On 8/23/23 09:19, Bob Pearson wrote:
I have also seen the same hangs in siw. Not as frequently but the same symptoms.
About every month or so I take another run at trying to find and fix this bug but
I have not succeeded yet. I haven't seen anything that looks like bad behavior from
the rxe side but that doesn't prove anything. I also saw these hangs on my system
before the WQ patch went in if my memory serves. Out main application for this
driver at HPE is Lustre which is a little different than SRP but uses the same
general approach with fast MRs. Currently we are finding the driver to be quite stable
even under very heavy stress.
I would be happy to collaborate with someone (you?) who knows the SRP side well to resolve
this hang. I think that is the quickest way to fix this. I have no idea what SRP is waiting for.
Hi Bob,
I cannot reproduce these issues. All SRP tests work reliably on my test setup on
top of the v6.5-rc7 kernel, whether I use the siw driver or whether I use the
rdma_rxe driver. Additionally, I do not see any SRP abort messages.
# uname -a
Linux opensuse-vm 6.5.0-rc7 #28 SMP PREEMPT_DYNAMIC Wed Aug 23 10:42:35 PDT 2023 x86_64 x86_64 x86_64 GNU/Linux
# journalctl --since=today | grep 'SRP abort' | wc
0 0 0
Since I installed openSUSE Tumbleweed in the VM in which I run kernel tests: if
you are using a Linux distro that is based on Debian it may include a buggy
version of multipathd. Last time I ran the SRP tests in a Debian VM I had to
build multipathd from source - the SRP tests did not work with the Debian version
of multipathd. The shell script that I use to build and install multipathd is as
follows (must be run in the multipath-tools source directory):
#!/bin/bash
scriptdir="$(dirname "$0")"
if type -p zypper >/dev/null 2>&1; then
rpms=(device-mapper-devel libaio-devel libjson-c-devel librados-devel
liburcu-devel readline-devel systemd-devel)
for p in "${rpms[@]}"; do
sudo zypper install -y "$p"
done
elif type -p apt-get >/dev/null 2>&1; then
export LIB=/lib
sudo apt-get install -y libaio-dev libdevmapper-dev libjson-c-dev librados-dev \
libreadline-dev libsystemd-dev liburcu-dev
fi
git clean -f
make -s "$@"
sudo make -s "$@" install
Bart.
