 1
fi
DEVICE="$1"
LOCAL_ADDRESS="$2"
REMOTE_ADDRESS="$3"
NWMASK="$4"
echo "Driver: $(ethtool -i ${DEVICE} | awk '/^driver:/{print $2}') "
ethtool -k "${DEVICE}" | grep tx-udp
echo
echo "Set up NIC and tunnel..."
ip addr add "${LOCAL_ADDRESS}/${NWMASK}" dev "${DEVICE}"
ip link set "${DEVICE}" up
sleep 2
ip link add vxlan1 type vxlan id 42 \
   remote "${REMOTE_ADDRESS}" \
   local "${LOCAL_ADDRESS}" \
   dstport 0 \
   dev "${DEVICE}"
ip addr add fc00::1/64 dev vxlan1
ip link set vxlan1 up
sleep 2
rm -f vxlan.pcap
echo "Running tcpdump and iperf3..."
( nohup tcpdump -i any -w vxlan.pcap >/dev/null 2>&1 ) &
sleep 2
iperf3 -c fc00::2 >/dev/null
pkill tcpdump
echo
echo -n "Max. Paket Size: "
tcpdump -r vxlan.pcap -nnle 2>/dev/null \
| grep "${LOCAL_ADDRESS}.*> ${REMOTE_ADDRESS}.*OTV" \
| awk '{print $8}' | awk -F ':' '{print $1}' \
| sort -n | tail -1
echo
ip link del vxlan1
ip addr del ${LOCAL_ADDRESS}/${NWMASK} dev "${DEVICE}"
=== SNAP ===

The expected outcome is

  Max. Paket Size: 64904

This is what you see on igb, the code igc has been taken from.
However, on igc the output is

  Max. Paket Size: 1516

so the GSO aggregate packets are segmented by the kernel before calling
igc_xmit_frame.  Inside the subsequent call to igc_tso, the check for
skb_is_gso(skb) fails and the function returns prematurely.

It turns out that this occurs because the feature flags aren't set
entirely correctly in igc_probe.  In contrast to the original code
from igb_probe, igc_probe neglects to set the flags required to allow
tunnel offloading.

Setting the same flags as igb fixes the issue on igc.

Fixes: 34428dff3679 ("igc: Add GSO partial support")
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Tested-by: Corinna Vinschen <vinschen@redhat.com>
Acked-by: Sasha Neftin <sasha.neftin@intel.com>
Tested-by: Nechama Kraus <nechamax.kraus@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet/{mlx5|nfp|bnxt}: Remove unnecessary RTNL lock assert
Eli Cohen [Wed, 15 Sep 2021 04:47:27 +0000 (07:47 +0300)]
net/{mlx5|nfp|bnxt}: Remove unnecessary RTNL lock assert

Remove the assert from the callback priv lookup function since it does
not require RTNL lock and is already protected by flow_indr_block_lock.

This will avoid warnings from being emitted to dmesg if the driver
registers its callback after an ingress qdisc was created for a
netdevice.

The warnings started after the following patch was merged:
commit 74fc4f828769 ("net: Fix offloading indirect devices dependency on qdisc order creation")

Signed-off-by: Eli Cohen <elic@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: wan: wanxl: define CROSS_COMPILE_M68K
Adam Borowski [Sun, 12 Sep 2021 21:23:21 +0000 (23:23 +0200)]
net: wan: wanxl: define CROSS_COMPILE_M68K

It was used but never set.  The hardcoded value from before the dawn of
time was non-standard; the usual name for cross-tools is $TRIPLET-$TOOL

Signed-off-by: Adam Borowski <kilobyte@angband.pl>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoselftests: nci: replace unsigned int with int
Xiang wangx [Thu, 16 Sep 2021 12:24:42 +0000 (20:24 +0800)]
selftests: nci: replace unsigned int with int

Should not use comparison of unsigned expressions < 0.

Signed-off-by: Xiang wangx <wangxiang@cdjrlc.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoparisc: Use absolute_pointer() to define PAGE0
Helge Deller [Thu, 16 Sep 2021 06:35:42 +0000 (08:35 +0200)]
parisc: Use absolute_pointer() to define PAGE0

Use absolute_pointer() wrapper for PAGE0 to avoid this compiler warning:

  arch/parisc/kernel/setup.c: In function 'start_parisc':
  error: '__builtin_memcmp_eq' specified bound 8 exceeds source size 0

Signed-off-by: Helge Deller <deller@gmx.de>
Co-Developed-by: Guenter Roeck <linux@roeck-us.net>
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agoMerge tag 'hyperv-fixes-signed-20210915' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Thu, 16 Sep 2021 00:18:56 +0000 (17:18 -0700)]
Merge tag 'hyperv-fixes-signed-20210915' of git://git./linux/kernel/git/hyperv/linux

Pull hyperv fixes from Wei Liu:

 - Fix kernel crash caused by uio driver (Vitaly Kuznetsov)

 - Remove on-stack cpumask from HV APIC code (Wei Liu)

* tag 'hyperv-fixes-signed-20210915' of git://git.kernel.org/pub/scm/linux/kernel/git/hyperv/linux:
  x86/hyperv: remove on-stack cpumask from hv_send_ipi_mask_allbutself
  asm-generic/hyperv: provide cpumask_to_vpset_noself
  Drivers: hv: vmbus: Fix kernel crash upon unbinding a device from uio_hv_generic driver

3 years agoMerge tag 'rtc-5.15-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/abelloni...
Linus Torvalds [Thu, 16 Sep 2021 00:06:01 +0000 (17:06 -0700)]
Merge tag 'rtc-5.15-fixes' of git://git./linux/kernel/git/abelloni/linux

Pull RTC fix from Alexandre Belloni:
 "Fix a locking issue in the cmos rtc driver"

* tag 'rtc-5.15-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/abelloni/linux:
  rtc: cmos: Disable irq around direct invocation of cmos_interrupt()

3 years agonet: dsa: flush switchdev workqueue before tearing down CPU/DSA ports
Vladimir Oltean [Tue, 14 Sep 2021 13:47:26 +0000 (16:47 +0300)]
net: dsa: flush switchdev workqueue before tearing down CPU/DSA ports

Sometimes when unbinding the mv88e6xxx driver on Turris MOX, these error
messages appear:

mv88e6085 d0032004.mdio-mii:12: port 1 failed to delete be:79:b4:9e:9e:96 vid 1 from fdb: -2
mv88e6085 d0032004.mdio-mii:12: port 1 failed to delete be:79:b4:9e:9e:96 vid 0 from fdb: -2
mv88e6085 d0032004.mdio-mii:12: port 1 failed to delete d8:58:d7:00:ca:6d vid 100 from fdb: -2
mv88e6085 d0032004.mdio-mii:12: port 1 failed to delete d8:58:d7:00:ca:6d vid 1 from fdb: -2
mv88e6085 d0032004.mdio-mii:12: port 1 failed to delete d8:58:d7:00:ca:6d vid 0 from fdb: -2

(and similarly for other ports)

What happens is that DSA has a policy "even if there are bugs, let's at
least not leak memory" and dsa_port_teardown() clears the dp->fdbs and
dp->mdbs lists, which are supposed to be empty.

But deleting that cleanup code, the warnings go away.

=> the FDB and MDB lists (used for refcounting on shared ports, aka CPU
and DSA ports) will eventually be empty, but are not empty by the time
we tear down those ports. Aka we are deleting them too soon.

The addresses that DSA complains about are host-trapped addresses: the
local addresses of the ports, and the MAC address of the bridge device.

The problem is that offloading those entries happens from a deferred
work item scheduled by the SWITCHDEV_FDB_DEL_TO_DEVICE handler, and this
races with the teardown of the CPU and DSA ports where the refcounting
is kept.

In fact, not only it races, but fundamentally speaking, if we iterate
through the port list linearly, we might end up tearing down the shared
ports even before we delete a DSA user port which has a bridge upper.

So as it turns out, we need to first tear down the user ports (and the
unused ones, for no better place of doing that), then the shared ports
(the CPU and DSA ports). In between, we need to ensure that all work
items scheduled by our switchdev handlers (which only run for user
ports, hence the reason why we tear them down first) have finished.

