| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
Currently MLX5E_TC_MAX_INT_PORT_NUM is 8. Usually int port has one
ingress and one egress rules. But sometimes, a temporary rule can be
offloaded as well, eg:
recirc_id(0),in_port(br-phy),eth(src=10:70:fd:87:57:c0,dst=33:33:00:00:00:16),
eth_type(0x86dd),ipv6(frag=no), packets:2, bytes:180, used:0.060s,
actions:enp8s0f0
If one int port device offloads 3 rules, only 2 devices can offload.
Other devices will hit the limit and fail to offload. Actually it is
insufficient for customers. So increase the number to 32.
Signed-off-by: Chris Mi <[email protected]>
Reviewed-by: Roi Dayan <[email protected]>
Signed-off-by: Tariq Toukan <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
f1294617d2f38bd2b9f6cce516b0326858b61182 removed the custom function for
ndo_eth_ioctl and used the standard phy_do_ioctl which calls
phy_mii_ioctl. However since then, this driver was ported to phylink
where it makes more sense to call phylink_mii_ioctl.
Bring back custom function that calls phylink_mii_ioctl.
Signed-off-by: Rosen Penev <[email protected]>
Reviewed-by: Andrew Lunn <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
During initialization with the vnic server, a bitstring is communicated
to the client regarding header info needed during CSO (See "VNIC
Capabilities" in PAPR). Most of the time, to be safe, vnic server
requests header info for CSO. When header info is needed, multiple TX
descriptors are required per skb; This limits the driver to use
send_subcrq_indirect instead of send_subcrq_direct.
Previously, the vnic server request for header info was ignored. This
allowed the use of send_sub_crq_direct. Transmissions were successful
because the bitstring returned by vnic server is broad and over
cautionary. It was observed that mlx backing devices could actually
transmit and handle CSO packets without the vnic server receiving
header info (despite the fact that the bitstring requested it).
There was a trust issue: The bitstring was overcautionary. This extra
precaution (requesting header info when the backing device may not use
it) comes at the cost of performance (using direct vs indirect hcalls
has a 30% delta in small packet RR transaction rate). So it has been
requested that the vnic server team tries to ensure that the bitstring
is more exact. In the meantime, disable CSO when it is possible to use
the skb in the send_subcrq_direct path. In other words, calculate the
checksum before handing the packet to FW when the packet is not
segmented and xmit_more is false.
Since the code path is only possible if the skb is non GSO and xmit_more
is false, the cost of doing checksum in the send_subcrq_direct path is
minimal. Any large segmented skb will have xmit_more set to true more
frequently and it is inexpensive to do checksumming on a small skb.
The worst-case workload would be a 9000 MTU TCP_RR test with close
to MTU sized packets (and TSO off). This allows xmit_more to be false
more frequently and open the code path up to use send_subcrq_direct.
Observing trace data (graph-time = 1) and packet rate with this workload
shows minimal performance degradation:
1. NIC does checksum w headers, safely use send_subcrq_indirect:
- Packet rate: 631k txs
- Trace data:
ibmvnic_xmit = 44344685.87 us / 6234576 hits = AVG 7.11 us
skb_checksum_help = 4.07 us / 2 hits = AVG 2.04 us
^ Notice hits, tracing this just for reassurance
ibmvnic_tx_scrq_flush = 33040649.69 us / 5638441 hits = AVG 5.86 us
send_subcrq_indirect = 37438922.24 us / 6030859 hits = AVG 6.21 us
2. NIC does checksum w/o headers, dangerously use send_subcrq_direct:
- Packet rate: 831k txs
- Trace data:
ibmvnic_xmit = 48940092.29 us / 8187630 hits = AVG 5.98 us
skb_checksum_help = 2.03 us / 1 hits = AVG 2.03
ibmvnic_tx_scrq_flush = 31141879.57 us / 7948960 hits = AVG 3.92 us
send_subcrq_indirect = 8412506.03 us / 728781 hits = AVG 11.54
^ notice hits is much lower b/c send_subcrq_direct was called
^ wasn't traceable
3. driver does checksum, safely use send_subcrq_direct (THIS PATCH):
- Packet rate: 829k txs
- Trace data:
ibmvnic_xmit = 56696077.63 us / 8066168 hits = AVG 7.03 us
skb_checksum_help = 8587456.16 us / 7526072 hits = AVG 1.14 us
ibmvnic_tx_scrq_flush = 30219545.55 us / 7782409 hits = AVG 3.88 us
send_subcrq_indirect = 8638326.44 us / 763693 hits = AVG 11.31 us
When the bitstring ever specifies that CSO does not require headers
(dependent on VIOS vnic server changes), then this patch should be
removed and replaced with one that investigates the bitstring before
using send_subcrq_direct.
Signed-off-by: Nick Child <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
Byte Queue Limits depends on dql_completed being called once per tx
completion round in order to adjust its algorithm appropriately. The
dql->limit value is an approximation of the amount of bytes that the NIC
can consume per irq interval. If this approximation is too high then the
NIC will become over-saturated. Too low and the NIC will starve.
The dql->limit depends on dql->prev-* stats to calculate an optimal
value. If dql_completed() is called more than once per irq handler then
those prev-* values become unreliable (because they are not an accurate
representation of the previous state of the NIC) resulting in a
sub-optimal limit value.
Therefore, move the call to netdev_tx_completed_queue() to the end of
ibmvnic_complete_tx().
When performing 150 sessions of TCP rr (request-response 1 byte packets)
workloads, one could observe:
PREVIOUSLY: - limit and inflight values hovering around 130
- transaction rate of around 750k pps.
