| Age | Commit message (Collapse) | Author |
|---|
|
Completion queues are used for completions of RDQ or SDQ. Each
completion queue is used for one DQ. The first CQs are used for SDQs and
the rest are used for RDQs.
Currently, for each CQE (completion queue element), we check 'sr' value
(send/receive) to know if it is completion of RDQ or SDQ. Actually, we
do not really have to check it, as according to the queue number we know
if it handles completions of Rx or Tx.
Break the tasklet into two - one for Rx (RDQ) and one for Tx (SDQ). Then,
setup the appropriate tasklet for each queue as part of queue
initialization. Use 'sr' value for unlikely case that we get completion
with type that we do not expect. Call WARN_ON_ONCE() only after checking
the value, to avoid calling this method for each completion.
A next patch set will use NAPI to handle events, then we will have a
separate poll method for Rx and Tx. This change is a preparation for
NAPI usage.
Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/50fbc366f8de54cb5dc72a7c4f394333ef71f1d0.1712062203.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
This function will be broken into several functions later. As preparation,
reorder variables to reverse xmas tree.
Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/7170a8f4429ecb5a539b0374c621697778ff8363.1712062203.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The previous patch changed the code to do not handle command interface
from event queue. With this change the wait queue is not used anymore.
Remove it and 'wait_done' variable.
Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/f3af6a5a9dabd97d2920cefe475c6aa57767f504.1712062203.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The device supports two event queues. EQ0 is used for command interface
completion events. EQ1 is used for completion events of RDQ or SDQ.
Currently, for each EQE (event queue element), we check the queue number
and handle accordingly. More than that, for each interrupt we schedule
tasklets for both EQs. This is really ineffective, especially because of
the fact that EQ0 is used only as part of driver init/fini, when EMADs are
not available. There is no point to schedule the tasklet for it and check
each EQE.
A previous patch changed the code to poll command interface for each use of
it. It means that now there is no real reason to use EQ0, as we poll the
command interface.
Initialize only one event queue and use it as EQ1 (this is determined by
queue number). Then, for each interrupt we can schedule the tasklet only
for one queue and we do not have to check the queue number. This
simplifies the code and should improve performance. Note that polling
command interface is ok as we use it only as part of driver init/fini.
Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/23d764f5c032e4c363b98590b746a4b32d2bf900.1712062203.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Currently we use MLXSW_PCI_EQS_COUNT event queues. A next patch will
change the driver to initialize only EQ1, as EQ0 is not required anymore
when we poll command interface.
Rename the macro to MLXSW_PCI_EQS_MAX as later we will not initialize
the maximum supported EQs, this value represents the maximum and a new
macro will be added to represent the actual used queues.
Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/b08df430b62f23ca1aa3aaa257896d2d95aa7691.1712062203.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Command interface is used for configuring and querying FW when EMADs are
not available. During the time that the driver sets up the asynchronous
queues, it polls the command interface for getting completions. Then,
there is a short period when asynchronous queues work, but EMADs are not
available (marked in the code as nopoll = true). During this time, we
send commands via command interface, but we do not poll it, as we can get
an interrupt for the completion. Completions of command interface are
received from HW in EQ0 (event queue 0).
The usage of EQ0 instead of polling is done only 4 times during
initialization and one time during tear down, but it makes an overhead
during lifetime of the driver. For each interrupt, we have to check if
we get events in EQ0 or EQ1 and handle them. This is really ineffective,
especially because of the fact that EQ0 is used only as part of driver
init/fini.
Instead, we can poll command interface for each call of cmd_exec(). It
means that when we send a command via command interface (as EMADs are
not available), we will poll it, regardless of availability of the
asynchronous queues. This will allow us to configure later only EQ1 and
simplify the flow.
Remove 'nopoll' indication and change mlxsw_pci_cmd_exec() to poll till
answer/timeout regardless of queues' state. For now, completions are
handled also by EQ0, but it will be removed in next patch. Additional
cleanups will be added in next patches.
Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/e674c70380ceda953e0e45a77334c5d22e69938f.1712062203.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
This function will be used later only for EQ1. As preparation, reorder
variables to reverse xmas tree and return earlier when it is possible, to
simplify the code.
Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/2412d6c135b2a6aedb4484f5d8baab3aecd7b9ae.1712062203.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The structure 'mlxsw_pci_queue' stores several counters which were consumed
via debugfs. Since commit 9a32562becd9 ("mlxsw: Remove debugfs interface"),
these counters are not used. Remove them. This makes the 'union u' and
'struct eq' redundant. Maintain 'struct cq' as it will be extended later.
Replace increasing 'q->u.eq.ev_other_count' with WARN_ON_ONCE(), as it is
used in an unreasonable case of receiving event in EQ which is not EQ0 or
EQ1. When the queues are initialized, we check number of event queues and
fail with the print "Unsupported number of queues" in case that the driver
tries to initialize more than two queues.
Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/ee9e658800aa0390e08342100bc27daff4c176c0.1712062203.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Currently, as part of mlxsw_pci_cq_tasklet(), we check if any item
was handled, and only in such case we arm doorbell. This is unlikely case,
as we schedule tasklet only for CQs that we get an event for them, which
means that they contain completions to handle. Remove this check, which
is supposed to be true always, and even if it is false, it is not a mistake
to ring the doorbell. We can warn on such case, but it is not really worth
to add a check which will be run for each CQ handling when we do not expect
to reach it and it does not point to logic error that should be handled.
Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/f8efa481bfe7bebb9f93bb803f44ab7da77f53e6.1712062203.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Currently, the structure 'mlxsw_pci_queue_ops' holds a pointer to the
callback function of tasklet. This is used only for EQ and CQ. mlxsw
driver will use NAPI in a following patch set, so CQ will not use tasklet
anymore. As preparation, remove this pointer from the shared operation
structure and setup the tasklet as part of queue initialization.
For now, setup tasklet for EQ and CQ. Later, CQ code will be changed.
Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/a326cae5fc1ad085a1a063c004983de6fe389414.1712062203.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Move mlxsw_pci_cq_{init, fini}() after mlxsw_pci_cq_tasklet() as a next
patch will setup the tasklet as part of initialization.
Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/25196cb5baf5acf6ec1e956203790e018ba8e306.1712062203.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Move mlxsw_pci_eq_{init, fini}() after mlxsw_pci_eq_tasklet() as a next
patch will setup the tasklet as part of initialization.
Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/7ae120a02e1c490084daae7e684a0d40b7cce4e7.1712062203.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Firmware will return 0 on query BOOT/INIT PAGES for non-page supplier
functions (external host PF/VF/SF), so no page is needed to be
allocated for them.
Signed-off-by: Jianbo Liu <jianbol@nvidia.com>
Reviewed-by: Parav Pandit <parav@nvidia.com>
Reviewed-by: Moshe Shemesh <moshe@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://lore.kernel.org/r/20240402133043.56322-12-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Page events are not issued by device on the function if
page_request_disable is set, so no need to create pages EQ.
Signed-off-by: Jianbo Liu <jianbol@nvidia.com>
Reviewed-by: Parav Pandit <parav@nvidia.com>
Reviewed-by: Moshe Shemesh <moshe@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://lore.kernel.org/r/20240402133043.56322-11-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Replace matching on TCP and UDP protocols with new l4_type field which
is parsed by steering for ttc_table. It is enabled by the
outer_l4_type or inner_l4_type bits in nic_rx or port_sel flow table
capabilities and used only if pcc_ifa2 bit in HCA capabilities is set.
Signed-off-by: Jianbo Liu <jianbol@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://lore.kernel.org/r/20240402133043.56322-10-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Add support for 800Gbps speed, link modes of 100Gbps per lane.
Signed-off-by: Gal Pressman <gal@nvidia.com>
Reviewed-by: Cosmin Ratiu <cratiu@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://lore.kernel.org/r/20240402133043.56322-9-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
In the kernel, the preferred types are uX.
Signed-off-by: Gal Pressman <gal@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/20240402133043.56322-8-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Starting from commit
eb9b9fdcafe2 ("net/mlx5e: Introduce extended version for mlx5e_xmit_data")
sinfo is no longer passed as an argument to
mlx5e_xmit_xdp_frame(), the comment is inconsistent.
check_result must be zero when the packet is fragmented.
Signed-off-by: Carolina Jubran <cjubran@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/20240402133043.56322-7-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Resetting stats just before some test/debug case allows us to eliminate
out the impact of previous commands. Useful in particular for the
average latency calculation.
The average_write() callback was unreachable, as "average" is a
read-only file. Extend, rename, and use it for a newly exposed
write-only "reset" file.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Saeed Mahameed <saeedm@nvidia.com>
Reviewed-by: Gal Pressman <gal@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/20240402133043.56322-6-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The fill_strings() callbacks were changed to accept a **data pointer,
and not rely on propagating the index value.
Make a similar change to fill_stats() callbacks to keep the API
consistent.
Signed-off-by: Gal Pressman <gal@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/20240402133043.56322-5-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Use ethtool_sprintf/puts() helper functions which handle the common
pattern of printing a string into the ethtool strings interface and
incrementing the string pointer by ETH_GSTRING_LEN.
Change the fill_strings callback to accept a **data pointer, and remove
the index and return value.
Signed-off-by: Gal Pressman <gal@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/20240402133043.56322-4-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Use ethtool_sprintf/puts() helper functions which handle the common
pattern of printing a string into the ethtool strings interface and
incrementing the string pointer by ETH_GSTRING_LEN.
The int return value in mlx5e_self_test_fill_strings() is not removed as
it is still used to return the number of selftests.