Fixes: 161ca59d39e9 ("net: dsa: reference count the MDB entries at the cross-chip notifier level")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Link: https://lore.kernel.org/r/20210914134726.2305133-1-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agoRevert "net: phy: Uniform PHY driver access"
Vladimir Oltean [Tue, 14 Sep 2021 14:05:15 +0000 (17:05 +0300)]
Revert "net: phy: Uniform PHY driver access"

This reverts commit 3ac8eed62596387214869319379c1fcba264d8c6, which did
more than it said on the box, and not only it replaced to_phy_driver
with phydev->drv, but it also removed the "!drv" check, without actually
explaining why that is fine.

That patch in fact breaks suspend/resume on any system which has PHY
devices with no drivers bound.

The stack trace is:

Unable to handle kernel NULL pointer dereference at virtual address 00000000000000e8
pc : mdio_bus_phy_suspend+0xd8/0xec
lr : dpm_run_callback+0x38/0x90
Call trace:
 mdio_bus_phy_suspend+0xd8/0xec
 dpm_run_callback+0x38/0x90
 __device_suspend+0x108/0x3cc
 dpm_suspend+0x140/0x210
 dpm_suspend_start+0x7c/0xa0
 suspend_devices_and_enter+0x13c/0x540
 pm_suspend+0x2a4/0x330

Examples why that assumption is not fine:

- There is an MDIO bus with a PHY device that doesn't have a specific
  PHY driver loaded, because mdiobus_register() automatically creates a
  PHY device for it but there is no specific PHY driver in the system.
  Normally under those circumstances, the generic PHY driver will be
  bound lazily to it (at phy_attach_direct time). But some Ethernet
  drivers attach to their PHY at .ndo_open time. Until then it, the
  to-be-driven-by-genphy PHY device will not have a driver. The blamed
  patch amounts to saying "you need to open all net devices before the
  system can suspend, to avoid the NULL pointer dereference".

- There is any raw MDIO device which has 'plausible' values in the PHY
  ID registers 2 and 3, which is located on an MDIO bus whose driver
  does not set bus->phy_mask = ~0 (which prevents auto-scanning of PHY
  devices). An example could be a MAC's internal MDIO bus with PCS
  devices on it, for serial links such as SGMII. PHY devices will get
  created for those PCSes too, due to that MDIO bus auto-scanning, and
  although those PHY devices are not used, they do not bother anybody
  either. PCS devices are usually managed in Linux as raw MDIO devices.
  Nonetheless, they do not have a PHY driver, nor does anybody attempt
  to connect to them (because they are not a PHY), and therefore this
  patch breaks that.

The goal itself of the patch is questionable, so I am going for a
straight revert. to_phy_driver does not seem to have a need to be
replaced by phydev->drv, in fact that might even trigger code paths
which were not given too deep of a thought.

For instance:

phy_probe populates phydev->drv at the beginning, but does not clean it
up on any error (including EPROBE_DEFER). So if the phydev driver
requests probe deferral, phydev->drv will remain populated despite there
being no driver bound.

If a system suspend starts in between the initial probe deferral request
and the subsequent probe retry, we will be calling the phydev->drv->suspend
method, but _before_ any phydev->drv->probe call has succeeded.

That is to say, if the phydev->drv is allocating any driver-private data
structure in ->probe, it pretty much expects that data structure to be
available in ->suspend. But it may not. That is a pretty insane
environment to present to PHY drivers.

In the code structure before the blamed patch, mdio_bus_phy_may_suspend
would just say "no, don't suspend" to any PHY device which does not have
a driver pointer _in_the_device_structure_ (not the phydev->drv). That
would essentially ensure that ->suspend will never get called for a
device that has not yet successfully completed probe. This is the code
structure the patch is returning to, via the revert.

Fixes: 3ac8eed62596 ("net: phy: Uniform PHY driver access")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Acked-by: Florian Fainelli <f.fainelli@gmail.com>
Link: https://lore.kernel.org/r/20210914140515.2311548-1-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agonet: dsa: destroy the phylink instance on any error in dsa_slave_phy_setup
Vladimir Oltean [Tue, 14 Sep 2021 13:43:31 +0000 (16:43 +0300)]
net: dsa: destroy the phylink instance on any error in dsa_slave_phy_setup

DSA supports connecting to a phy-handle, and has a fallback to a non-OF
based method of connecting to an internal PHY on the switch's own MDIO
bus, if no phy-handle and no fixed-link nodes were present.

The -ENODEV error code from the first attempt (phylink_of_phy_connect)
is what triggers the second attempt (phylink_connect_phy).

However, when the first attempt returns a different error code than
-ENODEV, this results in an unbalance of calls to phylink_create and
phylink_destroy by the time we exit the function. The phylink instance
has leaked.

There are many other error codes that can be returned by
phylink_of_phy_connect. For example, phylink_validate returns -EINVAL.
So this is a practical issue too.

Fixes: aab9c4067d23 ("net: dsa: Plug in PHYLINK support")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Link: https://lore.kernel.org/r/20210914134331.2303380-1-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agoqnx4: avoid stringop-overread errors
Linus Torvalds [Wed, 15 Sep 2021 20:56:37 +0000 (13:56 -0700)]
qnx4: avoid stringop-overread errors

The qnx4 directory entries are 64-byte blocks that have different
contents depending on the a status byte that is in the last byte of the
block.

In particular, a directory entry can be either a "link info" entry with
a 48-byte name and pointers to the real inode information, or an "inode
entry" with a smaller 16-byte name and the full inode information.

But the code was written to always just treat the directory name as if
it was part of that "inode entry", and just extend the name to the
longer case if the status byte said it was a link entry.

That work just fine and gives the right results, but now that gcc is
tracking data structure accesses much more, the code can trigger a
compiler error about using up to 48 bytes (the long name) in a structure
that only has that shorter name in it:

   fs/qnx4/dir.c: In function ‘qnx4_readdir’:
   fs/qnx4/dir.c:51:32: error: ‘strnlen’ specified bound 48 exceeds source size 16 [-Werror=stringop-overread]
      51 |                         size = strnlen(de->di_fname, size);
         |                                ^~~~~~~~~~~~~~~~~~~~~~~~~~~
   In file included from fs/qnx4/qnx4.h:3,
                    from fs/qnx4/dir.c:16:
   include/uapi/linux/qnx4_fs.h:45:25: note: source object declared here
      45 |         char            di_fname[QNX4_SHORT_NAME_MAX];
         |                         ^~~~~~~~

which is because the source code doesn't really make this whole "one of
two different types" explicit.

Fix this by introducing a very explicit union of the two types, and
basically explaining to the compiler what is really going on.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agosparc: avoid stringop-overread errors
Linus Torvalds [Mon, 6 Sep 2021 23:06:04 +0000 (16:06 -0700)]
sparc: avoid stringop-overread errors

The sparc mdesc code does pointer games with 'struct mdesc_hdr', but
didn't describe to the compiler how that header is then followed by the
data that the header describes.

As a result, gcc is now unhappy since it does stricter pointer range
tracking, and doesn't understand about how these things work.  This
results in various errors like:

    arch/sparc/kernel/mdesc.c: In function ‘mdesc_node_by_name’:
    arch/sparc/kernel/mdesc.c:647:22: error: ‘strcmp’ reading 1 or more bytes from a region of size 0 [-Werror=stringop-overread]
      647 |                 if (!strcmp(names + ep[ret].name_offset, name))
          |                      ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

which are easily avoided by just describing 'struct mdesc_hdr' better,
and making the node_block() helper function look into that unsized
data[] that follows the header.

This makes the sparc64 build happy again at least for my cross-compiler
version (gcc version 11.2.1).

Link: https://lore.kernel.org/lkml/CAHk-=wi4NW3NC0xWykkw=6LnjQD6D_rtRtxY9g8gQAJXtQMi8A@mail.gmail.com/
Cc: Guenter Roeck <linux@roeck-us.net>
Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agoMerge branch 'absolute-pointer' (patches from Guenter)
Linus Torvalds [Wed, 15 Sep 2021 19:11:48 +0000 (12:11 -0700)]
Merge branch 'absolute-pointer' (patches from Guenter)

Merge absolute_pointer macro series from Guenter Roeck:
 "Kernel test builds currently fail for several architectures with error
  messages such as the following.

  drivers/net/ethernet/i825xx/82596.c: In function 'i82596_probe':
  arch/m68k/include/asm/string.h:72:25: error:
        '__builtin_memcpy' reading 6 bytes from a region of size 0
                [-Werror=stringop-overread]

  Such warnings may be reported by gcc 11.x for string and memory
  operations on fixed addresses if gcc's builtin functions are used for
  those operations.