NOW: - limit rises and falls in response to inflight (130-900)
- transaction rate of around 1M pps (33% improvement)
Signed-off-by: Nick Child <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
Firmware supports two hcalls to send a sub-crq request:
H_SEND_SUB_CRQ_INDIRECT and H_SEND_SUB_CRQ. The indirect hcall allows
for submission of batched messages while the other hcall is limited to
only one message. This protocol is defined in PAPR section 17.2.3.3.
Previously, the ibmvnic xmit function only used the indirect hcall. This
allowed the driver to batch it's skbs. A single skb can occupy a few
entries per hcall depending on if FW requires skb header information or
not. The FW only needs header information if the packet is segmented.
By this logic, if an skb is not GSO then it can fit in one sub-crq
message and therefore is a candidate for H_SEND_SUB_CRQ.
Batching skb transmission is only useful when there are more packets
coming down the line (ie netdev_xmit_more is true).
As it turns out, H_SEND_SUB_CRQ induces less latency than
H_SEND_SUB_CRQ_INDIRECT. Therefore, use H_SEND_SUB_CRQ where
appropriate.
Small latency gains seen when doing TCP_RR_150 (request/response
workload). Ftrace results (graph-time=1):
Previous:
ibmvnic_xmit = 29618270.83 us / 8860058.0 hits = AVG 3.34
ibmvnic_tx_scrq_flush = 21972231.02 us / 6553972.0 hits = AVG 3.35
Now:
ibmvnic_xmit = 22153350.96 us / 8438942.0 hits = AVG 2.63
ibmvnic_tx_scrq_flush = 15858922.4 us / 6244076.0 hits = AVG 2.54
Signed-off-by: Nick Child <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
send_subcrq_[in]direct() already has a dma memory barrier.
Remove the earlier one.
Signed-off-by: Nick Child <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
Previously when creating the header descriptors, the driver would:
1. allocate a temporary buffer on the stack (in build_hdr_descs_arr)
2. memcpy the header info into the temporary buffer (in build_hdr_data)
3. memcpy the temp buffer into a local variable (in create_hdr_descs)
4. copy the local variable into the return buffer (in create_hdr_descs)
Since, there is no opportunity for errors during this process, the temp
buffer is not needed and work can be done on the return buffer directly.
Repurpose build_hdr_data() to only calculate the header lengths. Rename
it to get_hdr_lens().
Edit create_hdr_descs() to read from the skb directly and copy directly
into the returned useful buffer.
The process now involves less memory and write operations while
also being more readable.
Signed-off-by: Nick Child <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
Use the header length helper functions rather than trying to calculate
it within the driver. There are defined functions for mac and network
headers (skb_mac_header_len and skb_network_header_len) but no such
function exists for the transport header length.
Also, hdr_data was memset during allocation to all 0's so no need to
memset again.
Signed-off-by: Nick Child <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
Previously, the driver would replenish the rx pool if the polling
function consumed less than the budget. The logic being that the driver
did not exhaust its budget so that must mean that the driver is not busy
and has cycles to spare for replenishing the pool.
So pool replenishment happens on every poll which did not consume
the budget. This can very costly during request-response tests.
In fact, an extra ~100pps can be seen in TCP_RR_150 tests when we remove
this conditional. Trace results (ftrace, graph-time=1) for the poll
function are below:
Previous results:
ibmvnic_poll = 64951846.0 us / 4167628.0 hits = AVG 15.58
replenish_rx_pool = 17602846.0 us / 4710437.0 hits = AVG 3.74
Now:
ibmvnic_poll = 57673941.0 us / 4791737.0 hits = AVG 12.04
replenish_rx_pool = 3938171.6 us / 4314.0 hits = AVG 912.88
While the replenish function takes longer, it is hit less frequently
meaning the ibmvnic_poll function, on average, is faster.
Furthermore, this change does not have a negative effect on
performance bandwidth/latency measurements.
Signed-off-by: Nick Child <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
Building fs_enet on powerpc e500 leads to following warning:
CC drivers/net/ethernet/freescale/fs_enet/mac-scc.o
In file included from ./include/linux/build_bug.h:5,
from ./include/linux/container_of.h:5,
from ./include/linux/list.h:5,
from ./include/linux/module.h:12,
from drivers/net/ethernet/freescale/fs_enet/mac-scc.c:15:
drivers/net/ethernet/freescale/fs_enet/mac-scc.c: In function 'allocate_bd':
./include/linux/err.h:28:49: warning: cast to pointer from integer of different size [-Wint-to-pointer-cast]
28 | #define IS_ERR_VALUE(x) unlikely((unsigned long)(void *)(x) >= (unsigned long)-MAX_ERRNO)
| ^
./include/linux/compiler.h:77:45: note: in definition of macro 'unlikely'
77 | # define unlikely(x) __builtin_expect(!!(x), 0)
| ^
drivers/net/ethernet/freescale/fs_enet/mac-scc.c:138:13: note: in expansion of macro 'IS_ERR_VALUE'
138 | if (IS_ERR_VALUE(fep->ring_mem_addr))
| ^~~~~~~~~~~~
This is due to fep->ring_mem_addr not being a pointer but a DMA
address which is 64 bits on that platform while pointers are
32 bits as this is a 32 bits platform with wider physical bus.
However, using fep->ring_mem_addr is just wrong because
cpm_muram_alloc() returns an offset within the muram and not
a physical address directly. So use fpi->dpram_offset instead.
