Public bug reported:
Summary:
Customer is reporting performance degradation from Bluefield Focal 5.4.0-1087
to 5.4.0-1088
They run iperf3 based performance test, and see performance drop around 10-15%.
Test scenarios:
===
1. Baseline: BF - BFB 3.9.8 (OFED Driver 5.8, FW 35.4
Public bug reported:
Summary:
ip6tables --set-xmark failed on 1057.59
A degradation
Working version: 1053.55
bf-bundle-2.9.1-13_24.11_ubuntu-22.04_prod.bfb has kernel
5.15.0-1053.55.24.g9cc17fe-bluefield
root@r-qa-hbn00-host00-00:mgmt:/home/ubuntu# /usr/sbin/ip6tables -t
** Tags removed: verification-needed-jammy-linux-bluefield
--
You received this bug notification because you are a member of Kernel
Packages, which is subscribed to linux-bluefield in Ubuntu.
https://bugs.launchpad.net/bugs/2077887
Title:
Backport VFIO P2P patch for BlueField 5.15
Status in l
can we skip the verification?
this needs a very special setup to verify.
--
You received this bug notification because you are a member of Kernel
Packages, which is subscribed to linux-bluefield in Ubuntu.
https://bugs.launchpad.net/bugs/2077887
Title:
Backport VFIO P2P patch for BlueField 5.1
bisect shows
# first bad commit: [9c901808b3affa5b21bb146148ae2c39dfd732ee] drivers:
core: synchronize really_probe() and dev_uevent()
--
You received this bug notification because you are a member of Kernel
Packages, which is subscribed to linux-bluefield in Ubuntu.
https://bugs.launchpad.net/b
** Description changed:
* introduction
Found a regression on 5.15.0-1053, which worked OK on 5.15.0-1050
At DPU, when using the 5.15.0-1053-bluefield kernel, when the user-space
process uses the OFED driver to create 2000 of SF devices in a batch mode.
At host side, the ubuntu kernel w
Public bug reported:
* introduction
Found a regression on 5.15.0-1053, which worked OK on 5.15.0-1050
At DPU, when using the 5.15.0-1053-bluefield kernel, when the user-space
process uses the OFED driver to create 2000 of SF devices in a batch mode.
At host side, the ubuntu kernel will prevent t
also backport the two patches
---
net/sched: act_ct: Take per-cb reference to tcf_ct_flow_table
net: sched: call tcf_ct_params_free to free params in tcf_ct_init
--
You received this bug notification because you are a member of Kernel
Packages, which is subscribed to linux-bluefield in Ubuntu.
ht
the original patch series for 6.0 kernel is herhe
https://patchwork.kernel.org/project/linux-rdma/cover/0-v2-472615b3877e+28f7-vfio_dma_buf_...@nvidia.com/
backport to 5.15 upstream kernel (not Bluefield)
https://github.com/torvalds/linux/compare/master...jgunthorpe:linux:for-william
v3 patch has
Public bug reported:
* Intro
Got call trace RIP below, internal bug ID: 4059202
[ 747.269287] WARNING: CPU: 3 PID: 8360 at
include/net/netfilter/nf_flow_table.h:261
mlx5_tc_ct_del_ft_cb.part.0+0x17b/0x1c0 [mlx5_core]
* Solution
Need to backport
"sched: act_ct: take care of padding in struct
** Description changed:
* Summary
Current BF kernel doesn't support VFIO P2P, the following patches need
to backport, based on 6.0 kernel.
1.
https://patchwork.kernel.org/project/kvm/cover/0-v2-1bd95d72f298+e0e-vfio_pci_priv_...@nvidia.com/
2.
https://patchwork.kernel.org/project
Public bug reported:
* Summary
Current BF kernel doesn't support VFIO P2P, the following patches need
to backport, based on 6.0 kernel.
1.
https://patchwork.kernel.org/project/kvm/cover/0-v2-1bd95d72f298+e0e-vfio_pci_priv_...@nvidia.com/
2.
https://patchwork.kernel.org/project/kvm/cover/0-v2-0
** Description changed:
* Explain the bug(s)
We run a simple CPU-only xor application.
int main(void) {
- long i;
- int count = 0;
- for (i=0; i< 100; i++) {
- count = count ^ i;
- }
- return 0;
+ long i;
+ in
Public bug reported:
* Explain the bug(s)
We run a simple CPU-only xor application.
int main(void) {
long i;
int count = 0;
for (i=0; i< 100; i++) {
count = count ^ i;
}
return 0;
}
tested on BlueField-2 and found consistent performa
** Description changed:
* Explain the bug(s)
We run a simple CPU-only xor application.