  This series introduces absolute_pointer() to fix the problem.
  absolute_pointer() disassociates a pointer from its originating symbol
  type and context, and thus prevents gcc from making assumptions about
  pointers passed to memory operations"

* emailed patches from Guenter Roeck <linux@roeck-us.net>:
  alpha: Use absolute_pointer to define COMMAND_LINE
  alpha: Move setup.h out of uapi
  net: i825xx: Use absolute_pointer for memcpy from fixed memory location
  compiler.h: Introduce absolute_pointer macro

3 years agoalpha: Use absolute_pointer to define COMMAND_LINE
Guenter Roeck [Wed, 15 Sep 2021 03:52:27 +0000 (20:52 -0700)]
alpha: Use absolute_pointer to define COMMAND_LINE

alpha:allmodconfig fails to build with the following error
when using gcc 11.x.

  arch/alpha/kernel/setup.c: In function 'setup_arch':
  arch/alpha/kernel/setup.c:493:13: error:
'strcmp' reading 1 or more bytes from a region of size 0

Avoid the problem by declaring COMMAND_LINE as absolute_pointer().

Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agoalpha: Move setup.h out of uapi
Guenter Roeck [Wed, 15 Sep 2021 03:52:26 +0000 (20:52 -0700)]
alpha: Move setup.h out of uapi

Most of the contents of setup.h have no value for userspace
applications.  The file was probably moved to uapi accidentally.

Keep the file in uapi to define the alpha-specific COMMAND_LINE_SIZE.
Move all other defines to arch/alpha/include/asm/setup.h.

Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agonet: i825xx: Use absolute_pointer for memcpy from fixed memory location
Guenter Roeck [Wed, 15 Sep 2021 03:52:25 +0000 (20:52 -0700)]
net: i825xx: Use absolute_pointer for memcpy from fixed memory location

gcc 11.x reports the following compiler warning/error.

  drivers/net/ethernet/i825xx/82596.c: In function 'i82596_probe':
  arch/m68k/include/asm/string.h:72:25: error:
'__builtin_memcpy' reading 6 bytes from a region of size 0 [-Werror=stringop-overread]

Use absolute_pointer() to work around the problem.

Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Reviewed-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agocompiler.h: Introduce absolute_pointer macro
Guenter Roeck [Wed, 15 Sep 2021 03:52:24 +0000 (20:52 -0700)]
compiler.h: Introduce absolute_pointer macro

absolute_pointer() disassociates a pointer from its originating symbol
type and context. Use it to prevent compiler warnings/errors such as

  drivers/net/ethernet/i825xx/82596.c: In function 'i82596_probe':
  arch/m68k/include/asm/string.h:72:25: error:
'__builtin_memcpy' reading 6 bytes from a region of size 0 [-Werror=stringop-overread]

Such warnings may be reported by gcc 11.x for string and memory
operations on fixed addresses.

Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Reviewed-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agotools/bootconfig: Define memblock_free_ptr() to fix build error
Masami Hiramatsu [Wed, 15 Sep 2021 00:53:38 +0000 (09:53 +0900)]
tools/bootconfig: Define memblock_free_ptr() to fix build error

The lib/bootconfig.c file is shared with the 'bootconfig' tooling, and
as a result, the changes incommit 77e02cf57b6c ("memblock: introduce
saner 'memblock_free_ptr()' interface") need to also be reflected in the
tooling header file.

So define the new memblock_free_ptr() wrapper, and remove unused __pa()
and memblock_free().

Fixes: 77e02cf57b6c ("memblock: introduce saner 'memblock_free_ptr()' interface")
Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agoptp: dp83640: don't define PAGE0
Randy Dunlap [Mon, 13 Sep 2021 22:06:05 +0000 (15:06 -0700)]
ptp: dp83640: don't define PAGE0

Building dp83640.c on arch/parisc/ produces a build warning for
PAGE0 being redefined. Since the macro is not used in the dp83640
driver, just make it a comment for documentation purposes.

In file included from ../drivers/net/phy/dp83640.c:23:
../drivers/net/phy/dp83640_reg.h:8: warning: "PAGE0" redefined
    8 | #define PAGE0                     0x0000
                 from ../drivers/net/phy/dp83640.c:11:
../arch/parisc/include/asm/page.h:187: note: this is the location of the previous definition
  187 | #define PAGE0   ((struct zeropage *)__PAGE_OFFSET)

Fixes: cb646e2b02b2 ("ptp: Added a clock driver for the National Semiconductor PHYTER.")
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Reported-by: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Richard Cochran <richard.cochran@omicron.at>
Cc: John Stultz <john.stultz@linaro.org>
Cc: Heiner Kallweit <hkallweit1@gmail.com>
Cc: Russell King <linux@armlinux.org.uk>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://lore.kernel.org/r/20210913220605.19682-1-rdunlap@infradead.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agobnx2x: Fix enabling network interfaces without VFs
Adrian Bunk [Sun, 12 Sep 2021 19:05:23 +0000 (22:05 +0300)]
bnx2x: Fix enabling network interfaces without VFs

This function is called to enable SR-IOV when available,
not enabling interfaces without VFs was a regression.

Fixes: 65161c35554f ("bnx2x: Fix missing error code in bnx2x_iov_init_one()")
Signed-off-by: Adrian Bunk <bunk@kernel.org>
Reported-by: YunQiang Su <wzssyqa@gmail.com>
Tested-by: YunQiang Su <wzssyqa@gmail.com>
Cc: stable@vger.kernel.org
Acked-by: Shai Malin <smalin@marvell.com>
Link: https://lore.kernel.org/r/20210912190523.27991-1-bunk@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
3 years agodrm/ttm: fix type mismatch error on sparc64
Huang Rui [Tue, 7 Sep 2021 10:03:02 +0000 (18:03 +0800)]
drm/ttm: fix type mismatch error on sparc64

On sparc64, __fls() returns an "int", but the drm TTM code expected it
to be "unsigned long" as on x86.  As a result, on sparc (and arc, and
m68k) you get build errors because 'min()' checks that the types match.

As suggested by Linus, it can use min_t instead of min to force the type
to be "unsigned int".

Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Huang Rui <ray.huang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Cc: Alex Deucher <alexdeucher@gmail.com>
Cc: David Airlie <airlied@linux.ie>
Cc: Daniel Vetter <daniel@ffwll.ch>
Cc: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomemblock: introduce saner 'memblock_free_ptr()' interface
Linus Torvalds [Tue, 14 Sep 2021 20:23:22 +0000 (13:23 -0700)]
memblock: introduce saner 'memblock_free_ptr()' interface

The boot-time allocation interface for memblock is a mess, with
'memblock_alloc()' returning a virtual pointer, but then you are
supposed to free it with 'memblock_free()' that takes a _physical_
address.

Not only is that all kinds of strange and illogical, but it actually
causes bugs, when people then use it like a normal allocation function,
and it fails spectacularly on a NULL pointer:

   https://lore.kernel.org/all/20210912140820.GD25450@xsang-OptiPlex-9020/

or just random memory corruption if the debug checks don't catch it:

   https://lore.kernel.org/all/61ab2d0c-3313-aaab-514c-e15b7aa054a0@suse.cz/

I really don't want to apply patches that treat the symptoms, when the
fundamental cause is this horribly confusing interface.

I started out looking at just automating a sane replacement sequence,
but because of this mix or virtual and physical addresses, and because
people have used the "__pa()" macro that can take either a regular
kernel pointer, or just the raw "unsigned long" address, it's all quite
messy.

So this just introduces a new saner interface for freeing a virtual
address that was allocated using 'memblock_alloc()', and that was kept
as a regular kernel pointer.  And then it converts a couple of users
that are obvious and easy to test, including the 'xbc_nodes' case in
lib/bootconfig.c that caused problems.

Reported-by: kernel test robot <oliver.sang@intel.com>
Fixes: 40caa127f3c7 ("init: bootconfig: Remove all bootconfig data when the init memory is removed")
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agoipc: remove memcg accounting for sops objects in do_semtimedop()
Vasily Averin [Sat, 11 Sep 2021 07:40:08 +0000 (10:40 +0300)]
ipc: remove memcg accounting for sops objects in do_semtimedop()

Linus proposes to revert an accounting for sops objects in
do_semtimedop() because it's really just a temporary buffer
for a single semtimedop() system call.