Signed-off-by: Gal Pressman <gal@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/20240402133043.56322-3-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Use ethtool_sprintf/puts() helper functions which handle the common
pattern of printing a string into the ethtool strings interface and
incrementing the string pointer by ETH_GSTRING_LEN.
Signed-off-by: Gal Pressman <gal@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/20240402133043.56322-2-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Historically, tunnel flags like TUNNEL_CSUM or TUNNEL_ERSPAN_OPT
have been defined as __be16. Now all of those 16 bits are occupied
and there's no more free space for new flags.
It can't be simply switched to a bigger container with no
adjustments to the values, since it's an explicit Endian storage,
and on LE systems (__be16)0x0001 equals to
(__be64)0x0001000000000000.
We could probably define new 64-bit flags depending on the
Endianness, i.e. (__be64)0x0001 on BE and (__be64)0x00010000... on
LE, but that would introduce an Endianness dependency and spawn a
ton of Sparse warnings. To mitigate them, all of those places which
were adjusted with this change would be touched anyway, so why not
define stuff properly if there's no choice.
Define IP_TUNNEL_*_BIT counterparts as a bit number instead of the
value already coded and a fistful of <16 <-> bitmap> converters and
helpers. The two flags which have a different bit position are
SIT_ISATAP_BIT and VTI_ISVTI_BIT, as they were defined not as
__cpu_to_be16(), but as (__force __be16), i.e. had different
positions on LE and BE. Now they both have strongly defined places.
Change all __be16 fields which were used to store those flags, to
IP_TUNNEL_DECLARE_FLAGS() -> DECLARE_BITMAP(__IP_TUNNEL_FLAG_NUM) ->
unsigned long[1] for now, and replace all TUNNEL_* occurrences to
their bitmap counterparts. Use the converters in the places which talk
to the userspace, hardware (NFP) or other hosts (GRE header). The rest
must explicitly use the new flags only. This must be done at once,
otherwise there will be too many conversions throughout the code in
the intermediate commits.
Finally, disable the old __be16 flags for use in the kernel code
(except for the two 'irregular' flags mentioned above), to prevent
any accidental (mis)use of them. For the userspace, nothing is
changed, only additions were made.
Most noticeable bloat-o-meter difference (.text):
vmlinux: 307/-1 (306)
gre.ko: 62/0 (62)
ip_gre.ko: 941/-217 (724) [*]
ip_tunnel.ko: 390/-900 (-510) [**]
ip_vti.ko: 138/0 (138)
ip6_gre.ko: 534/-18 (516) [*]
ip6_tunnel.ko: 118/-10 (108)
[*] gre_flags_to_tnl_flags() grew, but still is inlined
[**] ip_tunnel_find() got uninlined, hence such decrease
The average code size increase in non-extreme case is 100-200 bytes
per module, mostly due to sizeof(long) > sizeof(__be16), as
%__IP_TUNNEL_FLAG_NUM is less than %BITS_PER_LONG and the compilers
are able to expand the majority of bitmap_*() calls here into direct
operations on scalars.
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Unlike IPv6 tunnels which use purely-kernel __ip6_tnl_parm structure
to store params inside the kernel, IPv4 tunnel code uses the same
ip_tunnel_parm which is being used to talk with the userspace.
This makes it difficult to alter or add any fields or use a
different format for whatever data.
Define struct ip_tunnel_parm_kern, a 1:1 copy of ip_tunnel_parm for
now, and use it throughout the code. Define the pieces, where the copy
user <-> kernel happens, as standalone functions, and copy the data
there field-by-field, so that the kernel-side structure could be easily
modified later on and the users wouldn't have to care about this.
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
There are, especially with multi-attr arrays, many cases
of needing to iterate all attributes of a specific type
in a netlink message or a nested attribute. Add specific
macros to support that case.
Also convert many instances using this spatch:
@@
iterator nla_for_each_attr;
iterator name nla_for_each_attr_type;
identifier nla;
expression head, len, rem;
expression ATTR;
type T;
identifier x;
@@
-nla_for_each_attr(nla, head, len, rem)
+nla_for_each_attr_type(nla, ATTR, head, len, rem)
{
<... T x; ...>
-if (nla_type(nla) == ATTR) {
...
-}
}
@@
identifier nla;
iterator nla_for_each_nested;
iterator name nla_for_each_nested_type;
expression attr, rem;
expression ATTR;
type T;
identifier x;
@@
-nla_for_each_nested(nla, attr, rem)
+nla_for_each_nested_type(nla, ATTR, attr, rem)
{
<... T x; ...>
-if (nla_type(nla) == ATTR) {
...
-}
}
@@
iterator nla_for_each_attr;
iterator name nla_for_each_attr_type;
identifier nla;
expression head, len, rem;
expression ATTR;
type T;
identifier x;
@@
-nla_for_each_attr(nla, head, len, rem)
+nla_for_each_attr_type(nla, ATTR, head, len, rem)
{
<... T x; ...>
-if (nla_type(nla) != ATTR) continue;
...