Signed-off-by: Christophe Leroy <[email protected]>
Reviewed-by: Simon Horman <[email protected]>
Link: https://patch.msgid.link/ec67ea3a3bef7e58b8dc959f7c17d405af0d27e4.1723101144.git.christophe.leroy@csgroup.eu
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
Jiri Slaby advises me that the preferred mechanism for declaring
string constants is static char arrays, so use that here.
This mostly reverts
commit 1692b9775e74 ("net: stmmac: xgmac: use #define for string constants")
That commit was a fix for
commit 46eba193d04f ("net: stmmac: xgmac: fix handling of DPP safety error for DMA channels").
The fix being replacing const char * with #defines in order to address
compilation failures observed on GCC 6 through 10.
Compile tested only.
No functional change intended.
Suggested-by: Jiri Slaby <[email protected]>
Link: https://lore.kernel.org/netdev/[email protected]/
Signed-off-by: Simon Horman <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
|
|
Allow of_find_net_device_by_node() to find icssg_prueth ports and make
the individual ports' of_nodes available in sysfs.
Signed-off-by: Matthias Schiffer <[email protected]>
Reviewed-by: MD Danish Anwar <[email protected]>
Reviewed-by: Andrew Lunn <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
Allows simplifying get_strings and avoids manual pointer manipulation.
Signed-off-by: Rosen Penev <[email protected]>
Reviewed-by: Joe Damato <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
Provide declaration of dmae_reg_go_c in header.
This symbol is defined in bnx2x_main.c.
And used in that file and bnx2x_stats.c.
However, Sparse complains that there is no declaration
of the symbol in dmae_reg_go_c nor is the symbol static.
.../bnx2x_main.c:291:11: warning: symbol 'dmae_reg_go_c' was not declared. Should it be static?
Address this by moving the declaration from bnx2x_stats.c to bnx2x_reg.h.
No functional change intended.
Compile tested only.
Signed-off-by: Simon Horman <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
Cross-merge networking fixes after downstream PR.
No conflicts or adjacent changes.
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue
Tony Nguyen says:
====================
Intel Wired LAN Driver Updates 2024-08-07 (ice)
This series contains updates to ice driver only.
Grzegorz adds IRQ synchronization call before performing reset and
prevents writing to hardware when it is resetting.
Mateusz swaps incorrect assignment of FEC statistics.
* '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue:
ice: Fix incorrect assigns of FEC counts
ice: Skip PTP HW writes during PTP reset procedure
ice: Fix reset handler
====================
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
Both ethtool_ops.rxfh_max_context_id and the default value used when
it's not specified are supposed to be exclusive maxima (the former
is documented as such; the latter, U32_MAX, cannot be used as an ID
since it equals ETH_RXFH_CONTEXT_ALLOC), but xa_alloc() expects an
inclusive maximum.
Subtract one from 'limit' to produce an inclusive maximum, and pass
that to xa_alloc().
Increase bnxt's max by one to prevent a (very minor) regression, as
BNXT_MAX_ETH_RSS_CTX is an inclusive max. This is safe since bnxt
is not actually hard-limited; BNXT_MAX_ETH_RSS_CTX is just a
leftover from old driver code that managed context IDs itself.
Rename rxfh_max_context_id to rxfh_max_num_contexts to make its
semantics (hopefully) more obvious.
Fixes: 847a8ab18676 ("net: ethtool: let the core choose RSS context IDs")
Signed-off-by: Edward Cree <[email protected]>
Link: https://patch.msgid.link/5a2d11a599aa5b0cc6141072c01accfb7758650c.1723045898.git.ecree.xilinx@gmail.com
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
PPS was not stopped in `fec_ptp_stop()`, called when
the adapter was removed. Consequentially, you couldn't
safely reload the driver with the PPS signal on.
Fixes: 32cba57ba74b ("net: fec: introduce fec_ptp_stop and use in probe fail path")
Reviewed-by: Fabio Estevam <[email protected]>
Link: https://lore.kernel.org/netdev/CAOMZO5BzcZR8PwKKwBssQq_wAGzVgf1ffwe_nhpQJjviTdxy-w@mail.gmail.com/T/#m01dcb810bfc451a492140f6797ca77443d0cb79f
Signed-off-by: Csókás, Bence <[email protected]>
Reviewed-by: Andrew Lunn <[email protected]>
Reviewed-by: Frank Li <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
Some Wake-on-LAN modes such as WAKE_FILTER may only be supported by the MAC,
while others might be only supported by the PHY. Make sure that the .get_wol()
returns the union of both rather than only that of the PHY if the PHY supports
Wake-on-LAN.
Fixes: 7e400ff35cbe ("net: bcmgenet: Add support for PHY-based Wake-on-LAN")
Signed-off-by: Florian Fainelli <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
dwmac4 was decoding the duplex mode from the GMAC_PHYIF_CONTROL_STATUS
register incorrectly, using GMAC_PHYIF_CTRLSTATUS_LNKMOD_MASK (value 1)
rather than GMAC_PHYIF_CTRLSTATUS_LNKMOD (bit 16). Fix this.
Fixes: 70523e639bf8c ("drivers: net: stmmac: reworking the PCS code.")
Reviewed-by: Andrew Halaney <[email protected]>
Reviewed-by: Andrew Lunn <[email protected]>
Reviewed-by: Serge Semin <[email protected]>
Signed-off-by: Russell King (Oracle) <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
The second tagged commit started sometimes (very rarely, but possible)
throwing WARNs from
net/core/page_pool.c:page_pool_disable_direct_recycling().
Turned out idpf frees interrupt vectors with embedded NAPIs *before*
freeing the queues making page_pools' NAPI pointers lead to freed
memory before these pools are destroyed by libeth.