This object can consume up to 2 pages, syscall is sleeping
one, size and duration can be controlled by user, and this
allocation can be repeated by many thread at the same time.

However Shakeel Butt pointed that there are much more popular
objects with the same life time and similar memory
consumption, the accounting of which was decided to be
rejected for performance reasons.

Considering at least 2 pages for task_struct and 2 pages for
the kernel stack, a back of the envelope calculation gives a
footprint amplification of <1.5 so this temporal buffer can be
safely ignored.

The factor would IMO be interesting if it was >> 2 (from the
PoV of excessive (ab)use, fine-grained accounting seems to be
currently unfeasible due to performance impact).

Link: https://lore.kernel.org/lkml/90e254df-0dfe-f080-011e-b7c53ee7fd20@virtuozzo.com/
Fixes: 18319498fdd4 ("memcg: enable accounting of ipc resources")
Signed-off-by: Vasily Averin <vvs@virtuozzo.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Michal Koutný <mkoutny@suse.com>
Acked-by: Shakeel Butt <shakeelb@google.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agopowerpc/boot: Fix build failure since GCC 4.9 removal
Michael Ellerman [Tue, 14 Sep 2021 12:17:23 +0000 (22:17 +1000)]
powerpc/boot: Fix build failure since GCC 4.9 removal

Stephen reported that the build was broken since commit
6d2ef226f2f1 ("compiler_attributes.h: drop __has_attribute() support for
gcc4"), with errors such as:

  include/linux/compiler_attributes.h:296:5: warning: "__has_attribute" is not defined, evaluates to 0 [-Wundef]
    296 | #if __has_attribute(__warning__)
        |     ^~~~~~~~~~~~~~~
  make[2]: *** [arch/powerpc/boot/Makefile:225: arch/powerpc/boot/crt0.o] Error 1

But we expect __has_attribute() to always be defined now that we've
stopped using GCC 4.

Linus debugged it to the point of reading the GCC sources, and noticing
that the problem is that __has_attribute() is not defined when
preprocessing assembly files, which is what we're doing here.

Our assembly files don't include, or need, compiler_attributes.h, but
they are getting it unconditionally from the -include in BOOT_CFLAGS,
which is then added in its entirety to BOOT_AFLAGS.

That -include was added in commit 77433830ed16 ("powerpc: boot: include
compiler_attributes.h") so that we'd have "fallthrough" and other
attributes defined for the C files in arch/powerpc/boot. But it's not
needed for assembly files.

The minimal fix is to move the addition to BOOT_CFLAGS of -include
compiler_attributes.h until after we've copied BOOT_CFLAGS into
BOOT_AFLAGS. That avoids including compiler_attributes.h for asm files,
but makes no other change to BOOT_CFLAGS or BOOT_AFLAGS.

Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Debugged-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agoRevert "Revert "ipv4: fix memory leaks in ip_cmsg_send() callers""
Eric Dumazet [Tue, 14 Sep 2021 05:18:51 +0000 (22:18 -0700)]
Revert "Revert "ipv4: fix memory leaks in ip_cmsg_send() callers""

This reverts commit d7807a9adf4856171f8441f13078c33941df48ab.

As mentioned in https://lkml.org/lkml/2021/9/13/1819
5 years old commit 919483096bfe ("ipv4: fix memory leaks in ip_cmsg_send() callers")
was a correct fix.

  ip_cmsg_send() can loop over multiple cmsghdr()

  If IP_RETOPTS has been successful, but following cmsghdr generates an error,
  we do not free ipc.ok

  If IP_RETOPTS is not successful, we have freed the allocated temporary space,
  not the one currently in ipc.opt.

Sure, code could be refactored, but let's not bring back old bugs.

Fixes: d7807a9adf48 ("Revert "ipv4: fix memory leaks in ip_cmsg_send() callers"")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Yajun Deng <yajun.deng@linux.dev>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agotcp: fix tp->undo_retrans accounting in tcp_sacktag_one()
zhenggy [Tue, 14 Sep 2021 01:51:15 +0000 (09:51 +0800)]
tcp: fix tp->undo_retrans accounting in tcp_sacktag_one()

Commit 10d3be569243 ("tcp-tso: do not split TSO packets at retransmit
time") may directly retrans a multiple segments TSO/GSO packet without
split, Since this commit, we can no longer assume that a retransmitted
packet is a single segment.

This patch fixes the tp->undo_retrans accounting in tcp_sacktag_one()
that use the actual segments(pcount) of the retransmitted packet.

Before that commit (10d3be569243), the assumption underlying the
tp->undo_retrans-- seems correct.

Fixes: 10d3be569243 ("tcp-tso: do not split TSO packets at retransmit time")
Signed-off-by: zhenggy <zhenggy@chinatelecom.cn>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Acked-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf
David S. Miller [Tue, 14 Sep 2021 12:09:54 +0000 (13:09 +0100)]
Merge git://git./pub/scm/linux/kernel/git/bpf/bpf

Daniel Borkmann says:

====================
pull-request: bpf 2021-09-14

The following pull-request contains BPF updates for your *net* tree.

We've added 7 non-merge commits during the last 13 day(s) which contain
a total of 18 files changed, 334 insertions(+), 193 deletions(-).

The main changes are:

1) Fix mmap_lock lockdep splat in BPF stack map's build_id lookup, from Yonghong Song.

2) Fix BPF cgroup v2 program bypass upon net_cls/prio activation, from Daniel Borkmann.

3) Fix kvcalloc() BTF line info splat on oversized allocation attempts, from Bixuan Cui.

4) Fix BPF selftest build of task_pt_regs test for arm64/s390, from Jean-Philippe Brucker.

5) Fix BPF's disasm.{c,h} to dual-license so that it is aligned with bpftool given the former
   is a build dependency for the latter, from Daniel Borkmann with ACKs from contributors.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet-caif: avoid user-triggerable WARN_ON(1)
Eric Dumazet [Mon, 13 Sep 2021 18:08:36 +0000 (11:08 -0700)]
net-caif: avoid user-triggerable WARN_ON(1)

syszbot triggers this warning, which looks something
we can easily prevent.

If we initialize priv->list_field in chnl_net_init(),
then always use list_del_init(), we can remove robust_list_del()
completely.

WARNING: CPU: 0 PID: 3233 at net/caif/chnl_net.c:67 robust_list_del net/caif/chnl_net.c:67 [inline]
WARNING: CPU: 0 PID: 3233 at net/caif/chnl_net.c:67 chnl_net_uninit+0xc9/0x2e0 net/caif/chnl_net.c:375
Modules linked in:
CPU: 0 PID: 3233 Comm: syz-executor.3 Not tainted 5.14.0-syzkaller #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
RIP: 0010:robust_list_del net/caif/chnl_net.c:67 [inline]
RIP: 0010:chnl_net_uninit+0xc9/0x2e0 net/caif/chnl_net.c:375
Code: 89 eb e8 3a a3 ba f8 48 89 d8 48 c1 e8 03 42 80 3c 28 00 0f 85 bf 01 00 00 48 81 fb 00 14 4e 8d 48 8b 2b 75 d0 e8 17 a3 ba f8 <0f> 0b 5b 5d 41 5c 41 5d e9 0a a3 ba f8 4c 89 e3 e8 02 a3 ba f8 4c
RSP: 0018:ffffc90009067248 EFLAGS: 00010202
RAX: 0000000000008780 RBX: ffffffff8d4e1400 RCX: ffffc9000fd34000
RDX: 0000000000040000 RSI: ffffffff88bb6e49 RDI: 0000000000000003
RBP: ffff88802cd9ee08 R08: 0000000000000000 R09: ffffffff8d0e6647
R10: ffffffff88bb6dc2 R11: 0000000000000000 R12: ffff88803791ae08
R13: dffffc0000000000 R14: 00000000e600ffce R15: ffff888073ed3480
FS:  00007fed10fa0700(0000) GS:ffff8880b9d00000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000001b2c322000 CR3: 00000000164a6000 CR4: 00000000001506e0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
Call Trace:
 register_netdevice+0xadf/0x1500 net/core/dev.c:10347
 ipcaif_newlink+0x4c/0x260 net/caif/chnl_net.c:468
 __rtnl_newlink+0x106d/0x1750 net/core/rtnetlink.c:3458
 rtnl_newlink+0x64/0xa0 net/core/rtnetlink.c:3506
 rtnetlink_rcv_msg+0x413/0xb80 net/core/rtnetlink.c:5572
 netlink_rcv_skb+0x153/0x420 net/netlink/af_netlink.c:2504
 netlink_unicast_kernel net/netlink/af_netlink.c:1314 [inline]
 netlink_unicast+0x533/0x7d0 net/netlink/af_netlink.c:1340
 netlink_sendmsg+0x86d/0xdb0 net/netlink/af_netlink.c:1929
 sock_sendmsg_nosec net/socket.c:704 [inline]
 sock_sendmsg+0xcf/0x120 net/socket.c:724
 __sys_sendto+0x21c/0x320 net/socket.c:2036
 __do_sys_sendto net/socket.c:2048 [inline]
 __se_sys_sendto net/socket.c:2044 [inline]
 __x64_sys_sendto+0xdd/0x1b0 net/socket.c:2044
 do_syscall_x64 arch/x86/entry/common.c:50 [inline]
 do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80
 entry_SYSCALL_64_after_hwframe+0x44/0xae

