David S. Miller [Tue, 14 Jul 2020 00:20:40 +0000 (17:20 -0700)]
Merge branch 'net-simple-kerneldoc-fixes'
Andrew Lunn says:
====================
net simple kerneldoc fixes
This is a collection of simple kerneldoc fixes. They are all low
hanging fruit, were not real understanding of the code was needed.
====================
Andrew Lunn [Sun, 12 Jul 2020 23:15:12 +0000 (01:15 +0200)]
net: socket: Move kerneldoc next to function it documents
Fix the warning "Function parameter or member 'inode' not described in
'__sock_release'' due to the kerneldoc being placed before
__sock_release() not sock_release(), which does not take an inode
parameter.
Alex Elder [Mon, 13 Jul 2020 12:24:18 +0000 (07:24 -0500)]
net: ipa: fix kerneldoc comments
This commit affects comments (and in one case, whitespace) only.
Throughout the IPA code, return statements are documented using
"@Return:", whereas they should use "Return:" instead. Fix these
mistakes.
In function definitions, some parameters are missing their comment
to describe them. And in structure definitions, some fields are
missing their comment to describe them. Add these missing
descriptions.
Some arguments changed name and type along the way, but their
descriptions were not updated (an endpoint pointer is now used in
many places that previously used an endpoint ID). Fix these
incorrect parameter descriptions.
In the description for the ipa_clock structure, one field had a
semicolon instead of a colon in its description. Fix this.
Add a missing function description for ipa_gsi_endpoint_data_empty().
All of these issues were identified when building with "W=1".
====================
Fix bpftool logic of stripping away const/volatile modifiers for all global
variables during BPF skeleton generation. See patch #1 for details on when
existing logic breaks and why it's important. Support special .strip_mods=true
mode in btf_dump__emit_type_decl.
Recent example of when this has caused problems can be found in [0].
tools/bpftool: Strip away modifiers from global variables
Reliably remove all the type modifiers from read-only (.rodata) global
variable definitions, including cases of inner field const modifiers and
arrays of const values.
Also modify one of selftests to ensure that const volatile struct doesn't
prevent user-space from modifying .rodata variable.
One important use case when emitting const/volatile/restrict is undesirable is
BPF skeleton generation of DATASEC layout. These are further memory-mapped and
can be written/read from user-space directly.
For important case of .rodata variables, bpftool strips away first-level
modifiers, to make their use on user-space side simple and not requiring extra
type casts to override compiler complaining about writing to const variables.
This logic works mostly fine, but breaks in some more complicated cases. E.g.:
const volatile int params[10];
Because in BTF it's a chain of ARRAY -> CONST -> VOLATILE -> INT, bpftool
stops at ARRAY and doesn't strip CONST and VOLATILE. In skeleton this variable
will be emitted as is. So when used from user-space, compiler will complain
about writing to const array. This is problematic, as also mentioned in [0].
To solve this for arrays and other non-trivial cases (e.g., inner
const/volatile fields inside the struct), teach btf_dump to strip away any
modifier, when requested. This is done as an extra option on
btf_dump__emit_type_decl() API.
Rationale:
Reduces attack surface on kernel devs opening the links for MITM
as HTTPS traffic is much harder to manipulate.
Deterministic algorithm:
For each file:
If not .svg:
For each line:
If doesn't contain `\bxmlns\b`:
For each link, `\bhttp://[^# \t\r\n]*(?:\w|/)`:
If neither `\bgnu\.org/license`, nor `\bmozilla\.org/MPL\b`:
If both the HTTP and HTTPS versions
return 200 OK and serve the same content:
Replace HTTP with HTTPS.
====================
Steven suggested a way to resolve the appearance of the warning banner
that appears as a result of using trace_printk() in BPF [1].
Applying the patch and testing reveals all works as expected; we
can call bpf_trace_printk() and see the trace messages in
/sys/kernel/debug/tracing/trace_pipe and no banner message appears.
Also add a test prog to verify basic bpf_trace_printk() helper behaviour.
Changes since v2:
- fixed stray newline in bpf_trace_printk(), use sizeof(buf)
rather than #defined value in vsnprintf() (Daniel, patch 1)
- Daniel also pointed out that vsnprintf() returns 0 on error rather
than a negative value; also turns out that a null byte is not
appended if the length of the string written is zero, so to fix
for cases where the string to be traced is zero length we set the
null byte explicitly (Daniel, patch 1)
- switch to using getline() for retrieving lines from trace buffer
to ensure we don't read a portion of the search message in one
read() operation and then fail to find it (Andrii, patch 2)
Changes since v1:
- reorder header inclusion in bpf_trace.c (Steven, patch 1)
- trace zero-length messages also (Andrii, patch 1)
- use a raw spinlock to ensure there are no issues for PREMMPT_RT
kernels when using bpf_trace_printk() within other raw spinlocks
(Steven, patch 1)
- always enable bpf_trace_printk() tracepoint when loading programs
using bpf_trace_printk() as this will ensure that a user disabling
that tracepoint will not prevent tracing output from being logged
(Steven, patch 1)
- use "tp/raw_syscalls/sys_enter" and a usleep(1) to trigger events
in the selftest ensuring test runs faster (Andrii, patch 2)
Rationale:
Reduces attack surface on kernel devs opening the links for MITM
as HTTPS traffic is much harder to manipulate.