It's not clear whether there are other accesses to the freed vectors
when destroying the queues, but anyway, we usually free queue/interrupt
vectors only when the queues are destroyed and the NAPIs are guaranteed
to not be referenced anywhere.
Invert the allocation and freeing logic making queue/interrupt vectors
be allocated first and freed last. Vectors don't require queues to be
present, so this is safe. Additionally, this change allows to remove
that useless queue->q_vector pointer cleanup, as vectors are still
valid when freeing the queues (+ both are freed within one function,
so it's not clear why nullify the pointers at all).
Fixes: 1c325aac10a8 ("idpf: configure resources for TX queues")
Fixes: 90912f9f4f2d ("idpf: convert header split mode to libeth + napi_build_skb()")
Reported-by: Michal Kubiak <[email protected]>
Signed-off-by: Alexander Lobakin <[email protected]>
Reviewed-by: Simon Horman <[email protected]>
Tested-by: Krishneil Singh <[email protected]>
Signed-off-by: Tony Nguyen <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
The initialization of vport interrupt consists of two functions:
1) idpf_vport_intr_init() where a generic configuration is done
2) idpf_vport_intr_req_irq() where the irq for each q_vector is
requested.
The first function used to create a base name for each interrupt using
"kasprintf()" call. Unfortunately, although that call allocated memory
for a text buffer, that memory was never released.
Fix this by removing creating the interrupt base name in 1).
Instead, always create a full interrupt name in the function 2), because
there is no need to create a base name separately, considering that the
function 2) is never called out of idpf_vport_intr_init() context.
Fixes: d4d558718266 ("idpf: initialize interrupts and enable vport")
Cc: [email protected] # 6.7
Signed-off-by: Michal Kubiak <[email protected]>
Reviewed-by: Pavan Kumar Linga <[email protected]>
Signed-off-by: Alexander Lobakin <[email protected]>
Reviewed-by: Simon Horman <[email protected]>
Tested-by: Krishneil Singh <[email protected]>
Signed-off-by: Tony Nguyen <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
The second tagged commit introduced a UAF, as it removed restoring
q_vector->vport pointers after reinitializating the structures.
This is due to that all queue allocation functions are performed here
with the new temporary vport structure and those functions rewrite
the backpointers to the vport. Then, this new struct is freed and
the pointers start leading to nowhere.
But generally speaking, the current logic is very fragile. It claims
to be more reliable when the system is low on memory, but in fact, it
consumes two times more memory as at the moment of running this
function, there are two vports allocated with their queues and vectors.
Moreover, it claims to prevent the driver from running into "bad state",
but in fact, any error during the rebuild leaves the old vport in the
partially allocated state.
Finally, if the interface is down when the function is called, it always
allocates a new queue set, but when the user decides to enable the
interface later on, vport_open() allocates them once again, IOW there's
a clear memory leak here.
Just don't allocate a new queue set when performing a reset, that solves
crashes and memory leaks. Readd the old queue number and reopen the
interface on rollback - that solves limbo states when the device is left
disabled and/or without HW queues enabled.
Fixes: 02cbfba1add5 ("idpf: add ethtool callbacks")
Fixes: e4891e4687c8 ("idpf: split &idpf_queue into 4 strictly-typed queue structures")
Signed-off-by: Alexander Lobakin <[email protected]>
Reviewed-by: Simon Horman <[email protected]>
Tested-by: Krishneil Singh <[email protected]>
Signed-off-by: Tony Nguyen <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
Increase size of queue_name buffer from 30 to 31 to accommodate
the largest string written to it. This avoids truncation in
the possibly unlikely case where the string is name is the
maximum size.
Flagged by gcc-14:
.../mvpp2_main.c: In function 'mvpp2_probe':
.../mvpp2_main.c:7636:32: warning: 'snprintf' output may be truncated before the last format character [-Wformat-truncation=]
7636 | "stats-wq-%s%s", netdev_name(priv->port_list[0]->dev),
| ^
.../mvpp2_main.c:7635:9: note: 'snprintf' output between 10 and 31 bytes into a destination of size 30
7635 | snprintf(priv->queue_name, sizeof(priv->queue_name),
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
7636 | "stats-wq-%s%s", netdev_name(priv->port_list[0]->dev),
| ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
7637 | priv->port_count > 1 ? "+" : "");
| ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Introduced by commit 118d6298f6f0 ("net: mvpp2: add ethtool GOP statistics").
I am not flagging this as a bug as I am not aware that it is one.
Compile tested only.
Signed-off-by: Simon Horman <[email protected]>
Reviewed-by: Marcin Wojtas <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
A recent commit has modified the code in __bnxt_reserve_rings() to
set the default RSS indirection table to default only when the number
of RX rings is changing. While this works for newer firmware that
requires RX ring reservations, it causes the regression on older
firmware not requiring RX ring resrvations (BNXT_NEW_RM() returns
false).
With older firmware, RX ring reservations are not required and so
hw_resc->resv_rx_rings is not always set to the proper value. The
comparison:
if (old_rx_rings != bp->hw_resc.resv_rx_rings)
in __bnxt_reserve_rings() may be false even when the RX rings are
changing. This will cause __bnxt_reserve_rings() to skip setting
the default RSS indirection table to default to match the current
number of RX rings. This may later cause bnxt_fill_hw_rss_tbl() to
use an out-of-range index.
