Skip to content

L
linux
Switch branch/tag
  1. 01 Nov, 2022 7 commits
  3. 31 Oct, 2022 18 commits
  5. 30 Oct, 2022 1 commit
  7. 29 Oct, 2022 14 commits
    • Jakub Kicinski's avatar
      Merge tag 'mlx5-updates-2022-10-24' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux · 02a97e02
      Jakub Kicinski authored 
      Saeed Mahameed says:

====================
mlx5-updates-2022-10-24

SW steering updates from Yevgeny Kliteynik:

1) 1st Four patches: small fixes / optimizations for SW steering:

 - Patch 1: Don't abort destroy flow if failed to destroy table - continue
   and free everything else.
 - Patches 2 and 3 deal with fast teardown:
    + Skip sync during fast teardown, as PCI device is not there any more.
    + Check device state when polling CQ - otherwise SW steering keeps polling
      the CQ forever, because nobody is there to flush it.
 - Patch 4: Removing unneeded function argument.

2) Deal with the hiccups that we get during rules insertion/deletion,
which sometimes reach 1/4 of a second. While insertion/deletion rate
improvement was not the focus here, it still is a by-product of removing these
hiccups.

Another by-product is the reduced standard deviation in measuring the duration
of rules insertion/deletion bursts.

In the testing we add K rules (warm-up phase), and then continuously do
insertion/deletion bursts of N rules.
During the test execution, the driver measures hiccups (amount and duration)
and total time for insertion/deletion of a batch of rules.

Here are some numbers, before and after these patches:

+--------------------------------------------+-----------------+----------------+
|                                            |   Create rules  |  Delete rules  |
|                                            +--------+--------+--------+-------+
|                                            | Before |  After | Before | After |
+--------------------------------------------+--------+--------+--------+-------+
| Max hiccup [msec]                          |    253 |     42 |    254 |    68 |
+--------------------------------------------+--------+--------+--------+-------+
| Avg duration of 10K rules add/remove [msec]| 140.07 | 124.32 | 106.99 | 99.51 |
+--------------------------------------------+--------+--------+--------+-------+
| Num of hiccups per 100K rules add/remove   |   7.77 |   7.97 |  12.60 | 11.57 |
+--------------------------------------------+--------+--------+--------+-------+
| Avg hiccup duration [msec]                 |  36.92 |  33.25 |  36.15 | 33.74 |
+--------------------------------------------+--------+--------+--------+-------+

 - Patch 5: Allocate a short array on stack instead of dynamically- it is
   destroyed at the end of the function.
 - Patch 6: Rather than cleaning the corresponding chunk's section of
   ste_arrays on chunk deletion, initialize these areas upon chunk creation.
   Chunk destruction tend to come in large batches (during pool syncing),
   so instead of doing huge memory initialization during pool sync,
   we amortize this by doing small initsializations on chunk creation.
 - Patch 7: In order to simplifies error flow and allows cleaner addition
   of new pools, handle creation/destruction of all the domain's memory pools
   and other memory-related fields in a separate init/uninit functions.
 - Patch 8: During rehash, write each table row immediately instead of waiting
   for the whole table to be ready and writing it all - saves allocations
   of ste_send_info structures and improves performance.
 - Patch 9: Instead of allocating/freeing send info objects dynamically,
   manage them in pool. The number of send info objects doesn't depend on
   number of rules, so after pre-populating the pool with an initial batch of
   send info objects, the pool is not expected to grow.
   This way we save alloc/free during writing STEs to ICM, which by itself can
   sometimes take up to 40msec.
 - Patch 10: Allocate icm_chunks from their own slab allocator, which lowered
   the alloc/free "hiccups" frequency.
 - Patch 11: Similar to patch 9, allocate htbl from its own slab allocator.
 - Patch 12: Lower sync threshold for ICM hot memory - set the threshold for
   sync to 1/4 of the pool instead of 1/2 of the pool. Although we will have
   more syncs, each     sync will be shorter and will help with insertion rate
   stability. Also, notice that the overall number of hiccups wasn't increased
   due to all the other patches.
 - Patch 13: Keep track of hot ICM chunks in an array instead of list.
   After steering sync, we traverse the hot list and finally free all the
   chunks. It appears that traversing a long list takes unusually long time
   due to cache misses on many entries, which causes a big "hiccup" during
   rule insertion. This patch replaces the list with pre-allocated array that
   stores only the bookkeeping information that is needed to later free the
   chunks in its buddy allocator.
 - Patch 14: Remove the unneeded buddy used_list - we don't need to have the
   list of used chunks, we only need the total amount of used memory.