Deterministic algorithm:
For each file:
If not .svg:
For each line:
If doesn't contain `\bxmlns\b`:
For each link, `\bhttp://[^# \t\r\n]*(?:\w|/)`:
If neither `\bgnu\.org/license`, nor `\bmozilla\.org/MPL\b`:
If both the HTTP and HTTPS versions
return 200 OK and serve the same content:
Replace HTTP with HTTPS.
Alan Maguire [Mon, 13 Jul 2020 11:52:33 +0000 (12:52 +0100)]
bpf: Use dedicated bpf_trace_printk event instead of trace_printk()
The bpf helper bpf_trace_printk() uses trace_printk() under the hood.
This leads to an alarming warning message originating from trace
buffer allocation which occurs the first time a program using
bpf_trace_printk() is loaded.
We can instead create a trace event for bpf_trace_printk() and enable
it in-kernel when/if we encounter a program using the
bpf_trace_printk() helper. With this approach, trace_printk()
is not used directly and no warning message appears.
This work was started by Steven (see Link) and finished by Alan; added
Steven's Signed-off-by with his permission.
====================
This series introduces new statistics for af_xdp:
1. drops due to rx ring being full
2. drops due to fill ring being empty
3. failures pulling an item from the tx ring
These statistics should assist users debugging and troubleshooting
peformance issues and packet drops.
The statistics are made available though the getsockopt and xsk_diag
interfaces, and the ability to dump these extended statistics is made
available in the xdpsock application via the --extra-stats or -x flag.
A separate patch which will add ss/iproute2 support will follow.
====================
samples: bpf: Add an option for printing extra statistics in xdpsock
Introduce the --extra-stats (or simply -x) flag to the xdpsock application
which prints additional statistics alongside the regular rx and tx
counters. The new statistics printed report error conditions eg. rx ring
full, invalid descriptors, etc.
It can be useful for the user to know the reason behind a dropped packet.
Introduce new counters which track drops on the receive path caused by:
1. rx ring being full
2. fill ring being empty
Also, on the tx path introduce a counter which tracks the number of times
we attempt pull from the tx ring when it is empty.
====================
This patchset adds:
- support to generate BTF ID lists that are resolved during
kernel linking and usable within kernel code with following
macros:
and access it in kernel code via:
extern u32 bpf_skb_output_btf_ids[];
- resolve_btfids tool that scans elf object for .BTF_ids
section and resolves its symbols with BTF ID values
- resolving of bpf_ctx_convert struct and several other
objects with BTF_ID_LIST
v7 changes:
- added more acks [Andrii]
- added some name-conflicting entries and fixed resolve_btfids
to process them properly [Andrii]
- changed bpf_get_task_stack_proto to use BTF_IDS_LIST/BTF_ID
macros [Andrii]
- fixed selftest build for resolve_btfids test
====================
Rationale:
Reduces attack surface on kernel devs opening the links for MITM
as HTTPS traffic is much harder to manipulate.
Deterministic algorithm:
For each file:
If not .svg:
For each line:
If doesn't contain `\bxmlns\b`:
For each link, `\bhttp://[^# \t\r\n]*(?:\w|/)`:
If neither `\bgnu\.org/license`, nor `\bmozilla\.org/MPL\b`:
If both the HTTP and HTTPS versions
return 200 OK and serve the same content:
Replace HTTP with HTTPS.
Rationale:
Reduces attack surface on kernel devs opening the links for MITM
as HTTPS traffic is much harder to manipulate.
Deterministic algorithm:
For each file:
If not .svg:
For each line:
If doesn't contain `\bxmlns\b`:
For each link, `\bhttp://[^# \t\r\n]*(?:\w|/)`:
If neither `\bgnu\.org/license`, nor `\bmozilla\.org/MPL\b`:
If both the HTTP and HTTPS versions
return 200 OK and serve the same content:
Replace HTTP with HTTPS.
kbuild test robot found that addr_to_string() is available only when
DEBUG is defined. And I found that what that function is doing is
what %pM will do. Thus, replace %s with %pM and remove thread-unsafe
addr_to_string() function.
net: bridge: fix undefined br_vlan_can_enter_range in tunnel code
If bridge vlan filtering is not defined we won't have
br_vlan_can_enter_range and thus will get a compile error as was
reported by Stephen and the build bot. So let's define a stub for when
vlan filtering is not used.
Fixes: 94339443686b ("net: bridge: notify on vlan tunnel changes done via the old api") Reported-by: Stephen Rothwell <[email protected]> Signed-off-by: Nikolay Aleksandrov <[email protected]> Signed-off-by: David S. Miller <[email protected]>
Jiri Olsa [Sat, 11 Jul 2020 21:53:29 +0000 (23:53 +0200)]
selftests/bpf: Add test for resolve_btfids
Adding resolve_btfids test under test_progs suite.
It's possible to use btf_ids.h header and its logic in
user space application, so we can add easy test for it.