We already have bnxt_check_rss_tbl_no_rmgr() to handle exactly this
scenario. We just need to move it up in bnxt_need_reserve_rings()
to be called unconditionally when using older firmware. Without the
fix, if the TX rings are changing, we'll skip the
bnxt_check_rss_tbl_no_rmgr() call and __bnxt_reserve_rings() may also
skip the bnxt_set_dflt_rss_indir_tbl() call for the reason explained
in the last paragraph. Without setting the default RSS indirection
table to default, it causes the regression:
BUG: KASAN: slab-out-of-bounds in __bnxt_hwrm_vnic_set_rss+0xb79/0xe40
Read of size 2 at addr ffff8881c5809618 by task ethtool/31525
Call Trace:
__bnxt_hwrm_vnic_set_rss+0xb79/0xe40
bnxt_hwrm_vnic_rss_cfg_p5+0xf7/0x460
__bnxt_setup_vnic_p5+0x12e/0x270
__bnxt_open_nic+0x2262/0x2f30
bnxt_open_nic+0x5d/0xf0
ethnl_set_channels+0x5d4/0xb30
ethnl_default_set_doit+0x2f1/0x620
Reported-by: Breno Leitao <[email protected]>
Closes: https://lore.kernel.org/netdev/[email protected]/
Fixes: 98ba1d931f61 ("bnxt_en: Fix RSS logic in __bnxt_reserve_rings()")
Reviewed-by: Pavan Chebbi <[email protected]>
Reviewed-by: Kalesh AP <[email protected]>
Reviewed-by: Somnath Kotur <[email protected]>
Signed-off-by: Michael Chan <[email protected]>
Tested-by: Breno Leitao <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
Replace SET_RUNTIME_PM_OPS()/SET SYSTEM_SLEEP_PM_OPS() with their modern
RUNTIME_PM_OPS() and SYSTEM_SLEEP_PM_OPS() alternatives.
The combined usage of pm_ptr() and RUNTIME_PM_OPS/SYSTEM_SLEEP_PM_OPS()
allows the compiler to evaluate if the runtime suspend/resume() functions
are used at build time or are simply dead code.
This allows removing the __maybe_unused notation from the runtime
suspend/resume() functions.
Signed-off-by: Fabio Estevam <[email protected]>
Reviewed-by: Peng Fan <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
Use the `DEFINE_RAW_FLEX()` helper for an on-stack definition of
a flexible structure where the size of the flexible-array member
is known at compile-time, and refactor the rest of the code,
accordingly.
So, with these changes, fix the following warning:
drivers/net/ethernet/fungible/funcore/fun_dev.c:550:43: warning: structure containing a flexible array member is not at the end of another structure [-Wflex-array-member-not-at-end]
Signed-off-by: Gustavo A. R. Silva <[email protected]>
Link: https://patch.msgid.link/ZrDwEugW7DR/FlP5@cute
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
A large tx latency issue was discovered during testing when only QBV was
enabled. The issue occurs because gtxoffset was not set when QBV is
active, it was only set when launch time is active.
The patch "igc: Correct the launchtime offset" only sets gtxoffset when
the launchtime_enable field is set by the user. Enabling launchtime_enable
ultimately sets the register IGC_TXQCTL_QUEUE_MODE_LAUNCHT (referred to as
LaunchT in the SW user manual).
Section 7.5.2.6 of the IGC i225/6 SW User Manual Rev 1.2.4 states:
"The latency between transmission scheduling (launch time) and the
time the packet is transmitted to the network is listed in Table 7-61."
However, the patch misinterprets the phrase "launch time" in that section
by assuming it specifically refers to the LaunchT register, whereas it
actually denotes the generic term for when a packet is released from the
internal buffer to the MAC transmit logic.
This launch time, as per that section, also implicitly refers to the QBV
gate open time, where a packet waits in the buffer for the QBV gate to
open. Therefore, latency applies whenever QBV is in use. TSN features such
as QBU and QAV reuse QBV, making the latency universal to TSN features.
Discussed with i226 HW owner (Shalev, Avi) and we were in agreement that
the term "launch time" used in Section 7.5.2.6 is not clear and can be
easily misinterpreted. Avi will update this section to:
"When TQAVCTRL.TRANSMIT_MODE = TSN, the latency between transmission
scheduling and the time the packet is transmitted to the network is listed
in Table 7-61."
Fix this issue by using igc_tsn_is_tx_mode_in_tsn() as a condition to
write to gtxoffset, aligning with the newly updated SW User Manual.
Tested:
1. Enrol taprio on talker board
base-time 0
cycle-time 1000000
flags 0x2
index 0 cmd S gatemask 0x1 interval1
index 0 cmd S gatemask 0x1 interval2
Note:
interval1 = interval for a 64 bytes packet to go through
interval2 = cycle-time - interval1
2. Take tcpdump on listener board
3. Use udp tai app on talker to send packets to listener
4. Check the timestamp on listener via wireshark
Test Result:
100 Mbps: 113 ~193 ns
1000 Mbps: 52 ~ 84 ns
2500 Mbps: 95 ~ 223 ns
Note that the test result is similar to the patch "igc: Correct the
launchtime offset".
Fixes: 790835fcc0cb ("igc: Correct the launchtime offset")
Signed-off-by: Faizal Rahim <[email protected]>
Reviewed-by: Simon Horman <[email protected]>
Acked-by: Vinicius Costa Gomes <[email protected]>
Tested-by: Mor Bar-Gabay <[email protected]>
Signed-off-by: Tony Nguyen <[email protected]>
|
|
Following the "igc: Fix TX Hang issue when QBV Gate is close" changes,
remaining issues with the reset adapter logic in igc_tsn_offload_apply()
have been observed:
1. The reset adapter logics for i225 and i226 differ, although they should
be the same according to the guidelines in I225/6 HW Design Section
7.5.2.1 on software initialization during tx mode changes.