* tag 'mlx5-updates-2022-10-24' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux:
  net/mlx5: DR, Remove the buddy used_list
  net/mlx5: DR, Keep track of hot ICM chunks in an array instead of list
  net/mlx5: DR, Lower sync threshold for ICM hot memory
  net/mlx5: DR, Allocate htbl from its own slab allocator
  net/mlx5: DR, Allocate icm_chunks from their own slab allocator
  net/mlx5: DR, Manage STE send info objects in pool
  net/mlx5: DR, In rehash write the line in the entry immediately
  net/mlx5: DR, Handle domain memory resources init/uninit separately
  net/mlx5: DR, Initialize chunk's ste_arrays at chunk creation
  net/mlx5: DR, For short chains of STEs, avoid allocating ste_arr dynamically
  net/mlx5: DR, Remove unneeded argument from dr_icm_chunk_destroy
  net/mlx5: DR, Check device state when polling CQ
  net/mlx5: DR, Fix the SMFS sync_steering for fast teardown
  net/mlx5: DR, In destroy flow, free resources even if FW command failed
====================

Link: https://lore.kernel.org/r/20221027145643.6618-1-saeed@kernel.orgSigned-off-by: default avatarJakub Kicinski <kuba@kernel.org>
      02a97e02
    • Jakub Kicinski's avatar
      Merge branch 'net-ipa-start-adding-ipa-v5-0-functionality' · eb288cbd
      Jakub Kicinski authored 
      Alex Elder says:

====================
net: ipa: start adding IPA v5.0 functionality

The biggest change for IPA v5.0 is that it supports more than 32
endpoints.  However there are two other unrelated changes:
  - The STATS_TETHERING memory region is not required
  - Filter tables no longer support a "global" filter

Beyond this, refactoring some code makes supporting more than 32
endpoints (in an upcoming series) easier.  So this series includes
a few other changes (not in this order):
  - The maximum endpoint ID in use is determined during config
  - Loops over all endpoints only involve those in use
  - Endpoints IDs and their directions are checked for validity
    differently to simplify comparison against the maximum
====================

Link: https://lore.kernel.org/r/20221027122632.488694-1-elder@linaro.orgSigned-off-by: default avatarJakub Kicinski <kuba@kernel.org>
      eb288cbd
    • Alex Elder's avatar
      net: ipa: record and use the number of defined endpoint IDs · b7aaff0b
      Alex Elder authored 
      Define a new field in the IPA structure that records the maximum
number of entries that will be used in the IPA endpoint array.  Use
that value rather than IPA_ENDPOINT_MAX to determine the end
condition for two loops that iterate over all endpoints.
Signed-off-by: default avatarAlex Elder <elder@linaro.org>
Signed-off-by: default avatarJakub Kicinski <kuba@kernel.org>
      b7aaff0b
    • Alex Elder's avatar
      net: ipa: determine the maximum endpoint ID · 5274c715
      Alex Elder authored 
      Each endpoint ID has an entry in the IPA endpoint array.  But the
size of that array is defined at compile time.  Instead, rename
ipa_endpoint_data_valid() to be ipa_endpoint_max() and have it
return the maximum endpoint ID defined in configuration data.
That function will still validate configuration data.

Zero is returned on error; it's a valid endpoint ID, but we need
more than one, so it can't be the maximum.  The next patch makes use
of the returned maximum value.

Finally, rename the "initialized" mask of endpoints defined by
configuration data to be "defined".
Signed-off-by: default avatarAlex Elder <elder@linaro.org>
Signed-off-by: default avatarJakub Kicinski <kuba@kernel.org>
      5274c715
    • Alex Elder's avatar
      net: ipa: refactor endpoint loops · e359ba89
      Alex Elder authored 
      Change two functions that iterate over all endpoints to use while
loops, using "endpoint_id" as the index variables in both spots.
Signed-off-by: default avatarAlex Elder <elder@linaro.org>
Signed-off-by: default avatarJakub Kicinski <kuba@kernel.org>
      e359ba89
    • Alex Elder's avatar
      net: ipa: more completely check endpoint validity · 2b87d721
      Alex Elder authored 
      Ensure all defined TX endpoints are in the range [0, CONS_PIPES) and
defined RX endpoints are within [PROD_LOWEST, PROD_LOWEST+PROD_PIPES).

Modify the way local variables are used to make the checks easier
to understand.  Check for each endpoint being in valid range in the
loop, and drop the logical-AND check of initialized against
unavailable IDs.
Signed-off-by: default avatarAlex Elder <elder@linaro.org>
Signed-off-by: default avatarJakub Kicinski <kuba@kernel.org>
      2b87d721
    • Alex Elder's avatar
      net: ipa: no more global filtering starting with IPA v5.0 · bd552493
      Alex Elder authored 
      IPA v5.0 eliminates the global filter table entry.  As a result,
there is no need to shift the filtered endpoint bitmap when it is
written to IPA local memory.