The test defines BTF_ID_LIST and checks it gets properly
resolved.
For this reason the test_progs binary (and other binaries
that use TRUNNER* macros) is processed with resolve_btfids
tool, which resolves BTF IDs in .BTF_ids section. The BTF
data are taken from btf_data.o object rceated from
progs/btf_data.c.
Jiri Olsa [Sat, 11 Jul 2020 21:53:23 +0000 (23:53 +0200)]
bpf: Add BTF_ID_LIST/BTF_ID/BTF_ID_UNUSED macros
Adding support to generate .BTF_ids section that will hold BTF
ID lists for verifier.
Adding macros that will help to define lists of BTF ID values
placed in .BTF_ids section. They are initially filled with zeros
(during compilation) and resolved later during the linking phase
by resolve_btfids tool.
The wrappers in include/linux/pci-dma-compat.h should go away.
The patch has been generated with the coccinelle script below and has been
hand modified to replace GPF_ with a correct flag.
It has been compile tested.
When memory is allocated in 'sky2_alloc_buffers()', GFP_KERNEL can be used
because some other memory allocations in the same function already use this
flag.
When memory is allocated in 'sky2_probe()', GFP_KERNEL can be used
because another memory allocations in the same function already uses this
flag.
The wrappers in include/linux/pci-dma-compat.h should go away.
The patch has been generated with the coccinelle script below and has been
hand modified to replace GPF_ with a correct flag.
It has been compile tested.
When memory is allocated in 'skge_up()', GFP_KERNEL can be used because
some other memory allocations done a few lines below in 'skge_ring_alloc()'
already use this flag.
====================
Fix MTU warnings for fec/mv886xxx combo
Since changing the MTU of dsa slave interfaces was implemented, the
fec/mv88e6xxx combo has been giving warnings:
[ 2.275925] mv88e6085 0.2:00: nonfatal error -95 setting MTU on port 9
[ 2.284306] eth1: mtu greater than device maximum
[ 2.287759] fec 400d1000.ethernet eth1: error -22 setting MTU to include DSA overhead
This patchset adds support for changing the MTU on mv88e6xxx switches,
which do support jumbo frames. And it modifies the FEC driver to
support its true MTU range, which is larger than the default Ethernet
MTU.
====================
Andrew Lunn [Sat, 11 Jul 2020 20:32:06 +0000 (22:32 +0200)]
net: fec: Set max MTU size to allow the MTU to be changed
The FEC allocates 2K buffers, but looses some of it due to
alignment. It can however support an MTU bigger than the default. This
is particularly interesting when used in combination with Ethernet
switches supporting DSA, which have extra headers. The DSA core will
try to increase the MTU to support these extra headers. If the max
size defaults to that of standard Ethernet we get a warning. By
setting the max to what the driver actually supports, we avoid this
warning.
net: bridge: notify on vlan tunnel changes done via the old api
If someone uses the old vlan API to configure tunnel mappings we'll only
generate the old-style full port notification. That would be a problem
if we are monitoring the new vlan notifications for changes. The patch
resolves the issue by adding vlan notifications to the old tunnel netlink
code. As usual we try to compress the notifications for as many vlans
in a range as possible, thus a vlan tunnel change is considered able
to enter the "current" vlan notification range if:
1. vlan exists
2. it has actually changed (curr_change == true)
3. it passes all standard vlan notification range checks done by
br_vlan_can_enter_range() such as option equality, id continuity etc
Note that vlan tunnel changes (add/del) are considered a part of vlan
options so only RTM_NEWVLAN notification is generated with the relevant
information inside.
Merge tag '5.8-rc4-smb3-fixes' of git://git.samba.org/sfrench/cifs-2.6
Pull cifs fixes from Steve French:
"Four cifs/smb3 fixes: the three for stable fix problems found recently
with change notification including a reference count leak"
* tag '5.8-rc4-smb3-fixes' of git://git.samba.org/sfrench/cifs-2.6:
cifs: update internal module version number
cifs: fix reference leak for tlink
smb3: fix unneeded error message on change notify
cifs: remove the retry in cifs_poxis_lock_set
smb3: fix access denied on change notify request to some servers
Merge tag 'inclusive-terminology' of git://git.kernel.org/pub/scm/linux/kernel/git/djbw/linux
Pull coding style terminology documentation from Dan Williams:
"The discussion has tapered off as well as the incoming ack, review,
and sign-off tags. I did not see a reason to wait for the next merge
window"
* tag 'inclusive-terminology' of git://git.kernel.org/pub/scm/linux/kernel/git/djbw/linux:
CodingStyle: Inclusive Terminology
1) Restore previous behavior of CAP_SYS_ADMIN wrt loading networking
BPF programs, from Maciej Żenczykowski.
2) Fix dropped broadcasts in mac80211 code, from Seevalamuthu
Mariappan.
3) Slay memory leak in nl80211 bss color attribute parsing code, from
Luca Coelho.
4) Get route from skb properly in ip_route_use_hint(), from Miaohe Lin.
5) Don't allow anything other than ARPHRD_ETHER in llc code, from Eric
Dumazet.