2. The i225 resets adapter every time, even though tx mode doesn't change.
This occurs solely based on the condition igc_is_device_id_i225() when
calling schedule_work().
3. i226 doesn't reset adapter for tsn->legacy tx mode changes. It only
resets adapter for legacy->tsn tx mode transitions.
4. qbv_count introduced in the patch is actually not needed; in this
context, a non-zero value of qbv_count is used to indicate if tx mode
was unconditionally set to tsn in igc_tsn_enable_offload(). This could
be replaced by checking the existing register
IGC_TQAVCTRL_TRANSMIT_MODE_TSN bit.
This patch resolves all issues and enters schedule_work() to reset the
adapter only when changing tx mode. It also removes reliance on qbv_count.
qbv_count field will be removed in a future patch.
Test ran:
1. Verify reset adapter behaviour in i225/6:
a) Enrol a new GCL
Reset adapter observed (tx mode change legacy->tsn)
b) Enrol a new GCL without deleting qdisc
No reset adapter observed (tx mode remain tsn->tsn)
c) Delete qdisc
Reset adapter observed (tx mode change tsn->legacy)
2. Tested scenario from "igc: Fix TX Hang issue when QBV Gate is closed"
to confirm it remains resolved.
Fixes: 175c241288c0 ("igc: Fix TX Hang issue when QBV Gate is closed")
Signed-off-by: Faizal Rahim <[email protected]>
Reviewed-by: Simon Horman <[email protected]>
Acked-by: Vinicius Costa Gomes <[email protected]>
Tested-by: Mor Bar-Gabay <[email protected]>
Signed-off-by: Tony Nguyen <[email protected]>
|
|
When user issues these cmds:
1. Either a) or b)
a) mqprio with hardware offload disabled
b) taprio with txtime-assist feature enabled
2. etf
3. tc qdisc delete
4. taprio with base time in the past
At step 4, qbv_config_change_errors wrongly increased by 1.
Excerpt from IEEE 802.1Q-2018 8.6.9.3.1:
"If AdminBaseTime specifies a time in the past, and the current schedule
is running, then: Increment ConfigChangeError counter"
qbv_config_change_errors should only increase if base time is in the past
and no taprio is active. In user perspective, taprio was not active when
first triggered at step 4. However, i225/6 reuses qbv for etf, so qbv is
enabled with a dummy schedule at step 2 where it enters
igc_tsn_enable_offload() and qbv_count got incremented to 1. At step 4, it
enters igc_tsn_enable_offload() again, qbv_count is incremented to 2.
Because taprio is running, tc_setup_type is TC_SETUP_QDISC_ETF and
qbv_count > 1, qbv_config_change_errors value got incremented.
This issue happens due to reliance on qbv_count field where a non-zero
value indicates that taprio is running. But qbv_count increases
regardless if taprio is triggered by user or by other tsn feature. It does
not align with qbv_config_change_errors expectation where it is only
concerned with taprio triggered by user.
Fixing this by relocating the qbv_config_change_errors logic to
igc_save_qbv_schedule(), eliminating reliance on qbv_count and its
inaccuracies from i225/6's multiple uses of qbv feature for other TSN
features.
The new function created: igc_tsn_is_taprio_activated_by_user() uses
taprio_offload_enable field to indicate that the current running taprio
was triggered by user, instead of triggered by non-qbv feature like etf.
Fixes: ae4fe4698300 ("igc: Add qbv_config_change_errors counter")
Signed-off-by: Faizal Rahim <[email protected]>
Reviewed-by: Simon Horman <[email protected]>
Acked-by: Vinicius Costa Gomes <[email protected]>
Tested-by: Mor Bar-Gabay <[email protected]>
Signed-off-by: Tony Nguyen <[email protected]>
|
|
Testing uncovered that even when the taprio gate is closed, some packets
still transmit.
According to i225/6 hardware errata [1], traffic might overflow the
planned QBV window. This happens because MAC maintains an internal buffer,
primarily for supporting half duplex retries. Therefore, even when the
gate closes, residual MAC data in the buffer may still transmit.
To mitigate this for i226, reduce the MAC's internal buffer from 192 bytes
to the recommended 88 bytes by modifying the RETX_CTL register value.
This follows guidelines from:
[1] Ethernet Controller I225/I22 Spec Update Rev 2.1 Errata Item 9:
TSN: Packet Transmission Might Cross Qbv Window
[2] I225/6 SW User Manual Rev 1.2.4: Section 8.11.5 Retry Buffer Control
Note that the RETX_CTL register can't be used in TSN mode because half
duplex feature cannot coexist with TSN.
Test Steps:
1. Send taprio cmd to board A:
tc qdisc replace dev enp1s0 parent root handle 100 taprio \
num_tc 4 \
map 3 2 1 0 3 3 3 3 3 3 3 3 3 3 3 3 \
queues 1@0 1@1 1@2 1@3 \
base-time 0 \
sched-entry S 0x07 500000 \
sched-entry S 0x0f 500000 \
flags 0x2 \
txtime-delay 0
Note that for TC3, gate should open for 500us and close for another
500us.
3. Take tcpdump log on Board B.
4. Send udp packets via UDP tai app from Board A to Board B.
5. Analyze tcpdump log via wireshark log on Board B. Ensure that the
total time from the first to the last packet received during one cycle
for TC3 does not exceed 500us.
