Re: [PATCH net-next v4 00/18] net/smc: implement virtual ISM extension and loopback-ism

From: Wen Gu
Date: Thu Sep 28 2023 - 12:42:31 EST




On 2023/9/27 23:16, Alexandra Winter wrote:


On 24.09.23 17:16, Wen Gu wrote:
Wen Gu (18):
net/smc: decouple ism_dev from SMC-D device dump
net/smc: decouple ism_dev from SMC-D DMB registration
net/smc: extract v2 check helper from SMC-D device registration
net/smc: support SMCv2.x supplemental features negotiation
net/smc: reserve CHID range for SMC-D virtual device
net/smc: extend GID to 128bits only for virtual ISM device
net/smc: disable SEID on non-s390 architecture
net/smc: enable virtual ISM device feature bit
net/smc: introduce SMC-D loopback device
net/smc: implement ID-related operations of loopback
net/smc: implement some unsupported operations of loopback
net/smc: implement DMB-related operations of loopback
net/smc: register loopback device as SMC-Dv2 device
net/smc: add operation for getting DMB attribute
net/smc: add operations for DMB attach and detach
net/smc: avoid data copy from sndbuf to peer RMB in SMC-D
net/smc: modify cursor update logic when sndbuf mapped to RMB
net/smc: add interface implementation of loopback device

drivers/s390/net/ism_drv.c | 20 +-
include/net/smc.h | 32 ++-
include/uapi/linux/smc.h | 3 +
include/uapi/linux/smc_diag.h | 2 +
net/smc/Kconfig | 13 ++
net/smc/Makefile | 2 +-
net/smc/af_smc.c | 88 ++++++--
net/smc/smc.h | 7 +
net/smc/smc_cdc.c | 56 ++++-
net/smc/smc_cdc.h | 1 +
net/smc/smc_clc.c | 64 ++++--
net/smc/smc_clc.h | 10 +-
net/smc/smc_core.c | 111 +++++++++-
net/smc/smc_core.h | 9 +-
net/smc/smc_diag.c | 11 +-
net/smc/smc_ism.c | 100 ++++++---
net/smc/smc_ism.h | 24 ++-
net/smc/smc_loopback.c | 489 ++++++++++++++++++++++++++++++++++++++++++
net/smc/smc_loopback.h | 54 +++++
net/smc/smc_pnet.c | 4 +-
20 files changed, 996 insertions(+), 104 deletions(-)
create mode 100644 net/smc/smc_loopback.c
create mode 100644 net/smc/smc_loopback.h


Hello Wen Gu,

I applied and built your patches and noticed some things that you may want to consider in the next version:

Series should be split up [2]

Several lines exceed 80 columns [1][3]

'git clang-format HEAD~18' finds several formatting issues.
Maybe not all of them need to be fixed.

codespell *.patch
0006-net-smc-extend-GID-to-128bits-only-for-virtual-ISM-d.patch:7: protocal ==> protocol

With your patches applied I get some new warnings [4]:
Seems there are some ntoh conversions missing

CHECK net/smc/af_smc.c
net/smc/af_smc.c:723:32: warning: cast to restricted __be64
net/smc/af_smc.c:1427:52: warning: cast to restricted __be64
CHECK net/smc/smc_pnet.c
CHECK net/smc/smc_ib.c
CHECK net/smc/smc_clc.c
net/smc/smc_clc.c:954:72: warning: incorrect type in argument 1 (different base types)
net/smc/smc_clc.c:954:72: expected unsigned short [usertype] chid
net/smc/smc_clc.c:954:72: got restricted __be16 [usertype] chid
net/smc/smc_clc.c:1050:29: warning: incorrect type in assignment (different base types)
net/smc/smc_clc.c:1050:29: expected unsigned long long [usertype] gid
net/smc/smc_clc.c:1050:29: got restricted __be64 [usertype]
net/smc/smc_clc.c:1051:31: warning: incorrect type in assignment (different base types)
net/smc/smc_clc.c:1051:31: expected unsigned long long [usertype] token
net/smc/smc_clc.c:1051:31: got restricted __be64 [usertype]


[1] linux/Documentation/process/coding-style.rst
[2] https://www.kernel.org/doc/html/v6.3/process/maintainer-netdev.html?highlight=network
[3] scripts/checkpatch.pl --strict --max-line-length=80 --git HEAD-18
[4] make C=2 CF=-D__CHECK_ENDIAN__ M=net/smc -Wunused-function -Wimplicit-fallthrough -Wincompatible-function-pointer-types-strict



Hi Sandy,

Thank you very much for your detailed comments. They are really helpful.
I will check and fix them in my next version.


When I installed the patches, I noticed that
smcd info
showed an SEID, even though I had no ISM device --> good


Yes, virtual ISM device will return the right SEID (same as that returned
by ISM device) on s390 arch.


smcd device
FID Type PCI-ID PCHID InUse #LGs PNET-ID
0000 0 0000 Yes 2

This needs some improvements.., but I'm not sure what is the best way to display virtual smcd interfaces in the smc-tools.


Right. My idea is to add the display of the device's name and GID
along with the PCI information (all zero) to indicate it's virtual
smcd device.


I was able to do SMC transfers via the smcd-loopback feature :-D


Glad to hear this. I will improve the issues you mentioned in the next
version and next version will be split up.

Thank you very much.
Wen Gu