}
@@
identifier nla;
iterator nla_for_each_nested;
iterator name nla_for_each_nested_type;
expression attr, rem;
expression ATTR;
type T;
identifier x;
@@
-nla_for_each_nested(nla, attr, rem)
+nla_for_each_nested_type(nla, ATTR, attr, rem)
{
<... T x; ...>
-if (nla_type(nla) != ATTR) continue;
...
}
Although I had to undo one bad change this made, and
I also adjusted some other code for whitespace and to
use direct variable initialization now.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Link: https://lore.kernel.org/r/20240328203144.b5a6c895fb80.I1869b44767379f204998ff44dd239803f39c23e0@changeid
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
When building with 64KB pages, clang points out that xsk->chunk_size
can never be PAGE_SIZE:
drivers/net/ethernet/mellanox/mlx5/core/en/xsk/setup.c:19:22: error: result of comparison of constant 65536 with expression of type 'u16' (aka 'unsigned short') is always false [-Werror,-Wtautological-constant-out-of-range-compare]
if (xsk->chunk_size > PAGE_SIZE ||
~~~~~~~~~~~~~~~ ^ ~~~~~~~~~
In older versions of this code, using PAGE_SIZE was the only
possibility, so this would have never worked on 64KB page kernels,
but the patch apparently did not address this case completely.
As Maxim Mikityanskiy suggested, 64KB chunks are really not all that
useful, so just shut up the warning by adding a cast.
Fixes: 282c0c798f8e ("net/mlx5e: Allow XSK frames smaller than a page")
Link: https://lore.kernel.org/netdev/20211013150232.2942146-1-arnd@kernel.org/
Link: https://lore.kernel.org/lkml/a7b27541-0ebb-4f2d-bd06-270a4d404613@app.fastmail.com/
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Maxim Mikityanskiy <maxtram95@gmail.com>
Reviewed-by: Justin Stitt <justinstitt@google.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://lore.kernel.org/r/20240328143051.1069575-9-arnd@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
clang warns that one error message is too long for its destination buffer:
drivers/net/ethernet/mellanox/mlx5/core/esw/bridge.c:1876:4: error: 'snprintf' will always be truncated; specified size is 80, but format string expands to at least 94 [-Werror,-Wformat-truncation-non-kprintf]
Reword it to be a bit shorter so it always fits.
Fixes: 70f0302b3f20 ("net/mlx5: Bridge, implement mdb offload")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Subbaraya Sundeep <sbhatta@marvell.com>
Link: https://lore.kernel.org/r/20240326223825.4084412-5-arnd@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The mlxbf_gige driver intermittantly encounters a NULL pointer
exception while the system is shutting down via "reboot" command.
The mlxbf_driver will experience an exception right after executing
its shutdown() method. One example of this exception is:
Unable to handle kernel NULL pointer dereference at virtual address 0000000000000070
Mem abort info:
ESR = 0x0000000096000004
EC = 0x25: DABT (current EL), IL = 32 bits
SET = 0, FnV = 0
EA = 0, S1PTW = 0
FSC = 0x04: level 0 translation fault
Data abort info:
ISV = 0, ISS = 0x00000004
CM = 0, WnR = 0
user pgtable: 4k pages, 48-bit VAs, pgdp=000000011d373000
[0000000000000070] pgd=0000000000000000, p4d=0000000000000000
Internal error: Oops: 96000004 [#1] SMP
CPU: 0 PID: 13 Comm: ksoftirqd/0 Tainted: G S OE 5.15.0-bf.6.gef6992a #1
Hardware name: https://www.mellanox.com BlueField SoC/BlueField SoC, BIOS 4.0.2.12669 Apr 21 2023
pstate: 20400009 (nzCv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
pc : mlxbf_gige_handle_tx_complete+0xc8/0x170 [mlxbf_gige]
lr : mlxbf_gige_poll+0x54/0x160 [mlxbf_gige]
sp : ffff8000080d3c10
x29: ffff8000080d3c10 x28: ffffcce72cbb7000 x27: ffff8000080d3d58
x26: ffff0000814e7340 x25: ffff331cd1a05000 x24: ffffcce72c4ea008
x23: ffff0000814e4b40 x22: ffff0000814e4d10 x21: ffff0000814e4128
x20: 0000000000000000 x19: ffff0000814e4a80 x18: ffffffffffffffff
x17: 000000000000001c x16: ffffcce72b4553f4 x15: ffff80008805b8a7
x14: 0000000000000000 x13: 0000000000000030 x12: 0101010101010101
x11: 7f7f7f7f7f7f7f7f x10: c2ac898b17576267 x9 : ffffcce720fa5404
x8 : ffff000080812138 x7 : 0000000000002e9a x6 : 0000000000000080
x5 : ffff00008de3b000 x4 : 0000000000000000 x3 : 0000000000000001
x2 : 0000000000000000 x1 : 0000000000000000 x0 : 0000000000000000
Call trace:
mlxbf_gige_handle_tx_complete+0xc8/0x170 [mlxbf_gige]
mlxbf_gige_poll+0x54/0x160 [mlxbf_gige]
__napi_poll+0x40/0x1c8
net_rx_action+0x314/0x3a0
__do_softirq+0x128/0x334
run_ksoftirqd+0x54/0x6c
smpboot_thread_fn+0x14c/0x190
kthread+0x10c/0x110
ret_from_fork+0x10/0x20
Code: 8b070000 f9000ea0 f95056c0 f86178a1 (b9407002)
---[ end trace 7cc3941aa0d8e6a4 ]---
Kernel panic - not syncing: Oops: Fatal exception in interrupt
Kernel Offset: 0x4ce722520000 from 0xffff800008000000
PHYS_OFFSET: 0x80000000
CPU features: 0x000005c1,a3330e5a
Memory Limit: none
---[ end Kernel panic - not syncing: Oops: Fatal exception in interrupt ]---
During system shutdown, the mlxbf_gige driver's shutdown() is always executed.