Fixes: 43546211738e ("igc: Add new device ID's")
Signed-off-by: Faizal Rahim <[email protected]>
Acked-by: Vinicius Costa Gomes <[email protected]>
Tested-by: Mor Bar-Gabay <[email protected]>
Signed-off-by: Tony Nguyen <[email protected]>
|
|
Commit ac21add2540e ("ice: Implement driver functionality to dump fec
statistics") introduces obtaining FEC correctable and uncorrectable
stats per netdev in ICE driver. Unfortunately the assignment of values
to fec_stats structure has been done incorrectly. This commit fixes the
assignments.
Fixes: ac21add2540e ("ice: Implement driver functionality to dump fec statistics")
Reviewed-by: Wojciech Drewek <[email protected]>
Signed-off-by: Mateusz Polchlopek <[email protected]>
Tested-by: Pucha Himasekhar Reddy <[email protected]> (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen <[email protected]>
|
|
Block HW write access for the driver while the device is in reset to
avoid potential race condition and access to the PTP HW in
non-nominal state which could lead to undesired effects
Fixes: 4aad5335969f ("ice: add individual interrupt allocation")
Signed-off-by: Grzegorz Nitka <[email protected]>
Co-developed-by: Karol Kolacinski <[email protected]>
Signed-off-by: Karol Kolacinski <[email protected]>
Signed-off-by: Sergey Temerkhanov <[email protected]>
Reviewed-by: Przemek Kitszel <[email protected]>
Tested-by: Pucha Himasekhar Reddy <[email protected]> (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen <[email protected]>
|
|
Synchronize OICR IRQ when preparing for reset to avoid potential
race conditions between the reset procedure and OICR
Fixes: 4aad5335969f ("ice: add individual interrupt allocation")
Signed-off-by: Grzegorz Nitka <[email protected]>
Signed-off-by: Sergey Temerkhanov <[email protected]>
Reviewed-by: Przemek Kitszel <[email protected]>
Tested-by: Pucha Himasekhar Reddy <[email protected]> (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen <[email protected]>
|
|
Compiling libcxgb_ppm.c results in several sparse warnings:
libcxgb_ppm.c:368:15: warning: incorrect type in assignment (different address spaces)
libcxgb_ppm.c:368:15: expected struct cxgbi_ppm_pool *pools
libcxgb_ppm.c:368:15: got void [noderef] __percpu *_res
libcxgb_ppm.c:374:48: warning: incorrect type in initializer (different address spaces)
libcxgb_ppm.c:374:48: expected void const [noderef] __percpu *__vpp_verify
libcxgb_ppm.c:374:48: got struct cxgbi_ppm_pool *
libcxgb_ppm.c:484:19: warning: incorrect type in assignment (different address spaces)
libcxgb_ppm.c:484:19: expected struct cxgbi_ppm_pool [noderef] __percpu *pool
libcxgb_ppm.c:484:19: got struct cxgbi_ppm_pool *[assigned] pool
libcxgb_ppm.c:511:21: warning: incorrect type in argument 1 (different address spaces)
libcxgb_ppm.c:511:21: expected void [noderef] __percpu *__pdata
libcxgb_ppm.c:511:21: got struct cxgbi_ppm_pool *[assigned] pool
Add __percpu annotation to *pools and *pool percpu pointers and to
ppm_alloc_cpu_pool() function that returns percpu pointer to fix
these warnings.
Compile tested only, but there is no difference in the resulting object file.
Signed-off-by: Uros Bizjak <[email protected]>
Reviewed-by: Simon Horman <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
Take into account return value from reset_control_bulk_assert and
reset_control_bulk_deassert routines in airoha_hw_init().
Signed-off-by: Lorenzo Bianconi <[email protected]>
Reviewed-by: Simon Horman <[email protected]>
Link: https://patch.msgid.link/f49dc04a87653e0155f4fab3e3eb584785c8ad6a.1722699555.git.lorenzo@kernel.org
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
Expose Precision Time Measurement support through related PTP ioctl.
The performance of PTM on ConnectX-7 was evaluated using both real-time
(RTC) and free-running (FRC) clocks under traffic and no traffic
conditions. Tests with phc2sys measured the maximum offset values at a 50Hz
rate, with and without PTM.
Results:
1. No traffic
+-----+--------+--------+
| | No-PTM | PTM |
+-----+--------+--------+
| FRC | 125 ns | <29 ns |
+-----+--------+--------+
| RTC | 248 ns | <34 ns |
+-----+--------+--------+
2. With traffic
+-----+--------+--------+
| | No-PTM | PTM |
+-----+--------+--------+
| FRC | 254 ns | <40 ns |
+-----+--------+--------+
| RTC | 255 ns | <45 ns |
+-----+--------+--------+
Signed-off-by: Rahul Rameshbabu <[email protected]>
Co-developed-by: Carolina Jubran <[email protected]>
Signed-off-by: Carolina Jubran <[email protected]>
Signed-off-by: Tariq Toukan <[email protected]>
Reviewed-by: Wojciech Drewek <[email protected]>
Tested-by: Vadim Fedorenko <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
Since the kernel doesn't support enabling Precision Time Measurement
for an endpoint device, enable the PTM PCI capability in the driver.
Signed-off-by: Carolina Jubran <[email protected]>
Signed-off-by: Tariq Toukan <[email protected]>
Reviewed-by: Wojciech Drewek <[email protected]>
Tested-by: Vadim Fedorenko <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
Make Management Precision Time Measurement (MTPTM) register and Management
Cross Timestamp (MTCTR) register usable in mlx5 driver.