6) xsk code dips too deeply into DMA mapping implementation internals.
Add dma_need_sync and use it. From Christoph Hellwig
7) Enforce power-of-2 for BPF ringbuf sizes. From Andrii Nakryiko.
8) Check for disallowed attributes when loading flow dissector BPF
programs. From Lorenz Bauer.
9) Correct packet injection to L3 tunnel devices via AF_PACKET, from
Jason A. Donenfeld.
10) Don't advertise checksum offload on ipa devices that don't support
it. From Alex Elder.
11) Resolve several issues in TCP MD5 signature support. Missing memory
barriers, bogus options emitted when using syncookies, and failure
to allow md5 key changes in established states. All from Eric
Dumazet.
12) Fix interface leak in hsr code, from Taehee Yoo.
13) VF reset fixes in hns3 driver, from Huazhong Tan.
14) Make loopback work again with ipv6 anycast, from David Ahern.
15) Fix TX starvation under high load in fec driver, from Tobias
Waldekranz.
16) MLD2 payload lengths not checked properly in bridge multicast code,
from Linus Lüssing.
17) Packet scheduler code that wants to find the inner protocol
currently only works for one level of VLAN encapsulation. Allow
Q-in-Q situations to work properly here, from Toke
Høiland-Jørgensen.
18) Fix route leak in l2tp, from Xin Long.
19) Resolve conflict between the sk->sk_user_data usage of bpf reuseport
support and various protocols. From Martin KaFai Lau.
20) Fix socket cgroup v2 reference counting in some situations, from
Cong Wang.
21) Cure memory leak in mlx5 connection tracking offload support, from
Eli Britstein.
* git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (146 commits)
mlxsw: pci: Fix use-after-free in case of failed devlink reload
mlxsw: spectrum_router: Remove inappropriate usage of WARN_ON()
net: macb: fix call to pm_runtime in the suspend/resume functions
net: macb: fix macb_suspend() by removing call to netif_carrier_off()
net: macb: fix macb_get/set_wol() when moving to phylink
net: macb: mark device wake capable when "magic-packet" property present
net: macb: fix wakeup test in runtime suspend/resume routines
bnxt_en: fix NULL dereference in case SR-IOV configuration fails
libbpf: Fix libbpf hashmap on (I)LP32 architectures
net/mlx5e: CT: Fix memory leak in cleanup
net/mlx5e: Fix port buffers cell size value
net/mlx5e: Fix 50G per lane indication
net/mlx5e: Fix CPU mapping after function reload to avoid aRFS RX crash
net/mlx5e: Fix VXLAN configuration restore after function reload
net/mlx5e: Fix usage of rcu-protected pointer
net/mxl5e: Verify that rpriv is not NULL
net/mlx5: E-Switch, Fix vlan or qos setting in legacy mode
net/mlx5: Fix eeprom support for SFP module
cgroup: Fix sock_cgroup_data on big-endian.
selftests: bpf: Fix detach from sockmap tests
...
CONFIG_CC_IS_GCC is undefined when Clang is used, which breaks the build
(see our Travis link below).
Clang 8 was chosen as a minimum version for this check because there
were some improvements around __builtin_constant_p in that release. In
reality, MIPS was not even buildable until clang 9 so that check was not
technically necessary. Just remove all compiler checks and just assume
that we have a working compiler.
inet: Remove an unnecessary argument of syn_ack_recalc().
Commit 0c3d79bce48034018e840468ac5a642894a521a3 ("tcp: reduce SYN-ACK
retrans for TCP_DEFER_ACCEPT") introduces syn_ack_recalc() which decides
if a minisock is held and a SYN+ACK is retransmitted or not.
If rskq_defer_accept is not zero in syn_ack_recalc(), max_retries always
has the same value because max_retries is overwritten by rskq_defer_accept
in reqsk_timer_handler().
This commit adds three changes:
- remove redundant non-zero check for rskq_defer_accept in
reqsk_timer_handler().
- remove max_retries from the arguments of syn_ack_recalc() and use
rskq_defer_accept instead.
- rename thresh to max_syn_ack_retries for readability.
mlxsw: pci: Fix use-after-free in case of failed devlink reload
In case devlink reload failed, it is possible to trigger a
use-after-free when querying the kernel for device info via 'devlink dev
info' [1].
This happens because as part of the reload error path the PCI command
interface is de-initialized and its mailboxes are freed. When the
devlink '->info_get()' callback is invoked the device is queried via the
command interface and the freed mailboxes are accessed.
Fix this by initializing the command interface once during probe and not
during every reload.
This is consistent with the other bus used by mlxsw (i.e., 'mlxsw_i2c')
and also allows user space to query the running firmware version (for
example) from the device after a failed reload.