Fixes: cc36a070b590 ("net-caif: add CAIF netdevice")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agortc: cmos: Disable irq around direct invocation of cmos_interrupt()
Chris Wilson [Fri, 5 Mar 2021 12:21:40 +0000 (12:21 +0000)]
rtc: cmos: Disable irq around direct invocation of cmos_interrupt()

As previously noted in commit 66e4f4a9cc38 ("rtc: cmos: Use
spin_lock_irqsave() in cmos_interrupt()"):

<4>[  254.192378] WARNING: inconsistent lock state
<4>[  254.192384] 5.12.0-rc1-CI-CI_DRM_9834+ #1 Not tainted
<4>[  254.192396] --------------------------------
<4>[  254.192400] inconsistent {IN-HARDIRQ-W} -> {HARDIRQ-ON-W} usage.
<4>[  254.192409] rtcwake/5309 [HC0[0]:SC0[0]:HE1:SE1] takes:
<4>[  254.192429] ffffffff8263c5f8 (rtc_lock){?...}-{2:2}, at: cmos_interrupt+0x18/0x100
<4>[  254.192481] {IN-HARDIRQ-W} state was registered at:
<4>[  254.192488]   lock_acquire+0xd1/0x3d0
<4>[  254.192504]   _raw_spin_lock+0x2a/0x40
<4>[  254.192519]   cmos_interrupt+0x18/0x100
<4>[  254.192536]   rtc_handler+0x1f/0xc0
<4>[  254.192553]   acpi_ev_fixed_event_detect+0x109/0x13c
<4>[  254.192574]   acpi_ev_sci_xrupt_handler+0xb/0x28
<4>[  254.192596]   acpi_irq+0x13/0x30
<4>[  254.192620]   __handle_irq_event_percpu+0x43/0x2c0
<4>[  254.192641]   handle_irq_event_percpu+0x2b/0x70
<4>[  254.192661]   handle_irq_event+0x2f/0x50
<4>[  254.192680]   handle_fasteoi_irq+0x9e/0x150
<4>[  254.192693]   __common_interrupt+0x76/0x140
<4>[  254.192715]   common_interrupt+0x96/0xc0
<4>[  254.192732]   asm_common_interrupt+0x1e/0x40
<4>[  254.192750]   _raw_spin_unlock_irqrestore+0x38/0x60
<4>[  254.192767]   resume_irqs+0xba/0xf0
<4>[  254.192786]   dpm_resume_noirq+0x245/0x3d0
<4>[  254.192811]   suspend_devices_and_enter+0x230/0xaa0
<4>[  254.192835]   pm_suspend.cold.8+0x301/0x34a
<4>[  254.192859]   state_store+0x7b/0xe0
<4>[  254.192879]   kernfs_fop_write_iter+0x11d/0x1c0
<4>[  254.192899]   new_sync_write+0x11d/0x1b0
<4>[  254.192916]   vfs_write+0x265/0x390
<4>[  254.192933]   ksys_write+0x5a/0xd0
<4>[  254.192949]   do_syscall_64+0x33/0x80
<4>[  254.192965]   entry_SYSCALL_64_after_hwframe+0x44/0xae
<4>[  254.192986] irq event stamp: 43775
<4>[  254.192994] hardirqs last  enabled at (43775): [<ffffffff81c00c42>] asm_sysvec_apic_timer_interrupt+0x12/0x20
<4>[  254.193023] hardirqs last disabled at (43774): [<ffffffff81aa691a>] sysvec_apic_timer_interrupt+0xa/0xb0
<4>[  254.193049] softirqs last  enabled at (42548): [<ffffffff81e00342>] __do_softirq+0x342/0x48e
<4>[  254.193074] softirqs last disabled at (42543): [<ffffffff810b45fd>] irq_exit_rcu+0xad/0xd0
<4>[  254.193101]
                  other info that might help us debug this:
<4>[  254.193107]  Possible unsafe locking scenario:

<4>[  254.193112]        CPU0
<4>[  254.193117]        ----
<4>[  254.193121]   lock(rtc_lock);
<4>[  254.193137]   <Interrupt>
<4>[  254.193142]     lock(rtc_lock);
<4>[  254.193156]
                   *** DEADLOCK ***

<4>[  254.193161] 6 locks held by rtcwake/5309:
<4>[  254.193174]  #0: ffff888104861430 (sb_writers#5){.+.+}-{0:0}, at: ksys_write+0x5a/0xd0
<4>[  254.193232]  #1: ffff88810f823288 (&of->mutex){+.+.}-{3:3}, at: kernfs_fop_write_iter+0xe7/0x1c0
<4>[  254.193282]  #2: ffff888100cef3c0 (kn->active#285
<7>[  254.192706] i915 0000:00:02.0: [drm:intel_modeset_setup_hw_state [i915]] [CRTC:51:pipe A] hw state readout: disabled
<4>[  254.193307] ){.+.+}-{0:0}, at: kernfs_fop_write_iter+0xf0/0x1c0
<4>[  254.193333]  #3: ffffffff82649fa8 (system_transition_mutex){+.+.}-{3:3}, at: pm_suspend.cold.8+0xce/0x34a
<4>[  254.193387]  #4: ffffffff827a2108 (acpi_scan_lock){+.+.}-{3:3}, at: acpi_suspend_begin+0x47/0x70
<4>[  254.193433]  #5: ffff8881019ea178 (&dev->mutex){....}-{3:3}, at: device_resume+0x68/0x1e0
<4>[  254.193485]
                  stack backtrace:
<4>[  254.193492] CPU: 1 PID: 5309 Comm: rtcwake Not tainted 5.12.0-rc1-CI-CI_DRM_9834+ #1
<4>[  254.193514] Hardware name: Google Soraka/Soraka, BIOS MrChromebox-4.10 08/25/2019
<4>[  254.193524] Call Trace:
<4>[  254.193536]  dump_stack+0x7f/0xad
<4>[  254.193567]  mark_lock.part.47+0x8ca/0xce0
<4>[  254.193604]  __lock_acquire+0x39b/0x2590
<4>[  254.193626]  ? asm_sysvec_apic_timer_interrupt+0x12/0x20
<4>[  254.193660]  lock_acquire+0xd1/0x3d0
<4>[  254.193677]  ? cmos_interrupt+0x18/0x100
<4>[  254.193716]  _raw_spin_lock+0x2a/0x40
<4>[  254.193735]  ? cmos_interrupt+0x18/0x100
<4>[  254.193758]  cmos_interrupt+0x18/0x100
<4>[  254.193785]  cmos_resume+0x2ac/0x2d0
<4>[  254.193813]  ? acpi_pm_set_device_wakeup+0x1f/0x110
<4>[  254.193842]  ? pnp_bus_suspend+0x10/0x10
<4>[  254.193864]  pnp_bus_resume+0x5e/0x90
<4>[  254.193885]  dpm_run_callback+0x5f/0x240
<4>[  254.193914]  device_resume+0xb2/0x1e0
<4>[  254.193942]  ? pm_dev_err+0x25/0x25
<4>[  254.193974]  dpm_resume+0xea/0x3f0
<4>[  254.194005]  dpm_resume_end+0x8/0x10
<4>[  254.194030]  suspend_devices_and_enter+0x29b/0xaa0
<4>[  254.194066]  pm_suspend.cold.8+0x301/0x34a
<4>[  254.194094]  state_store+0x7b/0xe0
<4>[  254.194124]  kernfs_fop_write_iter+0x11d/0x1c0
<4>[  254.194151]  new_sync_write+0x11d/0x1b0
<4>[  254.194183]  vfs_write+0x265/0x390
<4>[  254.194207]  ksys_write+0x5a/0xd0
<4>[  254.194232]  do_syscall_64+0x33/0x80
<4>[  254.194251]  entry_SYSCALL_64_after_hwframe+0x44/0xae
<4>[  254.194274] RIP: 0033:0x7f07d79691e7
<4>[  254.194293] Code: 64 89 02 48 c7 c0 ff ff ff ff eb bb 0f 1f 80 00 00 00 00 f3 0f 1e fa 64 8b 04 25 18 00 00 00 85 c0 75 10 b8 01 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 51 c3 48 83 ec 28 48 89 54 24 18 48 89 74 24
<4>[  254.194312] RSP: 002b:00007ffd9cc2c768 EFLAGS: 00000246 ORIG_RAX: 0000000000000001
<4>[  254.194337] RAX: ffffffffffffffda RBX: 0000000000000004 RCX: 00007f07d79691e7
<4>[  254.194352] RDX: 0000000000000004 RSI: 0000556ebfc63590 RDI: 000000000000000b
<4>[  254.194366] RBP: 0000556ebfc63590 R08: 0000000000000000 R09: 0000000000000004
<4>[  254.194379] R10: 0000556ebf0ec2a6 R11: 0000000000000246 R12: 0000000000000004