+ test.c
+ --
int main(void) {
long i;
int count = 0;
for (i=0; i< 100; i++) {
count = count ^ i;
}
return 0;
}
tested
root@c-237-155-120-p88-00-0-bf2:/boot# diff config-5.15.0-1043-bluefield
config-5.15.0-1044-bluefield
3c3
< # Linux/arm64 5.15.0-1043-bluefield Kernel Configuration
---
> # Linux/arm64 5.15.0-1044-bluefield Kernel Configuration
37c37
< CONFIG_VERSION_SIGNATURE="Ubuntu 5.15.0-1043.45-bluefield 5.15
Public bug reported:
intro
-
Our internal test triggers a kernel crash dump below
[ 888.690348] Sun Mar 24 23:51:59 2024: DriVerTest - Start Test
[ 888.691834]
[ 888.983912] mlx5_core 00
** Description changed:
Intro:
==
- When hit a kernel warning when using openvswitch kernel module. Digging into
the source code, we found it's due to the code snippet
- if (WARN_ON(i.cmd >= family->resv_start_op &&
-(i.doit.validate || i.dumpit.validate)))
-
Public bug reported:
Intro:
==
When hit a kernel warning when using openvswitch kernel module. Digging into
the source code, we found it's due to the code snippet
if (WARN_ON(i.cmd >= family->resv_start_op &&
(i.doit.validate || i.dumpit.validate)))
Public bug reported:
* intro
In bug 2053155 "Add DPLL and syncE support" below:
https://bugs.launchpad.net/ubuntu/+source/linux-bluefield/+bug/2053155
It requires using a yaml spec file, dpll.yaml, and a python script, cli.py, to
verify the correctness.
ex:
$ sudo ./tools/net/ynl/cli.py --spec
Public bug reported:
* intro
As IDR can't protect itself from the concurrent modification, place
idr_remove() under the protection of tp->lock.
* how to fix
need to backport the patch from Jianbo
"net/sched: flower: Add lock protection when remove filter handle"
https://git.kernel.org/pub/scm/lin
** Description changed:
* intro
Synchronous Ethernet, or SyncE, is an ITU-T standard for computer networking
that facilitates the transference of clock signals over the Ethernet physical
layer. It is used to pass timing from node to node and is particularly
important for mobile networks.
** Description changed:
* intro
Synchronous Ethernet, or SyncE, is an ITU-T standard for computer networking
that facilitates the transference of clock signals over the Ethernet physical
layer. It is used to pass timing from node to node and is particularly
important for mobile networks.
my first time sending pull request, do I make any mistake so it's
Invalid?
--
You received this bug notification because you are a member of Kernel
Packages, which is subscribed to linux-bluefield in Ubuntu.
https://bugs.launchpad.net/bugs/2053155
Title:
Add DPLL and syncE support
Status in l
Public bug reported:
* intro
Synchronous Ethernet, or SyncE, is an ITU-T standard for computer networking
that facilitates the transference of clock signals over the Ethernet physical
layer. It is used to pass timing from node to node and is particularly
important for mobile networks.
The DPLL
Hi Taihsiang,
cat /root@bu-lab6-oob:~# cat /etc/mlnx-release
DOCA_2.5.0_BSP_4.5.0_Ubuntu_22.04-1.20231116.dev
--
You received this bug notification because you are a member of Kernel
Packages, which is subscribed to linux-bluefield in Ubuntu.
https://bugs.launchpad.net/bugs/2021930
Title:
kdum
adding "-d" to kexec which shows debugging info. I did not see any error...
[ 441.468322] kdump-tools[7548]: arch_process_options:178: command_line:
BOOT_IMAGE=/boot/vmlinuz-5.15.0-1031-bluefield root=1
[ 441.468586] kdump-tools[7548]: arch_process_options:180: initrd:
/var/lib/kdump/initrd.im
I also tested it on 5.15.0-1031-bluefield and it also fails.
Configurations:
root@bu-oob:~# kdump-config show
DUMP_MODE: kdump
USE_KDUMP: 1
KDUMP_COREDIR: /var/crash
crashkernel addr: 0xbd00
/var/lib/kdump/vmlinuz: symbolic link to /boot/vmlinuz-5.15.0-10
Public bug reported:
Summary:
Machine hangs when loading OFED 2310 mlx5 driver at BlueField
How to reproduce:
# load the OFED driver
Reason:
BF got stuck and observed call trace "mlx5_sf_hw_table_init+0xf4/0x2d0
[mlx5_core]
dmesg from minicom:
[ 726.569928] INFO: task systemd-udevd:297 blocke
Public bug reported:
Summary:
Machine hangs when doing devlink reload
How to reproduce:
Host:
[root@bu-lab24v ~]# echo '2' > /sys/class/net/ens2f0np0/device/sriov_numvfs
Arm:
root@bu-lab24v-oob:~# uname -r
5.15.0-1027-bluefield
root@bu-lab24v-oob:~# devlink dev eswitch set pci/:03:00.0 mod
add merge request
https://code.launchpad.net/~wtu/ubuntu/+source/linux-bluefield/+git/linux-bluefield/+merge/450806
--
You received this bug notification because you are a member of Kernel
Packages, which is subscribed to linux-bluefield in Ubuntu.