[1]
BUG: KASAN: use-after-free in memcpy include/linux/string.h:406 [inline]
BUG: KASAN: use-after-free in mlxsw_pci_cmd_exec+0x177/0xa60 drivers/net/ethernet/mellanox/mlxsw/pci.c:1675
Write of size 4096 at addr ffff88810ae32000 by task syz-executor.1/2355
Fixes: a9c8336f6544 ("mlxsw: core: Add support for devlink info command") Signed-off-by: Ido Schimmel <[email protected]> Reviewed-by: Jiri Pirko <[email protected]> Signed-off-by: David S. Miller <[email protected]>
David S. Miller [Fri, 10 Jul 2020 21:32:02 +0000 (14:32 -0700)]
Merge branch 'devlink-health'
Moshe Shemesh says:
====================
Add devlink-health support for devlink ports
Implement support for devlink health reporters on per-port basis.
This patchset comes to fix a design issue as some health reporters report
on errors and run recovery on device level while the actual functionality
is on port level. As for the current implemented devlink health reporters
it is relevant only to Tx and Rx reporters of mlx5, which has only one
port, so no real effect on functionality, but this should be fixed before
more drivers will use devlink health reporters.
First part in the series prepares common functions parts for health
reporter implementation. Second introduces required API to devlink-health
and mlx5e ones demonstrate its usage and implement the feature for mlx5
driver.
The per-port reporter functionality is achieved by adding a list of
devlink_health_reporters to devlink_port struct in a manner similar to
existing device infrastructure. This is the only major difference and
it makes possible to fully reuse device reporters operations.
The effect will be seen in conjunction with iproute2 additions and
will affect all devlink health commands. User can distinguish between
device and port reporters by looking at a devlink handle. Port reporters
have a port index at the end of the address and such addresses can be
provided as a parameter in every place where devlink-health accepted it.
These can be obtained from devlink port show command.
For example:
$ devlink health show
pci/0000:00:0a.0:
reporter fw
state healthy error 0 recover 0 auto_dump true
pci/0000:00:0a.0/1:
reporter tx
state healthy error 0 recover 0 grace_period 500 auto_recover true auto_dump true
$ devlink health set pci/0000:00:0a.0/1 reporter tx grace_period 1000 \
auto_recover false auto_dump false
$ devlink health show pci/0000:00:0a.0/1 reporter tx
pci/0000:00:0a.0/1:
reporter tx
state healthy error 0 recover 0 grace_period 1000 auto_recover flase auto_dump false
Note: User can use the same devlink health uAPI commands can get now either
port health reporter or device health reporter.
For example, the recover command:
Before this patchset: devlink health recover DEV reporter REPORTER_NAME
After this patchset: devlink health recover { DEV | DEV/PORT_INDEX } reporter REPORTER_NAME
Changes v1 -> v2:
Fixed functions comment to match parameters list.
Changes v2 -> v3:
Added motivation to cover letter and note on uAPI.
====================
net/mlx5e: Move devlink port register and unregister calls
Register devlink ports upon NIC init. TX and RX health reporters handle
errors which may occur early on at driver initialization. And because
these reporters are to be moved to port context, they require devlink
ports to be already registered.
devlink: Implement devlink health reporters on per-port basis
Add devlink-health reporter support on per-port basis.
The main difference existing devlink-health is that port reporters are
stored in per-devlink_port lists. Upon creation of such health reporter the
reference to a port it belongs to is stored in reporter struct.
Fill the port index attribute in devlink-health response to
allow devlink userspace utility to distinguish between device and port
reporters.
devlink: Rework devlink health reporter destructor
Devlink keeps its own reference to every reporter in a list and inits
refcount to 1 upon reporter's creation. Existing destructor waits to
free the memory indefinitely using msleep() until all references except
devlink's own are put.
Rework this mechanism by moving memory free routine to a separate
function, which is called when the last reporter reference is put.
Besides, it allows to call __devlink_health_reporter_destroy() while
locked on a reporters list mutex in symmetry to
__devlink_health_reporter_create(), which is required in follow-up
patch.
David S. Miller [Fri, 10 Jul 2020 21:29:38 +0000 (14:29 -0700)]
Merge branch 'macb-WOL-fixes'
Nicolas Ferre says:
====================
net: macb: Wake-on-Lan magic packet fixes and GEM handling
Here is a split series to fix WoL magic-packet on the current macb driver. Only
fixes in this one based on current net/master.
Changes in v5:
- Addressed the error code returned by phylink_ethtool_set_wol() as suggested
by Russell.
If PHY handles WoL, MAC doesn't stay in the way.
- Removed Florian's tag on 3/5 because of the above changes.
- Correct the "Fixes" tag on 1/5.
Changes in v4:
- Pure bug fix series for 'net'. GEM addition and MACB update removed: will be
sent later.
Changes in v3:
- Revert some of the v2 changes done in macb_resume(). Now the resume function
supports in-depth re-configuration of the controller in order to deal with
deeper sleep states. Basically as it was before changes introduced by this
series
- Tested for non-regression with our deeper Power Management mode which cuts
power to the controller completely
Changes in v2:
- Add patch 4/7 ("net: macb: fix macb_suspend() by removing call to netif_carrier_off()")
needed for keeping phy state consistent
- Add patch 5/7 ("net: macb: fix call to pm_runtime in the suspend/resume functions") that prevent
putting the macb in runtime pm suspend mode when WoL is used
- Collect review tags on 3 first patches from Florian: Thanks!
