I have no idea what this all means, this needs a review from somebody from the SCSI community.

(Oh, and I think scsi_id really should move somewhere else – sg3_utils? –, otherwise it will always be under-supported)

btw, we don't do "S-o-b" in systemd. That's a kernel thing.

I have no idea what this all means, this needs a review from somebody from the SCSI community.

I think @hreinecke is part of the SCSI community

I think scsi_id really should move somewhere else – sg3_utils?

It is the current plan and it's mostly done: sq_inq is the tool that is supposed the replace scsi_id IIRC.

We still have one use of scsi_id in systemd which is in 60-persistent-storage.rules:

# SCSI devices
KERNEL=="sd*[!0-9]|sr*", ENV{ID_SERIAL}!="?*", IMPORT{program}="scsi_id --export --whitelisted -d $devnode", ENV{ID_BUS}="scsi"
KERNEL=="cciss*", ENV{DEVTYPE}=="disk", ENV{ID_SERIAL}!="?*", IMPORT{program}="scsi_id --export --whitelisted -d $devnode", ENV{ID_BUS}="cciss"
KERNEL=="sd*|sr*|cciss*", ENV{DEVTYPE}=="disk", ENV{ID_SERIAL}=="?*", SYMLINK+="disk/by-id/$env{ID_BUS}-$env{ID_SERIAL}"
KERNEL=="sd*|cciss*", ENV{DEVTYPE}=="partition", ENV{ID_SERIAL}=="?*", SYMLINK+="disk/by-id/$env{ID_BUS}-$env{ID_SERIAL}-part%n"

@hreinecke, can sg_inq be used to generate these symlinks ? IOW Could these rules be moved in the sg3_utils package ?

If so that would be the last bits in systemd that still requires the presence of scsi_id.

Thanks.

I have no idea what this all means, this needs a review from somebody from the SCSI community.

I think @hreinecke is part of the SCSI community

I know Hannes, but we require at least one other person than the original author to look at a patch before we merge it.

I'm unsure if I, being a coworker of Hannes, qualify as a member of the SCSI community. Doug Gilbert probably does. In his patch for sg3_utils for EMC Symmetrix support (hreinecke/sg3_utils@8cf2275) he hardcoded the desig_type variable (which corresponds to scsi_id_search_values->id_type in scsi_id) to 3, which is the same thing Hannes did in this patch.

So, ACK from me.

Also no progress here for almost 2 months.

@poettering according to @mwilck the same change has been merged in sg3_utils. That surely means that it has been reviewed by the SCSI community.

And FWIW, SLE has the patch and no regressions have been reported so far.

Thanks.

@fbuihuu, the sg3_utils changes aren't merged yet. I was discussing my patches with @hreinecke right now.

EDIT: Forget it, I thought this was #7594. Sorry. The patch that we're talking about here is of course merged in sg3_utils.

My thinking was that #7594 might make this one obsolete, but that might take even more time than this one.

My thinking was that #7594 might make this one obsolete, but that might take even more time than this one.

Yes and as long as we ship scsi_id, we still need to maintain it.

@poettering, it's still stuck after almost 8 months.

The patch has been reviewed by the SCSI community since a similar change has been merged in sg3_utils, see #7190 (comment)

And it seems that no other reviews will happen.

Could we make a decision here ?

I'm setting the v239 label so hopefully a decision will be taken for that milestone.

quite frankly i have zero idea what this is. but if @hreinecke and @mwilck say it's fine, let's merge it.

Thanks !