However, the driver's stop() method will only execute if networking interface
configuration logic within the Linux distribution has been setup to do so.
If shutdown() executes but stop() does not execute, NAPI remains enabled
and this can lead to an exception if NAPI is scheduled while the hardware
interface has only been partially deinitialized.
The networking interface managed by the mlxbf_gige driver must be properly
stopped during system shutdown so that IFF_UP is cleared, the hardware
interface is put into a clean state, and NAPI is fully deinitialized.
Fixes: f92e1869d74e ("Add Mellanox BlueField Gigabit Ethernet driver")
Signed-off-by: David Thompson <davthompson@nvidia.com>
Link: https://lore.kernel.org/r/20240325210929.25362-1-davthompson@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The mlxbf_gige driver encounters a NULL pointer exception in
mlxbf_gige_open() when kdump is enabled. The sequence to reproduce
the exception is as follows:
a) enable kdump
b) trigger kdump via "echo c > /proc/sysrq-trigger"
c) kdump kernel executes
d) kdump kernel loads mlxbf_gige module
e) the mlxbf_gige module runs its open() as the
the "oob_net0" interface is brought up
f) mlxbf_gige module will experience an exception
during its open(), something like:
Unable to handle kernel NULL pointer dereference at virtual address 0000000000000000
Mem abort info:
ESR = 0x0000000086000004
EC = 0x21: IABT (current EL), IL = 32 bits
SET = 0, FnV = 0
EA = 0, S1PTW = 0
FSC = 0x04: level 0 translation fault
user pgtable: 4k pages, 48-bit VAs, pgdp=00000000e29a4000
[0000000000000000] pgd=0000000000000000, p4d=0000000000000000
Internal error: Oops: 0000000086000004 [#1] SMP
CPU: 0 PID: 812 Comm: NetworkManager Tainted: G OE 5.15.0-1035-bluefield #37-Ubuntu
Hardware name: https://www.mellanox.com BlueField-3 SmartNIC Main Card/BlueField-3 SmartNIC Main Card, BIOS 4.6.0.13024 Jan 19 2024
pstate: 80400009 (Nzcv daif +PAN -UAO -TCO -DIT -SSBS BTYPE=--)
pc : 0x0
lr : __napi_poll+0x40/0x230
sp : ffff800008003e00
x29: ffff800008003e00 x28: 0000000000000000 x27: 00000000ffffffff
x26: ffff000066027238 x25: ffff00007cedec00 x24: ffff800008003ec8
x23: 000000000000012c x22: ffff800008003eb7 x21: 0000000000000000
x20: 0000000000000001 x19: ffff000066027238 x18: 0000000000000000
x17: ffff578fcb450000 x16: ffffa870b083c7c0 x15: 0000aaab010441d0
x14: 0000000000000001 x13: 00726f7272655f65 x12: 6769675f6662786c
x11: 0000000000000000 x10: 0000000000000000 x9 : ffffa870b0842398
x8 : 0000000000000004 x7 : fe5a48b9069706ea x6 : 17fdb11fc84ae0d2
x5 : d94a82549d594f35 x4 : 0000000000000000 x3 : 0000000000400100
x2 : 0000000000000000 x1 : 0000000000000000 x0 : ffff000066027238
Call trace:
0x0
net_rx_action+0x178/0x360
__do_softirq+0x15c/0x428
__irq_exit_rcu+0xac/0xec
irq_exit+0x18/0x2c
handle_domain_irq+0x6c/0xa0
gic_handle_irq+0xec/0x1b0
call_on_irq_stack+0x20/0x2c
do_interrupt_handler+0x5c/0x70
el1_interrupt+0x30/0x50
el1h_64_irq_handler+0x18/0x2c
el1h_64_irq+0x7c/0x80
__setup_irq+0x4c0/0x950
request_threaded_irq+0xf4/0x1bc
mlxbf_gige_request_irqs+0x68/0x110 [mlxbf_gige]
mlxbf_gige_open+0x5c/0x170 [mlxbf_gige]
__dev_open+0x100/0x220
__dev_change_flags+0x16c/0x1f0
dev_change_flags+0x2c/0x70
do_setlink+0x220/0xa40
__rtnl_newlink+0x56c/0x8a0
rtnl_newlink+0x58/0x84
rtnetlink_rcv_msg+0x138/0x3c4
netlink_rcv_skb+0x64/0x130
rtnetlink_rcv+0x20/0x30
netlink_unicast+0x2ec/0x360
netlink_sendmsg+0x278/0x490
__sock_sendmsg+0x5c/0x6c
____sys_sendmsg+0x290/0x2d4
___sys_sendmsg+0x84/0xd0
__sys_sendmsg+0x70/0xd0
__arm64_sys_sendmsg+0x2c/0x40
invoke_syscall+0x78/0x100
el0_svc_common.constprop.0+0x54/0x184
do_el0_svc+0x30/0xac
el0_svc+0x48/0x160
el0t_64_sync_handler+0xa4/0x12c
el0t_64_sync+0x1a4/0x1a8
Code: bad PC value
---[ end trace 7d1c3f3bf9d81885 ]---