- Review of macb_resume() function
- Addition of pm_wakeup_event() in both MACB and GEM WoL IRQ handlers
====================
Nicolas Ferre [Fri, 10 Jul 2020 12:46:45 +0000 (14:46 +0200)]
net: macb: fix call to pm_runtime in the suspend/resume functions
The calls to pm_runtime_force_suspend/resume() functions are only
relevant if the device is not configured to act as a WoL wakeup source.
Add the device_may_wakeup() test before calling them.
Nicolas Ferre [Fri, 10 Jul 2020 12:46:44 +0000 (14:46 +0200)]
net: macb: fix macb_suspend() by removing call to netif_carrier_off()
As we now use the phylink call to phylink_stop() in the non-WoL path,
there is no need for this call to netif_carrier_off() anymore. It can
disturb the underlying phylink FSM.
Nicolas Ferre [Fri, 10 Jul 2020 12:46:43 +0000 (14:46 +0200)]
net: macb: fix macb_get/set_wol() when moving to phylink
Keep previous function goals and integrate phylink actions to them.
phylink_ethtool_get_wol() is not enough to figure out if Ethernet driver
supports Wake-on-Lan.
Initialization of "supported" and "wolopts" members is done in phylink
function, no need to keep them in calling function.
phylink_ethtool_set_wol() return value is considered and determines
if the MAC has to handle WoL or not. The case where the PHY doesn't
implement WoL leads to the MAC configuring it to provide this feature.
Nicolas Ferre [Fri, 10 Jul 2020 12:46:42 +0000 (14:46 +0200)]
net: macb: mark device wake capable when "magic-packet" property present
Change the way the "magic-packet" DT property is handled in the
macb_probe() function, matching DT binding documentation.
Now we mark the device as "wakeup capable" instead of calling the
device_init_wakeup() function that would enable the wakeup source.
For Ethernet WoL, enabling the wakeup_source is done by
using ethtool and associated macb_set_wol() function that
already calls device_set_wakeup_enable() for this purpose.
That would reduce power consumption by cutting more clocks if
"magic-packet" property is set but WoL is not configured by ethtool.
Nicolas Ferre [Fri, 10 Jul 2020 12:46:41 +0000 (14:46 +0200)]
net: macb: fix wakeup test in runtime suspend/resume routines
Use the proper struct device pointer to check if the wakeup flag
and wakeup source are positioned.
Use the one passed by function call which is equivalent to
&bp->dev->dev.parent.
It's preventing the trigger of a spurious interrupt in case the
Wake-on-Lan feature is used.
bnxt_en: fix NULL dereference in case SR-IOV configuration fails
we need to set 'active_vfs' back to 0, if something goes wrong during the
allocation of SR-IOV resources: otherwise, further VF configurations will
wrongly assume that bp->pf.vf[x] are valid memory locations, and commands
like the ones in the following sequence:
# echo 2 >/sys/bus/pci/devices/${ADDR}/sriov_numvfs
# ip link set dev ens1f0np0 up
# ip link set dev ens1f0np0 vf 0 trust on
1) Restore CT state from lookup in zone instead of tupleid
On a miss, Use this zone + 5 tuple taken from the skb, to lookup the CT
entry and restore it, instead of the driver allocated tuple id.
This improves flow insertion rate by avoiding the allocation of a header
rewrite context to maintain the tupleid.
2) Re-use modify header HW objects for identical modify actions.
3) Expand tunnel register mappings
Reg_c1 is 32 bits wide. Before this patchset, 24 bit were allocated
for the tuple_id, 6 bits for tunnel mapping and 2 bits for tunnel
options mappings.
Restoring the ct state from zone lookup instead of tuple id requires
reg_c1 to store 8 bits mapping the ct zone, leaving 24 bits for tunnel
mappings.
Expand tunnel and tunnel options register mappings to 12 bit each.
4) Trivial cleanup and fixes.
====================
David S. Miller [Fri, 10 Jul 2020 21:02:01 +0000 (14:02 -0700)]
Merge tag 'mlx5-fixes-2020-07-02' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux
Saeed Mahameed says:
====================
mlx5 fixes 2020-07-02
This series introduces some fixes to mlx5 driver.
V1->v2:
- Drop "ip -s" patch and mirred device hold reference patch.
- Will revise them in a later submission.
Please pull and let me know if there is any problem.
For -stable v5.2
('net/mlx5: Fix eeprom support for SFP module')
For -stable v5.4
('net/mlx5e: Fix 50G per lane indication')
For -stable v5.5
('net/mlx5e: Fix CPU mapping after function reload to avoid aRFS RX crash')
('net/mlx5e: Fix VXLAN configuration restore after function reload')
For -stable v5.7
('net/mlx5e: CT: Fix memory leak in cleanup')
====================
====================
udp_tunnel: add NIC RX port offload infrastructure
Kernel has a facility to notify drivers about the UDP tunnel ports
so that devices can recognize tunneled packets. This is important
mostly for RX - devices which don't support CHECKSUM_COMPLETE can
report checksums of inner packets, and compute RSS over inner headers.