https://bugs.launchpad.net/bugs/2032378
Title:
** Also affects: linux-bluefield (Ubuntu)
Importance: Undecided
Status: New
** No longer affects: linux-bluefield (Ubuntu Impish)
** No longer affects: linux-bluefield (Ubuntu Kinetic)
** Changed in: linux-bluefield (Ubuntu)
Assignee: (unassigned) => William Tu (wtu)
** Chan
Public bug reported:
* Explain the bug(s)
when setting up conntrack offload with tcp pedit (test-ct-tcp-pedit.sh),
encounter kernel panic.
* brief explanation of fixes
In tc_setup_flow_action, need to properly assign action.
Which fixes previous commit ("UBUNTU: SAUCE: net/sched: Provide act to
Public bug reported:
* Explain the bug(s)
We've found that for Jammy 5.15 and also 5.4.0-1049 kernel running on
Bluefield, the kdump doesn't work when enabling secure boot[1].
* How to test
Make sure kernel config:
CONFIG_SECURITY_LOCKDOWN_LSM=y
CONFIG_SECURITY_LOCKDOWN_LSM_EARLY=y
CONFIG_LOCK_
With this fix, does kdump/kexec work with
CONFIG_SECURITY_LOCKDOWN_LSM=y?
--
You received this bug notification because you are a member of Kernel
Packages, which is subscribed to kexec-tools in Ubuntu.
https://bugs.launchpad.net/bugs/2008523
Title:
Make kexec work by default on secureboot sys
Public bug reported:
fix race on port output
* Explain the bug(s)
there is a race condition on port output
* brief explanation of fixes
We found that upstream linux has the same issue and fix
commit 066b86787fa3d97b7aefb5ac0a99a22dad2d15f8
Author: Felix Huettner
Date: Wed Apr 5 07:53:41 20
** Description changed:
* Explain the bug(s)
While conntrack tuples are offloaded to hardware and conntrack packet
accounting is enabled, offloaded packets aren't counted.
* Brief explanation of fixes
Cherry-pick. No adaptation. First commit for SW, second commit of HW
offloade
Public bug reported:
* Explain the bug(s)
conntrack -D or conntrack -F doesn't delete offloaded tuples.
* brief explanation of fixes
Add support for to delete offloaded tuples via netlink interface and
userspace conntrack utility.
* How to test
Create OVS bridge with 2 devices mlx5 rep d
Public bug reported:
Support hardware stats for tc actions
* Explain the bug(s)
tc hardware stats are not offloaded from some offloaded actions (pedit, sample,
skbedit)
* brief explanation of fixes
Implement the necessary callbacks to report back stats for these actions.
* How to test
Public bug reported:
* Explain the bug(s)
Currently tc miss interface only supports resuming from a specific tc chain.
If a packet modification is done before a missable action such as CT and
there is a miss in CT after it, this may cause a miss match when resuming
re-executing the same chain in
** Changed in: linux-bluefield (Ubuntu Focal)
Status: Confirmed => In Progress
--
You received this bug notification because you are a member of Kernel
Packages, which is subscribed to linux-bluefield in Ubuntu.
https://bugs.launchpad.net/bugs/2008136
Title:
netfilter: flowtable: add co
** Changed in: linux-bluefield (Ubuntu Focal)
Status: In Progress => Confirmed
--
You received this bug notification because you are a member of Kernel
Packages, which is subscribed to linux-bluefield in Ubuntu.
https://bugs.launchpad.net/bugs/2008136
Title:
netfilter: flowtable: add co
** Description changed:
* Explain the bug(s)
When CT HW offload is enabled, the CT stats does not show the stats of the
offloaded flow.
Ex: using
cat /proc/net/nf_conntrack, or conntrack -L
shows only the software CT stats, not the offloaded CT stats
* Brief explanation of fixes
** Description changed:
* Explain the bug(s)
When CT HW offload is enabled, the CT stats does not show the stats of the
offloaded flow.
- Ex: using
+ Ex: using
cat /proc/net/nf_conntrack, or conntrack -L
shows only the software CT stats, not the offloaded CT stats
* Brief explana
Public bug reported:
* Explain the bug(s)
When CT HW offload is enabled, the CT stats does not show the stats of the
offloaded flow.
Ex: using
cat /proc/net/nf_conntrack, or conntrack -L
shows only the software CT stats, not the offloaded CT stats
* Brief explanation of fixes
Cherry-pick. No
45 matches
Mail list logo