Signed-off-by: Rahul Rameshbabu <[email protected]>
Signed-off-by: Tariq Toukan <[email protected]>
Reviewed-by: Wojciech Drewek <[email protected]>
Tested-by: Vadim Fedorenko <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
GVE driver wrongly relies on netif_carrier_ok() to check the
interface administrative state when resources are being
allocated/deallocated for queue(s). netif_carrier_ok() needs
to be replaced with netif_running() for all such cases.
Administrative state is the result of "ip link set dev <dev>
up/down". It reflects whether the administrator wants to use
the device for traffic and the corresponding resources have
been allocated.
Fixes: 5f08cd3d6423 ("gve: Alloc before freeing when adjusting queues")
Signed-off-by: Praveen Kaligineedi <[email protected]>
Reviewed-by: Shailend Chand <[email protected]>
Reviewed-by: Willem de Bruijn <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
When the length of a packet is under the rx_copybreak threshold, the
buffer is copied into a new skb and sent up the stack. This allows the
dma mapped memory to be recycled back to FW.
Previously, the reuse of the DMA space was handled immediately.
This means that further packet processing has to wait until
h_add_logical_lan finishes for this packet.
Therefore, when reusing a packet, offload the hcall to the replenish
function. As a result, much of the shared logic between the recycle and
replenish functions can be removed.
This change increases TCP_RR packet rate by another 15% (370k to 430k
txns). We can see the ftrace data supports this:
PREV: ibmveth_poll = 8078553.0 us / 190999.0 hits = AVG 42.3 us
NEW: ibmveth_poll = 7632787.0 us / 224060.0 hits = AVG 34.07 us
Signed-off-by: Nick Child <[email protected]>
Reviewed-by: Shannon Nelson <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
When the ibmveth driver processes less than the budget, it must call
napi_complete_done() to release the instance. This function will
return false if the driver should avoid rearming interrupts.
Previously, the driver was ignoring the return code of
napi_complete_done(). As a result, there were unnecessary calls to
enable the veth irq.
Therefore, use the return code napi_complete_done() to determine if
irq rearm is necessary.
Additionally, in the event that new data is received immediately after
rearming interrupts, rather than just rescheduling napi, also jump
back to the poll processing loop since we are already in the poll
function (and know that we did not expense all of budget).
This slight tweak results in a 15% increase in TCP_RR transaction rate
(320k to 370k txns). We can see the ftrace data supports this:
PREV: ibmveth_poll = 8818014.0 us / 182802.0 hits = AVG 48.24
NEW: ibmveth_poll = 8082398.0 us / 191413.0 hits = AVG 42.22
Signed-off-by: Nick Child <[email protected]>
Reviewed-by: Shannon Nelson <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
Link the running gdm port to the qdma controller used to connect with
the CPU. Moreover, load all QDMA controllers available on EN7581 SoC.
Signed-off-by: Lorenzo Bianconi <[email protected]>
Link: https://patch.msgid.link/95b515df34ba4727f7ae5b14a1d0462cceec84ff.1722522582.git.lorenzo@kernel.org
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
This is a preliminary patch to support multi-QDMA controllers.
Signed-off-by: Lorenzo Bianconi <[email protected]>
Link: https://patch.msgid.link/b51cf69c94d8cbc81e0a0b35587f024d01e6d9c0.1722522582.git.lorenzo@kernel.org
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
Map MMIO regions of both qdma controllers available on EN7581 SoC.
Run airoha_hw_cleanup routine for both QDMA controllers available on
EN7581 SoC removing airoha_eth module or in airoha_probe error path.
This is a preliminary patch to support multi-QDMA controllers.
Signed-off-by: Lorenzo Bianconi <[email protected]>
Link: https://patch.msgid.link/a734ae608da14b67ae749b375d880dbbc70868ea.1722522582.git.lorenzo@kernel.org
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
This is a preliminary patch to support multi-QDMA controllers.
Signed-off-by: Lorenzo Bianconi <[email protected]>
Link: https://patch.msgid.link/1e40c3cb973881c0eb3c3c247c78550da62054ab.1722522582.git.lorenzo@kernel.org
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
structures
Move airoha_eth pointer in airoha_qdma structure from
airoha_tx_irq_queue/airoha_queue ones. This is a preliminary patch to
introduce support for multi-QDMA controllers available on EN7581.
Signed-off-by: Lorenzo Bianconi <[email protected]>
Link: https://patch.msgid.link/074565b82fd0ceefe66e186f21133d825dbd48eb.1722522582.git.lorenzo@kernel.org
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
QDMA controllers have independent irq lines, so move irqmask in
airoha_qdma structure. This is a preliminary patch to support multiple
QDMA controllers.
Signed-off-by: Lorenzo Bianconi <[email protected]>
Link: https://patch.msgid.link/1c8a06e8be605278a7b2f3cd8ac06e74bf5ebf2b.1722522582.git.lorenzo@kernel.org
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
QDMA controllers available in EN7581 SoC have independent tx/rx hw queues
so move them in airoha_queues structure.
Signed-off-by: Lorenzo Bianconi <[email protected]>
Link: https://patch.msgid.link/795fc4797bffbf7f0a1351308aa9bf0e65b5126e.1722522582.git.lorenzo@kernel.org
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
Introduce airoha_qdma struct and move qdma IO register mapping in
airoha_qdma. This is a preliminary patch to enable both QDMA controllers
available on EN7581 SoC.
Signed-off-by: Lorenzo Bianconi <[email protected]>
Link: https://patch.msgid.link/7df163bdc72ee29c3d27a0cbf54522ffeeafe53c.1722522582.git.lorenzo@kernel.org
Signed-off-by: Jakub Kicinski <[email protected]>
|