Some drivers also match the UDP tunnel ports also for TX, although
doing so may lead to false positives and negatives.
Unfortunately the user experience when trying to take adavantage
of these facilities is suboptimal. First of all there is no way
for users to check which ports are offloaded. Many drivers resort
to printing messages to aid debugging, other use debugfs. Even worse
the availability of the RX features (NETIF_F_RX_UDP_TUNNEL_PORT)
is established purely on the basis of the driver having the ndos
installed. For most drivers, however, the ability to perform offloads
is contingent on device capabilities (driver support multiple device
and firmware versions). Unless driver resorts to hackish clearing
of features set incorrectly by the core - users are left guessing
whether their device really supports UDP tunnel port offload or not.
There is currently no way to indicate or configure whether RX
features include just the checksum offload or checksum and using
inner headers for RSS. Many drivers default to not using inner
headers for RSS because most implementations populate the source
port with entropy from the inner headers. This, however, is not
always the case, for example certain switches are only able to
use a fixed source port during encapsulation.
We have also seen many driver authors get the intricacies of UDP
tunnel port offloads wrong. Most commonly the drivers forget to
perform reference counting, or take sleeping locks in the callbacks.
This work tries to improve the situation by pulling the UDP tunnel
port table maintenance out of the drivers. It turns out that almost
all drivers maintain a fixed size table of ports (in most cases one
per tunnel type), so we can take care of all the refcounting in the
core, and let the driver specify if they need to sleep in the
callbacks or not. The new common implementation will also support
replacing ports - when a port is removed from a full table it will
try to find a previously missing port to take its place.
This patch only implements the core functionality along with a few
drivers I was hoping to test manually [1] along with a test based
on a netdevsim implementation. Following patches will convert all
the drivers. Once that's complete we can remove the ndos, and rely
directly on the new infrastrucutre.
Then after RSS (RXFH) is converted to netlink we can add the ability
to configure the use of inner RSS headers for UDP tunnels.
[1] Unfortunately I wasn't able to, turns out 2 of the devices
I had access to were older generation or had old FW, and they
did not actually support UDP tunnel port notifications (see
the second paragraph). The thrid device appears to program
the UDP ports correctly but it generates bad UDP checksums with
or without these patches. Long story short - I'd appreciate
reviews and testing here..
v4:
- better build fix (hopefully this one does it..)
v3:
- fix build issue;
- improve bnxt changes.
====================
Jakub Kicinski [Fri, 10 Jul 2020 00:42:51 +0000 (17:42 -0700)]
ixgbe: convert to new udp_tunnel_nic infra
Make use of new common udp_tunnel_nic infra. ixgbe supports
IPv4 only, and only single VxLAN and Geneve ports (one each).
v2:
- split out the RXCSUM feature handling to separate change;
- declare structs separately;
- use ti.type instead of assuming table 0 is VxLAN;
- move setting netdev->udp_tunnel_nic_info to its own switch.
Jakub Kicinski [Fri, 10 Jul 2020 00:42:50 +0000 (17:42 -0700)]
ixgbe: don't clear UDP tunnel ports when RXCSUM is disabled
It appears the clearing of UDP tunnel ports when RXCSUM
is disabled is unnecessary. Driver will not pay attention
to checksum bits if RXCSUM is not set, so we can let
the hardware parse the packets.
Note that the UDP tunnel port NDO handlers don't pay attention
to the state of RXCSUM, so the ports could had been re-programmed,
anyway.
Jakub Kicinski [Fri, 10 Jul 2020 00:42:47 +0000 (17:42 -0700)]
ethtool: add tunnel info interface
Add an interface to report offloaded UDP ports via ethtool netlink.
Now that core takes care of tracking which UDP tunnel ports the NICs
are aware of we can quite easily export this information out to
user space.
The responsibility of writing the netlink dumps is split between
ethtool code and udp_tunnel_nic.c - since udp_tunnel module may
not always be loaded, yet we should always report the capabilities
of the NIC.
$ ethtool --show-tunnels eth0
Tunnel information for eth0:
UDP port table 0:
Size: 4
Types: vxlan
No entries
UDP port table 1:
Size: 4
Types: geneve, vxlan-gpe
Entries (1):
port 1230, vxlan-gpe
v4:
- back to v2, build fix is now directly in udp_tunnel.h
v3:
- don't compile ETHTOOL_MSG_TUNNEL_INFO_GET in if CONFIG_INET
not set.
v2:
- fix string set count,
- reorder enums in the uAPI,
- fix type of ETHTOOL_A_TUNNEL_UDP_TABLE_TYPES to bitset
in docs and comments.
Jakub Kicinski [Fri, 10 Jul 2020 00:42:46 +0000 (17:42 -0700)]
udp_tunnel: add central NIC RX port offload infrastructure
Cater to devices which:
(a) may want to sleep in the callbacks;
(b) only have IPv4 support;
(c) need all the programming to happen while the netdev is up.
Drivers attach UDP tunnel offload info struct to their netdevs,
where they declare how many UDP ports of various tunnel types
they support. Core takes care of tracking which ports to offload.