Kernel panic - not syncing: Oops: Fatal exception in interrupt
Kernel Offset: 0x2870a7a00000 from 0xffff800008000000
PHYS_OFFSET: 0x80000000
CPU features: 0x0,000005c1,a3332a5a
Memory Limit: none
---[ end Kernel panic - not syncing: Oops: Fatal exception in interrupt ]---
The exception happens because there is a pending RX interrupt before the
call to request_irq(RX IRQ) executes. Then, the RX IRQ handler fires
immediately after this request_irq() completes. The RX IRQ handler runs
"napi_schedule()" before NAPI is fully initialized via "netif_napi_add()"
and "napi_enable()", both which happen later in the open() logic.
The logic in mlxbf_gige_open() must fully initialize NAPI before any calls
to request_irq() execute.
Fixes: f92e1869d74e ("Add Mellanox BlueField Gigabit Ethernet driver")
Signed-off-by: David Thompson <davthompson@nvidia.com>
Reviewed-by: Asmaa Mnebhi <asmaa@nvidia.com>
Link: https://lore.kernel.org/r/20240325183627.7641-1-davthompson@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The mlxbf_gige_open() routine starts the PHY as part of normal
initialization. The mlxbf_gige_open() routine must stop the
PHY during its error paths.
Fixes: f92e1869d74e ("Add Mellanox BlueField Gigabit Ethernet driver")
Signed-off-by: David Thompson <davthompson@nvidia.com>
Reviewed-by: Asmaa Mnebhi <asmaa@nvidia.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm
Pull thermal control updates from Rafael Wysocki:
"These mostly change the thermal core in a few ways allowing thermal
drivers to be simplified, in particular in their removal and failing
probe handling parts that are notoriously prone to errors, and
propagate the changes to several drivers.
Apart from that, support for a new platform is added (Intel Lunar
Lake-M), some bugs are fixed and some code is cleaned up, as usual.
Specifics:
- Store zone trips table and zone operations directly in struct
thermal_zone_device (Rafael Wysocki)
- Fix up flex array initialization during thermal zone device
registration (Nathan Chancellor)
- Rework writable trip points handling in the thermal core and
several drivers (Rafael Wysocki)
- Thermal core code cleanups (Dan Carpenter, Flavio Suligoi)
- Use thermal zone accessor functions in the int340x Intel thermal
driver (Rafael Wysocki)
- Add Lunar Lake-M PCI ID to the int340x Intel thermal driver
(Srinivas Pandruvada)
- Minor fixes for thermal governors (Rafael Wysocki, Di Shen)
- Trip point handling fixes for the iwlwifi wireless driver (Rafael
Wysocki)
- Code cleanups (Rafael J. Wysocki, AngeloGioacchino Del Regno)"
* tag 'thermal-6.9-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm: (29 commits)
thermal: core: remove unnecessary check in trip_point_hyst_store()
thermal: intel: int340x_thermal: Use thermal zone accessor functions
thermal: core: Remove excess empty line from a comment
thermal: int340x: processor_thermal: Add Lunar Lake-M PCI ID
thermal: core: Eliminate writable trip points masks
thermal: of: Set THERMAL_TRIP_FLAG_RW_TEMP directly
thermal: imx: Set THERMAL_TRIP_FLAG_RW_TEMP directly
wifi: iwlwifi: mvm: Set THERMAL_TRIP_FLAG_RW_TEMP directly
mlxsw: core_thermal: Set THERMAL_TRIP_FLAG_RW_TEMP directly
thermal: intel: Set THERMAL_TRIP_FLAG_RW_TEMP directly
thermal: core: Drop the .set_trip_hyst() thermal zone operation
thermal: core: Add flags to struct thermal_trip
thermal: core: Move initial num_trips assignment before memcpy()
thermal: Get rid of CONFIG_THERMAL_WRITABLE_TRIPS
thermal: intel: Adjust ops handling during thermal zone registration
thermal: ACPI: Constify acpi_thermal_zone_ops
thermal: core: Store zone ops in struct thermal_zone_device
thermal: intel: Discard trip tables after zone registration
thermal: ACPI: Discard trips table after zone registration
thermal: core: Store zone trips table in struct thermal_zone_device
...