which breaks S3-resume on fi-kbl-soraka presumably as that's slow enough
to trigger the alarm during the suspend.

Fixes: 6950d046eb6e ("rtc: cmos: Replace spin_lock_irqsave with spin_lock in hard IRQ")
References: 66e4f4a9cc38 ("rtc: cmos: Use spin_lock_irqsave() in cmos_interrupt()"):
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Xiaofei Tan <tanxiaofei@huawei.com>
Cc: Alexandre Belloni <alexandre.belloni@bootlin.com>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
Link: https://lore.kernel.org/r/20210305122140.28774-1-chris@chris-wilson.co.uk
3 years agobpf, selftests: Add test case for mixed cgroup v1/v2
Daniel Borkmann [Mon, 13 Sep 2021 23:07:59 +0000 (01:07 +0200)]
bpf, selftests: Add test case for mixed cgroup v1/v2

Minimal selftest which implements a small BPF policy program to the
connect(2) hook which rejects TCP connection requests to port 60123
with EPERM. This is being attached to a non-root cgroup v2 path. The
test asserts that this works under cgroup v2-only and under a mixed
cgroup v1/v2 environment where net_classid is set in the former case.

Before fix:

  # ./test_progs -t cgroup_v1v2
  test_cgroup_v1v2:PASS:server_fd 0 nsec
  test_cgroup_v1v2:PASS:client_fd 0 nsec
  test_cgroup_v1v2:PASS:cgroup_fd 0 nsec
  test_cgroup_v1v2:PASS:server_fd 0 nsec
  run_test:PASS:skel_open 0 nsec
  run_test:PASS:prog_attach 0 nsec
  test_cgroup_v1v2:PASS:cgroup-v2-only 0 nsec
  run_test:PASS:skel_open 0 nsec
  run_test:PASS:prog_attach 0 nsec
  run_test:PASS:join_classid 0 nsec
  (network_helpers.c:219: errno: None) Unexpected success to connect to server
  test_cgroup_v1v2:FAIL:cgroup-v1v2 unexpected error: -1 (errno 0)
  #27 cgroup_v1v2:FAIL
  Summary: 0/0 PASSED, 0 SKIPPED, 1 FAILED

After fix:

  # ./test_progs -t cgroup_v1v2
  #27 cgroup_v1v2:OK
  Summary: 1/0 PASSED, 0 SKIPPED, 0 FAILED

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20210913230759.2313-3-daniel@iogearbox.net
3 years agobpf, selftests: Add cgroup v1 net_cls classid helpers
Daniel Borkmann [Mon, 13 Sep 2021 23:07:58 +0000 (01:07 +0200)]
bpf, selftests: Add cgroup v1 net_cls classid helpers

Minimal set of helpers for net_cls classid cgroupv1 management in order
to set an id, join from a process, initiate setup and teardown. cgroupv2
helpers are left as-is, but reused where possible.

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20210913230759.2313-2-daniel@iogearbox.net
3 years agobpf, cgroups: Fix cgroup v2 fallback on v1/v2 mixed mode
Daniel Borkmann [Mon, 13 Sep 2021 23:07:57 +0000 (01:07 +0200)]
bpf, cgroups: Fix cgroup v2 fallback on v1/v2 mixed mode

Fix cgroup v1 interference when non-root cgroup v2 BPF programs are used.
Back in the days, commit bd1060a1d671 ("sock, cgroup: add sock->sk_cgroup")
embedded per-socket cgroup information into sock->sk_cgrp_data and in order
to save 8 bytes in struct sock made both mutually exclusive, that is, when
cgroup v1 socket tagging (e.g. net_cls/net_prio) is used, then cgroup v2
falls back to the root cgroup in sock_cgroup_ptr() (&cgrp_dfl_root.cgrp).

The assumption made was "there is no reason to mix the two and this is in line
with how legacy and v2 compatibility is handled" as stated in bd1060a1d671.
However, with Kubernetes more widely supporting cgroups v2 as well nowadays,
this assumption no longer holds, and the possibility of the v1/v2 mixed mode
with the v2 root fallback being hit becomes a real security issue.

Many of the cgroup v2 BPF programs are also used for policy enforcement, just
to pick _one_ example, that is, to programmatically deny socket related system
calls like connect(2) or bind(2). A v2 root fallback would implicitly cause
a policy bypass for the affected Pods.

In production environments, we have recently seen this case due to various
circumstances: i) a different 3rd party agent and/or ii) a container runtime
such as [0] in the user's environment configuring legacy cgroup v1 net_cls
tags, which triggered implicitly mentioned root fallback. Another case is
Kubernetes projects like kind [1] which create Kubernetes nodes in a container
and also add cgroup namespaces to the mix, meaning programs which are attached
to the cgroup v2 root of the cgroup namespace get attached to a non-root
cgroup v2 path from init namespace point of view. And the latter's root is
out of reach for agents on a kind Kubernetes node to configure. Meaning, any
entity on the node setting cgroup v1 net_cls tag will trigger the bypass
despite cgroup v2 BPF programs attached to the namespace root.

Generally, this mutual exclusiveness does not hold anymore in today's user
environments and makes cgroup v2 usage from BPF side fragile and unreliable.
This fix adds proper struct cgroup pointer for the cgroup v2 case to struct
sock_cgroup_data in order to address these issues; this implicitly also fixes
the tradeoffs being made back then with regards to races and refcount leaks
as stated in bd1060a1d671, and removes the fallback, so that cgroup v2 BPF
programs always operate as expected.

  [0] https://github.com/nestybox/sysbox/
  [1] https://kind.sigs.k8s.io/

Fixes: bd1060a1d671 ("sock, cgroup: add sock->sk_cgroup")
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Stanislav Fomichev <sdf@google.com>
Acked-by: Tejun Heo <tj@kernel.org>
Link: https://lore.kernel.org/bpf/20210913230759.2313-1-daniel@iogearbox.net
3 years agobpf: Add oversize check before call kvcalloc()
Bixuan Cui [Sat, 11 Sep 2021 00:55:57 +0000 (08:55 +0800)]
bpf: Add oversize check before call kvcalloc()

Commit 7661809d493b ("mm: don't allow oversized kvmalloc() calls") add the
oversize check. When the allocation is larger than what kmalloc() supports,
the following warning triggered:

WARNING: CPU: 0 PID: 8408 at mm/util.c:597 kvmalloc_node+0x108/0x110 mm/util.c:597
Modules linked in:
CPU: 0 PID: 8408 Comm: syz-executor221 Not tainted 5.14.0-syzkaller #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
RIP: 0010:kvmalloc_node+0x108/0x110 mm/util.c:597
Call Trace:
 kvmalloc include/linux/mm.h:806 [inline]
 kvmalloc_array include/linux/mm.h:824 [inline]
 kvcalloc include/linux/mm.h:829 [inline]
 check_btf_line kernel/bpf/verifier.c:9925 [inline]
 check_btf_info kernel/bpf/verifier.c:10049 [inline]
 bpf_check+0xd634/0x150d0 kernel/bpf/verifier.c:13759
 bpf_prog_load kernel/bpf/syscall.c:2301 [inline]
 __sys_bpf+0x11181/0x126e0 kernel/bpf/syscall.c:4587
 __do_sys_bpf kernel/bpf/syscall.c:4691 [inline]
 __se_sys_bpf kernel/bpf/syscall.c:4689 [inline]
 __x64_sys_bpf+0x78/0x90 kernel/bpf/syscall.c:4689
 do_syscall_x64 arch/x86/entry/common.c:50 [inline]
 do_syscall_64+0x3d/0xb0 arch/x86/entry/common.c:80
 entry_SYSCALL_64_after_hwframe+0x44/0xae

Reported-by: syzbot+f3e749d4c662818ae439@syzkaller.appspotmail.com
Signed-off-by: Bixuan Cui <cuibixuan@huawei.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Yonghong Song <yhs@fb.com>
Link: https://lore.kernel.org/bpf/20210911005557.45518-1-cuibixuan@huawei.com
3 years agotools: compiler-gcc.h: Guard error attribute use with __has_attribute
Nathan Chancellor [Mon, 13 Sep 2021 22:09:00 +0000 (15:09 -0700)]
tools: compiler-gcc.h: Guard error attribute use with __has_attribute

When building objtool with HOSTCC=clang, there are several errors along
the lines of

  orc_dump.c:201:28: error: unknown attribute 'error' ignored [-Werror,-Wunknown-attributes]

This occurs after commit 4e59869aa655 ("compiler-gcc.h: drop checks for
older GCC versions"), which removed the GCC_VERSION gating.  The removed
version check just so happened to prevent __compiletime_error() from
being defined with clang because it pretends to be GCC 4.2.1 for
compatibility but the error attribute was not added to clang until
14.0.0.

Commit 815f0ddb346c ("include/linux/compiler*.h: make compiler-*.h
mutually exclusive") and commit a3f8a30f3f00 ("Compiler Attributes: use
feature checks instead of version checks") refactored the handling of
attributes in the main kernel to avoid situations like this but that
refactoring has never been done for the tools directory.

Refactoring is a rather large undertaking and this has never been an
issue before so instead, just guard the definition of
__compiletime_error() with __has_attribute() so that there are no more
errors.

Fixes: 4e59869aa655 ("compiler-gcc.h: drop checks for older GCC versions")
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agoMerge branch 'gcc-min-version-5.1' (make gcc-5.1 the minimum version)
Linus Torvalds [Mon, 13 Sep 2021 17:43:04 +0000 (10:43 -0700)]
Merge branch 'gcc-min-version-5.1' (make gcc-5.1 the minimum version)

Merge patch series from Nick Desaulniers to update the minimum gcc
version to 5.1.

This is some of the left-overs from the merge window that I didn't want
to deal with yesterday, so it comes in after -rc1 but was sent before.

Gcc-4.9 support has been an annoyance for some time, and with -Werror I
had the choice of applying a fairly big patch from Kees Cook to remove a
fair number of initializer warnings (still leaving some), or this patch
series from Nick that just removes the source of the problem.

The initializer cleanups might still be worth it regardless, but
honestly, I preferred just tackling the problem with gcc-4.9 head-on.
We've been more aggressiuve about no longer having to care about
compilers that were released a long time ago, and I think it's been a
good thing.

I added a couple of patches on top to sort out a few left-overs now that
we no longer support gcc-4.x.

As noted by Arnd, as a result of this minimum compiler version upgrade
we can probably change our use of '--std=gnu89' to '--std=gnu11', and
finally start using local loop declarations etc.  But this series does
_not_ yet do that.

Link: https://lore.kernel.org/all/20210909182525.372ee687@canb.auug.org.au/
Link: https://lore.kernel.org/lkml/CAK7LNASs6dvU6D3jL2GG3jW58fXfaj6VNOe55NJnTB8UPuk2pA@mail.gmail.com/
Link: https://github.com/ClangBuiltLinux/linux/issues/1438
* emailed patches from Nick Desaulniers <ndesaulniers@google.com>:
  Drop some straggling mentions of gcc-4.9 as being stale
  compiler_attributes.h: drop __has_attribute() support for gcc4
  vmlinux.lds.h: remove old check for GCC 4.9
  compiler-gcc.h: drop checks for older GCC versions
  Makefile: drop GCC < 5 -fno-var-tracking-assignments workaround
  arm64: remove GCC version check for ARCH_SUPPORTS_INT128
  powerpc: remove GCC version check for UPD_CONSTR
  riscv: remove Kconfig check for GCC version for ARCH_RV64I
  Kconfig.debug: drop GCC 5+ version check for DWARF5
  mm/ksm: remove old GCC 4.9+ check
  compiler.h: drop fallback overflow checkers
  Documentation: raise minimum supported version of GCC to 5.1

3 years agoDrop some straggling mentions of gcc-4.9 as being stale
Linus Torvalds [Mon, 13 Sep 2021 17:29:44 +0000 (10:29 -0700)]
Drop some straggling mentions of gcc-4.9 as being stale

Fix up the admin-guide README file to the new gcc-5.1 requirement, and
remove a stale comment about gcc support for the __assume_aligned__
attribute.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agocompiler_attributes.h: drop __has_attribute() support for gcc4
Linus Torvalds [Mon, 13 Sep 2021 17:20:01 +0000 (10:20 -0700)]
compiler_attributes.h: drop __has_attribute() support for gcc4

Now that GCC 5.1 is the minimally supported default, the manual
workaround for older gcc versions not having __has_attribute() are no
longer relevant and can be removed.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agovmlinux.lds.h: remove old check for GCC 4.9
Nick Desaulniers [Fri, 10 Sep 2021 23:40:47 +0000 (16:40 -0700)]
vmlinux.lds.h: remove old check for GCC 4.9

Now that GCC 5.1 is the minimally supported version of GCC, we can
effectively revert commit 85c2ce9104eb ("sched, vmlinux.lds: Increase
STRUCT_ALIGNMENT to 64 bytes for GCC-4.9")

Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>
Acked-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agocompiler-gcc.h: drop checks for older GCC versions
Nick Desaulniers [Fri, 10 Sep 2021 23:40:46 +0000 (16:40 -0700)]
compiler-gcc.h: drop checks for older GCC versions

Now that GCC 5.1 is the minimally supported default, drop the values we
don't use.

Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agoMakefile: drop GCC < 5 -fno-var-tracking-assignments workaround
Nick Desaulniers [Fri, 10 Sep 2021 23:40:45 +0000 (16:40 -0700)]
Makefile: drop GCC < 5 -fno-var-tracking-assignments workaround

Now that GCC 5.1 is the minimally supported version, we can drop this
workaround for older versions of GCC.

Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agoarm64: remove GCC version check for ARCH_SUPPORTS_INT128
Nick Desaulniers [Fri, 10 Sep 2021 23:40:44 +0000 (16:40 -0700)]
arm64: remove GCC version check for ARCH_SUPPORTS_INT128

Now that GCC 5.1 is the minimally supported compiler version, this
Kconfig check is no longer necessary.

Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will@kernel.org>
Cc: linux-arm-kernel@lists.infradead.org
Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agopowerpc: remove GCC version check for UPD_CONSTR
Nick Desaulniers [Fri, 10 Sep 2021 23:40:43 +0000 (16:40 -0700)]
powerpc: remove GCC version check for UPD_CONSTR

Now that GCC 5.1 is the minimum supported version, we can drop this
workaround for older versions of GCC. This adversely affected clang,
too.

Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Segher Boessenkool <segher@kernel.crashing.org>
Cc: Christophe Leroy <christophe.leroy@csgroup.eu>
Cc: linuxppc-dev@lists.ozlabs.org
Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agoriscv: remove Kconfig check for GCC version for ARCH_RV64I
Nick Desaulniers [Fri, 10 Sep 2021 23:40:42 +0000 (16:40 -0700)]
riscv: remove Kconfig check for GCC version for ARCH_RV64I

The minimum supported version of GCC is now 5.1. The check wasn't
correct as written anyways since GCC_VERSION is 0 when CC=clang.