Use a fixed-size array since this matches what almost all drivers
do, and avoids a complexity and uncertainty around memory allocations
in an atomic context.
Make sure that tunnel drivers don't try to replay the ports when
new NIC netdev is registered. Automatic replays would mess up
reference counting, and will be removed completely once all drivers
are converted.
v4:
- use a #define NULL to avoid build issues with CONFIG_INET=n.
Jakub Kicinski [Fri, 10 Jul 2020 00:42:45 +0000 (17:42 -0700)]
udp_tunnel: re-number the offload tunnel types
Make it possible to use tunnel types as flags more easily.
There doesn't appear to be any user using the type as an
array index, so this should make no difference.
Jakub Kicinski [Fri, 10 Jul 2020 00:42:44 +0000 (17:42 -0700)]
debugfs: make sure we can remove u32_array files cleanly
debugfs_create_u32_array() allocates a small structure to wrap
the data and size information about the array. If users ever
try to remove the file this leads to a leak since nothing ever
frees this wrapper.
That said there are no upstream users of debugfs_create_u32_array()
that'd remove a u32 array file (we only have one u32 array user in
CMA), so there is no real bug here.
Make callers pass a wrapper they allocated. This way the lifetime
management of the wrapper is on the caller, and we can avoid the
potential leak in debugfs.
Merge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma
Pull rdma fixes from Jason Gunthorpe:
"Small update, a few more merge window bugs and normal driver bug
fixes:
- Two merge window regressions in mlx5: a error path bug found by
syzkaller and some lost code during a rework preventing ipoib from
working in some configurations
- Silence clang compilation warning in OPA related code
- Fix a long standing race condition in ib_nl for ACM
- Resolve when the HFI1 is shutdown"
* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma:
RDMA/mlx5: Set PD pointers for the error flow unwind
IB/mlx5: Fix 50G per lane indication
RDMA/siw: Fix reporting vendor_part_id
IB/sa: Resolv use-after-free in ib_nl_make_request()
IB/hfi1: Do not destroy link_wq when the device is shut down
IB/hfi1: Do not destroy hfi1_wq when the device is shut down
RDMA/mlx5: Fix legacy IPoIB QP initialization
IB/hfi1: Add explicit cast OPA_MTU_8192 to 'enum ib_mtu'
Merge tag 'linux-kselftest-fixes-5.8-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/shuah/linux-kselftest
Pull kselftest fixes from Shuah Khan:
"TPM2 test changes to run on python3 and kselftest framework fix to
incorrect return type"
* tag 'linux-kselftest-fixes-5.8-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/shuah/linux-kselftest:
kselftest: ksft_test_num return type should be unsigned
selftests: tpm: upgrade TPM2 tests from Python 2 to Python 3
Merge tag 'io_uring-5.8-2020-07-10' of git://git.kernel.dk/linux-block
Pull io_uring fixes from Jens Axboe:
- Fix memleak for error path in registered files (Yang)
- Export CQ overflow state in flags, necessary to fix a case where
liburing doesn't know if it needs to enter the kernel (Xiaoguang)
- Fix for a regression in when user memory is accounted freed, causing
issues with back-to-back ring exit + init if the ulimit -l setting is
very tight.
* tag 'io_uring-5.8-2020-07-10' of git://git.kernel.dk/linux-block:
io_uring: account user memory freed when exit has been queued
io_uring: fix memleak in io_sqe_files_register()
io_uring: fix memleak in __io_sqe_files_update()
io_uring: export cq overflow status to userspace
Merge tag 'block-5.8-2020-07-10' of git://git.kernel.dk/linux-block
Pull block fixes from Jens Axboe:
- Fix for inflight accounting, which affects only dm (Ming)
- Fix documentation error for bfq (Yufen)
- Fix memory leak for nbd (Zheng)
* tag 'block-5.8-2020-07-10' of git://git.kernel.dk/linux-block:
nbd: Fix memory leak in nbd_add_socket
blk-mq: consider non-idle request as "inflight" in blk_mq_rq_inflight()
docs: block: update and fix tiny error for bfq
Merge tag 'cleanup-kernel_read_write' of git://git.infradead.org/users/hch/misc
Pull in-kernel read and write op cleanups from Christoph Hellwig:
"Cleanup in-kernel read and write operations
Reshuffle the (__)kernel_read and (__)kernel_write helpers, and ensure
all users of in-kernel file I/O use them if they don't use iov_iter
based methods already.
The new WARN_ONs in combination with syzcaller already found a missing
input validation in 9p. The fix should be on your way through the
maintainer ASAP".
[ This is prep-work for the real changes coming 5.9 ]
* tag 'cleanup-kernel_read_write' of git://git.infradead.org/users/hch/misc:
fs: remove __vfs_read
fs: implement kernel_read using __kernel_read
integrity/ima: switch to using __kernel_read
fs: add a __kernel_read helper
fs: remove __vfs_write
fs: implement kernel_write using __kernel_write
fs: check FMODE_WRITE in __kernel_write
fs: unexport __kernel_write
bpfilter: switch to kernel_write
autofs: switch to kernel_write
cachefiles: switch to kernel_write
projects / linux.git / log