|
|
For resilient groups, we can reuse the same counter for all the buckets
that share the same nexthop. Keep a reference count per counter, and keep
all these counters in a per-next hop group xarray, which serves as a
NHID->counter cache. If a counter is already present for a given NHID, just
take a reference and use the same counter.
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Link: https://lore.kernel.org/r/cdd00084533fc83ac5917562f54642f008205bf3.1709901020.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
When hw_stats is set on a group, install nexthop counters on members of a
group.
Counter allocation request is moved from nexthop object initialization to
the update code. The previous placement made sense: when the counters are
enabled by dpipe, the counters are installed to all existing nexthops and
all nexthops created from then on get them. For the finer-grained nexthop
group statistics, this is unsuitable. The existing placement was kept for
the IPv4 and IPv6 nexthops.
Resilient group replacement emits a pre_replace notification, and then any
bucket_replace notifications if there were any replacements at all. If the
group is balanced and the nexthop composition of the replaced group didn't
change, there will be no such notifiers. Therefore hook to the pre_replace
notifier and mark all buckets for update, to un/install the counters.
When reporting deltas for resilient groups, use the nexthop ID that we
stored in a previous patch to look up to which nexthop a bucket
contributes.
Co-developed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Petr Machata <petrm@nvidia.com>
Link: https://lore.kernel.org/r/87495a72f187df2e5d491d02729c550d235fcc85.1709901020.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The core interfaces for collecting per-NH statistics are built around
nexthops even for resilient groups. Because mlxsw models each bucket as a
nexthop, the core next hop that a given bucket contributes to needs to be
looked up. In order to be able to match the two up, we need to track
nexthop ID for members of group nexthop objects. For simplicity, do it for
all nexthop objects, not just group members.
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Link: https://lore.kernel.org/r/184ceb6b154e08f5bcf116a705b0fcb01c31895c.1709901020.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The next patch will add the ability to share nexthop counters among
mlxsw nexthops backed by the same core nexthop. To have a place to store
reference count, the counter should be kept in a dedicated structure. In
this patch, introduce the structure together with the related helpers, sans
the refcount, which comes in the next patch.
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Link: https://lore.kernel.org/r/61f23fa4f8c5d7879f68dacd793d8ab7425f33c0.1709901020.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
mlxsw_sp_nexthop_counter_disable() decays to a nop when called on a
disabled counter, but mlxsw_sp_nexthop_counter_enable() can't similarly
be called on an enabled counter. This would be useful in the following
patches. Add the missing condition.
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Link: https://lore.kernel.org/r/0cc9050e196366c1387ab5ee47f1cee8ecde9c86.1709901020.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
For the report_delta-like interface like a previous patch has added for
collection of NH group statistics, it's easiest to read the counter and
have the HW clear it right away. Thus, change mlxsw_sp_flow_counter_get()
to take a bool indicating whether this should be done.
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Link: https://lore.kernel.org/r/6a096ede8ee92d5041e3832242c3bbc137198aba.1709901020.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
In order to be able to diagnose failures in counter allocation, have the
function mlxsw_sp_nexthop_counter_enable() return an error code.
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Link: https://lore.kernel.org/r/e0bb5c0cc6234ade2ade1e92abac991359c3f446.1709901020.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The function mlxsw_sp_nexthop_counter_alloc() doesn't directly allocate
anything, and mlxsw_sp_nexthop_counter_free() doesn't directly free. For
the following patches, we will need names for functions that actually do
those things. Therefore rename to mlxsw_sp_nexthop_counter_enable() and
mlxsw_sp_nexthop_counter_disable() to free up the namespace.
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Link: https://lore.kernel.org/r/f59272958697a718f090f59f892d32beabcd8972.1709901020.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux
Saeed Mahameed says:
====================
Support Multi-PF netdev (Socket Direct)
This series adds support for combining multiple devices (PFs) of the
same port under one netdev instance. Passing traffic through different
devices belonging to different NUMA sockets saves cross-numa traffic and
allows apps running on the same netdev from different numas to still
feel a sense of proximity to the device and achieve improved
performance.