Cc: Paul Walmsley <paul.walmsley@sifive.com>
Cc: Palmer Dabbelt <palmer@dabbelt.com>
Cc: Albert Ou <aou@eecs.berkeley.edu>
Cc: linux-riscv@lists.infradead.org
Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agoKconfig.debug: drop GCC 5+ version check for DWARF5
Nick Desaulniers [Fri, 10 Sep 2021 23:40:41 +0000 (16:40 -0700)]
Kconfig.debug: drop GCC 5+ version check for DWARF5

Now that the minimum supported version of GCC is 5.1, we no longer need
this Kconfig version check for CONFIG_DEBUG_INFO_DWARF5.

Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agomm/ksm: remove old GCC 4.9+ check
Nick Desaulniers [Fri, 10 Sep 2021 23:40:40 +0000 (16:40 -0700)]
mm/ksm: remove old GCC 4.9+ check

The minimum supported version of GCC has been raised to GCC 5.1.

Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agocompiler.h: drop fallback overflow checkers
Nick Desaulniers [Fri, 10 Sep 2021 23:40:39 +0000 (16:40 -0700)]
compiler.h: drop fallback overflow checkers

Once upgrading the minimum supported version of GCC to 5.1, we can drop
the fallback code for !COMPILER_HAS_GENERIC_BUILTIN_OVERFLOW.

This is effectively a revert of commit f0907827a8a9 ("compiler.h: enable
builtin overflow checkers and add fallback code")

Link: https://github.com/ClangBuiltLinux/linux/issues/1438#issuecomment-916745801
Suggested-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>
Acked-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agoDocumentation: raise minimum supported version of GCC to 5.1
Nick Desaulniers [Fri, 10 Sep 2021 23:40:38 +0000 (16:40 -0700)]
Documentation: raise minimum supported version of GCC to 5.1

commit fad7cd3310db ("nbd: add the check to prevent overflow in
__nbd_ioctl()") raised an issue from the fallback helpers added in
commit f0907827a8a9 ("compiler.h: enable builtin overflow checkers and
add fallback code")

Specifically, the helpers for checking whether the results of a
multiplication overflowed (__unsigned_mul_overflow,
__signed_add_overflow) use the division operator when
!COMPILER_HAS_GENERIC_BUILTIN_OVERFLOW.  This is problematic for 64b
operands on 32b hosts.

Also, because the macro is type agnostic, it is very difficult to write
a similarly type generic macro that dispatches to one of:
 * div64_s64
 * div64_u64
 * div_s64
 * div_u64

Raising the minimum supported versions allows us to remove all of the
fallback helpers for !COMPILER_HAS_GENERIC_BUILTIN_OVERFLOW, instead
dispatching the compiler builtins.

arm64 has already raised the minimum supported GCC version to 5.1, do
this for all targets now.  See the link below for the previous
discussion.

Link: https://lore.kernel.org/all/20210909182525.372ee687@canb.auug.org.au/
Link: https://lore.kernel.org/lkml/CAK7LNASs6dvU6D3jL2GG3jW58fXfaj6VNOe55NJnTB8UPuk2pA@mail.gmail.com/
Link: https://github.com/ClangBuiltLinux/linux/issues/1438
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Reported-by: Nathan Chancellor <nathan@kernel.org>
Suggested-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agox86/uaccess: Fix 32-bit __get_user_asm_u64() when CC_HAS_ASM_GOTO_OUTPUT=y
Will Deacon [Mon, 13 Sep 2021 16:35:47 +0000 (17:35 +0100)]
x86/uaccess: Fix 32-bit __get_user_asm_u64() when CC_HAS_ASM_GOTO_OUTPUT=y

Commit 865c50e1d279 ("x86/uaccess: utilize CONFIG_CC_HAS_ASM_GOTO_OUTPUT")
added an optimised version of __get_user_asm() for x86 using 'asm goto'.

Like the non-optimised code, the 32-bit implementation of 64-bit
get_user() expands to a pair of 32-bit accesses.  Unlike the
non-optimised code, the _original_ pointer is incremented to copy the
high word instead of loading through a new pointer explicitly
constructed to point at a 32-bit type.  Consequently, if the pointer
points at a 64-bit type then we end up loading the wrong data for the
upper 32-bits.

This was observed as a mount() failure in Android targeting i686 after
b0cfcdd9b967 ("d_path: make 'prepend()' fill up the buffer exactly on
overflow") because the call to copy_from_kernel_nofault() from
prepend_copy() ends up in __get_kernel_nofault() and casts the source
pointer to a 'u64 __user *'.  An attempt to mount at "/debug_ramdisk"
therefore ends up failing trying to mount "/debumdismdisk".

Use the existing '__gu_ptr' source pointer to unsigned int for 32-bit
__get_user_asm_u64() instead of the original pointer.

Cc: Bill Wendling <morbo@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Reported-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Fixes: 865c50e1d279 ("x86/uaccess: utilize CONFIG_CC_HAS_ASM_GOTO_OUTPUT")
Signed-off-by: Will Deacon <will@kernel.org>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
Tested-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
3 years agoMerge branch 'hns3-fixes'
David S. Miller [Mon, 13 Sep 2021 13:25:37 +0000 (14:25 +0100)]
Merge branch 'hns3-fixes'

Guangbin Huang says:

====================
net: hns3: add some fixes for -net

This series adds some fixes for the HNS3 ethernet driver.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: hns3: fix the timing issue of VF clearing interrupt sources
Jiaran Zhang [Mon, 13 Sep 2021 13:08:25 +0000 (21:08 +0800)]
net: hns3: fix the timing issue of VF clearing interrupt sources

Currently, the VF does not clear the interrupt source immediately after
receiving the interrupt. As a result, if the second interrupt task is
triggered when processing the first interrupt task, clearing the
interrupt source before exiting will clear the interrupt sources of the
two tasks at the same time. As a result, no interrupt is triggered for
the second task. The VF detects the missed message only when the next
interrupt is generated.

Clearing it immediately after executing check_evt_cause ensures that:
1. Even if two interrupt tasks are triggered at the same time, they can
be processed.
2. If the second task is triggered during the processing of the first
task and the interrupt source is not cleared, the interrupt is reported
after vector0 is enabled.

Fixes: b90fcc5bd904 ("net: hns3: add reset handling for VF when doing Core/Global/IMP reset")
Signed-off-by: Jiaran Zhang <zhangjiaran@huawei.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: hns3: fix the exception when query imp info
Jiaran Zhang [Mon, 13 Sep 2021 13:08:24 +0000 (21:08 +0800)]
net: hns3: fix the exception when query imp info

When the command for querying imp info is issued to the firmware,
if the firmware does not support the command, the returned value
of bd num is 0.
Add protection mechanism before alloc memory to prevent apply for
0-length memory.

Fixes: 0b198b0d80ea ("net: hns3: refactor dump m7 info of debugfs")
Signed-off-by: Jiaran Zhang <zhangjiaran@huawei.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: hns3: disable mac in flr process
Yufeng Mo [Mon, 13 Sep 2021 13:08:23 +0000 (21:08 +0800)]
net: hns3: disable mac in flr process

The firmware will not disable mac in flr process. Therefore, the driver
needs to proactively disable mac during flr, which is the same as the
function reset.

Fixes: 35d93a30040c ("net: hns3: adjust the process of PF reset")
Signed-off-by: Yufeng Mo <moyufeng@huawei.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: hns3: change affinity_mask to numa node range
Yufeng Mo [Mon, 13 Sep 2021 13:08:22 +0000 (21:08 +0800)]
net: hns3: change affinity_mask to numa node range

Currently, affinity_mask is set to a single cpu. As a result,
irqbalance becomes invalid in SUBSET or EXACT mode. To solve
this problem, change affinity_mask to numa node range. In this
way, irqbalance can be performed on the cpu of the numa node.

Fixes: 0812545487ec ("net: hns3: add interrupt affinity support for misc interrupt")
Signed-off-by: Yufeng Mo <moyufeng@huawei.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
3 years agonet: hns3: pad the short tunnel frame before sending to hardware
Yufeng Mo [Mon, 13 Sep 2021 13:08:21 +0000 (21:08 +0800)]
net: hns3: pad the short tunnel frame before sending to hardware

The hardware cannot handle short tunnel frames below 65 bytes,
and will cause vlan tag missing problem. So pads packet size to
65 bytes for tunnel frames to fix this bug.

Fixes: 3db084d28dc0("net: hns3: Fix for vxlan tx checksum bug")
Signed-off-by: Yufeng Mo <moyufeng@huawei.com>
Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>