Update comments to explain this.  Also delete a redundant block of
comments above the function.
Signed-off-by: default avatarAlex Elder <elder@linaro.org>
Signed-off-by: default avatarJakub Kicinski <kuba@kernel.org>
      bd552493
    • Alex Elder's avatar
      net: ipa: change an IPA v5.0 memory requirement · 5ba5faa2
      Alex Elder authored 
      Don't require IPA v5.0 to have a STATS_TETHERING memory region.
Downstream defines its size to 0, so it apparently is unused.
Signed-off-by: default avatarAlex Elder <elder@linaro.org>
Signed-off-by: default avatarJakub Kicinski <kuba@kernel.org>
      5ba5faa2
    • Alex Elder's avatar
      net: ipa: define IPA v5.0 · 5783c68a
      Alex Elder authored 
      In preparation for adding support for IPA v5.0, define it as an
understood version.
Signed-off-by: default avatarAlex Elder <elder@linaro.org>
Signed-off-by: default avatarJakub Kicinski <kuba@kernel.org>
      5783c68a
    • Willem de Bruijn's avatar
      net/packet: add PACKET_FANOUT_FLAG_IGNORE_OUTGOING · 58ba4263
      Willem de Bruijn authored 
      Extend packet socket option PACKET_IGNORE_OUTGOING to fanout groups.

The socket option sets ptype.ignore_outgoing, which makes
dev_queue_xmit_nit skip the socket.

When the socket joins a fanout group, the option is not reflected in
the struct ptype of the group. dev_queue_xmit_nit only tests the
fanout ptype, so the flag is ignored once a socket joins a
fanout group.

Inheriting the option from a socket would change established behavior.
Different sockets in the group can set different flags, and can also
change them at runtime.

Testing in packet_rcv_fanout defeats the purpose of the original
patch, which is to avoid skb_clone in dev_queue_xmit_nit (esp. for
MSG_ZEROCOPY packets).

Instead, introduce a new fanout group flag with the same behavior.

Tested with https://github.com/wdebruij/kerneltools/blob/master/tests/test_psock_fanout_ignore_outgoing.cSigned-off-by: default avatarWillem de Bruijn <willemb@google.com>
Reviewed-by: default avatarEric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/20221027211014.3581513-1-willemdebruijn.kernel@gmail.comSigned-off-by: default avatarJakub Kicinski <kuba@kernel.org>
      58ba4263
    • Lukasz Czapnik's avatar
      ice: Add additional CSR registers to ETHTOOL_GREGS · 637639cb
      Lukasz Czapnik authored 
      In the event of a Tx hang it can be useful to read a variety of hardware
registers to capture some state about why the transmit queue got stuck.

Extend the ETHTOOL_GREGS dump provided by the ice driver with several CSR
registers that provide such relevant information regarding the hardware Tx
state. This enables capturing relevant data to enable debugging such a Tx
hang.
Signed-off-by: default avatarLukasz Czapnik <lukasz.czapnik@intel.com>
Signed-off-by: default avatarMateusz Palczewski <mateusz.palczewski@intel.com>
Tested-by: Gurucharan <gurucharanx.g@intel.com> (A Contingent worker at Intel)
Link: https://lore.kernel.org/r/20221027104239.1691549-1-jacob.e.keller@intel.comSigned-off-by: default avatarJakub Kicinski <kuba@kernel.org>
      637639cb
    • Jakub Kicinski's avatar
      Merge branch 'clean-up-sfp-register-definitions' · 00643631
      Jakub Kicinski authored 
      Russell King says:

====================
Clean up SFP register definitions

This two-part patch series cleans up the SFP register definitions by
1. converting them from hex to decimal, as all the definitions in the
   documents use decimal, this makes it easier to cross-reference.
2. moving the bit definitions for each register along side their
   register address definition
====================

Link: https://lore.kernel.org/r/Y1qFvaDlLVM1fHdG@shell.armlinux.org.ukSigned-off-by: default avatarJakub Kicinski <kuba@kernel.org>
      00643631
    • Russell King (Oracle)'s avatar
      net: sfp: move field definitions along side register index · d83845d2
      Russell King (Oracle) authored 
      Just as we do for the A2h enum, arrange the A0h enum to have the
field definitions next to their corresponding register index.
Signed-off-by: default avatarRussell King (Oracle) <rmk+kernel@armlinux.org.uk>
Reviewed-by: default avatarAndrew Lunn <andrew@lunn.ch>
Signed-off-by: default avatarJakub Kicinski <kuba@kernel.org>
      d83845d2
    • Russell King (Oracle)'s avatar
      net: sfp: convert register indexes from hex to decimal · 17dd3611
      Russell King (Oracle) authored 
      The register indexes in the standards are in decimal rather than hex,
so lets specify them in decimal in the header file so we can easily
cross-reference without converting between hex and decimal.
Signed-off-by: default avatarRussell King (Oracle) <rmk+kernel@armlinux.org.uk>
Reviewed-by: default avatarAndrew Lunn <andrew@lunn.ch>
Signed-off-by: default avatarJakub Kicinski <kuba@kernel.org>
      17dd3611
GitLab Nexedi Edition | About GitLab | About Nexedi | 沪ICP备2021021310号-2 | 沪ICP备2021021310号-7