We achieve this by grouping PFs together, and creating the netdev only
once all group members are probed. Symmetrically, we destroy the netdev
once any of the PFs is removed.
The channels are distributed between all devices, a proper configuration
would utilize the correct close numa when working on a certain app/cpu.
We pick one device to be a primary (leader), and it fills a special
role. The other devices (secondaries) are disconnected from the network
in the chip level (set to silent mode). All RX/TX traffic is steered
through the primary to/from the secondaries.
Currently, we limit the support to PFs only, and up to two devices
(sockets).
* tag 'mlx5-socket-direct-v3' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux:
Documentation: networking: Add description for multi-pf netdev
net/mlx5: Enable SD feature
net/mlx5e: Block TLS device offload on combined SD netdev
net/mlx5e: Support per-mdev queue counter
net/mlx5e: Support cross-vhca RSS
net/mlx5e: Let channels be SD-aware
net/mlx5e: Create EN core HW resources for all secondary devices
net/mlx5e: Create single netdev per SD group
net/mlx5: SD, Add debugfs
net/mlx5: SD, Add informative prints in kernel log
net/mlx5: SD, Implement steering for primary and secondaries
net/mlx5: SD, Implement devcom communication and primary election
net/mlx5: SD, Implement basic query and instantiation
net/mlx5: SD, Introduce SD lib
net/mlx5: Add MPIR bit in mcam_access_reg
====================
Link: https://lore.kernel.org/r/20240307084229.500776-1-saeed@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Move RPS related structures and helpers from include/linux/netdevice.h
and include/net/sock.h to a new include file.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://lore.kernel.org/r/20240306160031.874438-18-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Cross-merge networking fixes after downstream PR.
No conflicts.
Adjacent changes:
net/core/page_pool_user.c
0b11b1c5c320 ("netdev: let netlink core handle -EMSGSIZE errors")
429679dcf7d9 ("page_pool: fix netlink dump stop/resume")
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Have an actual mlx5_sd instance in the core device, and fix the getter
accordingly. This allows SD stuff to flow, the feature becomes supported
only here.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Gal Pressman <gal@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
1) Each TX TLS device offloaded context has its own TIS object. Extra work
is needed to get it working in a SD environment, where a stream can move
between different SQs (belonging to different mdevs).
2) Each RX TLS device offloaded context needs a DEK object from the DEK
pool.
Extra work is needed to get it working in a SD environment, as the DEK
pool currently falsely depends on TX cap, and is on the primary device
only.
Disallow this combination for now.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Gal Pressman <gal@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
Each queue counter object counts some events (in hardware) for the RQs
that are attached to it, like events of packet drops due to no receive
WQE (rx_out_of_buffer).
Each RQ can be attached to a queue counter only within the same vhca. To
still cover all RQs with these counters, we create multiple instances,
one per vhca.
The result that's shown to the user is now the sum of all instances.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
Implement driver support for the HW feature that allows RX steering of
one device to target other device's RQs.
In SD multi-pf netdev mode, we set the secondaries into silent mode,
disconnecting them from the network. This feature is then used to steer
traffic from the primary to the secondaries.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Gal Pressman <gal@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
Distribute the channels between the different SD-devices to acheive
local numa node performance on multiple numas.
Each channel works against one specific mdev, creating all datapath
queues against it.
We distribute channels to mdevs in a round-robin policy.
Example for 2 mdevs and 6 channels:
+-------+---------+
| ch ix | mdev ix |
+-------+---------+
| 0 | 0 |
| 1 | 1 |
| 2 | 0 |
| 3 | 1 |
| 4 | 0 |
| 5 | 1 |
+-------+---------+
This round-robin distribution policy is preferred over another suggested
intuitive distribution, in which we first distribute one half of the
channels to mdev #0 and then the second half to mdev #1.
We prefer round-robin for a reason: it is less influenced by changes in
the number of channels. The mapping between channel index and mdev is
fixed, no matter how many channels the user configures. As the channel
stats are persistent to channels closure, changing the mapping every
single time would turn the accumulative stats less representing of the
channel's history.
Per-channel objects should stop using the primary mdev (priv->mdev)
directly, and instead move to using their own channel's mdev.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Gal Pressman <gal@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
Traffic queues will be created on all devices, including the
secondaries. Create the needed core layer resources for them as well.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Gal Pressman <gal@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
|
Integrate the SD library calls into the auxiliary_driver ops in
preparation for creating a single netdev for the multiple PFs belonging
to the same SD group.
SD is still disabled at this stage. It is enabled by a downstream patch
when all needed parts are implemented.
The netdev is created whenever the SD group, with all its participants,
are ready. It is later destroyed whenever any of the participating PFs
drops.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Gal Pressman <gal@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
|
