Hi

It’s me

Leo chat

If anybody at least somewhat familiar with viona wants to have a look code.illumos.org/c/illumos-gate/+/2980, 3rd attempt. Last time it was decided to just change the default until it can be properly fixed eventually. So here is that, including the man page change to indicate that and a way to 'restore the old behaviour' for those that want the extra performance back and were not hitting the problem. (I have personally not noticed any

performance loss, then again it works now instead of just stalling on big transfers xD so hard to compare really)

→ CODE REVIEW 2980: 15464 Change viona default to always copy tx buffers (NEW) | illumos.org/issues/15464

without source diving, is there an easy way anyone knows offhand to tell if an HBA driver is using SCSv3 or not?

look for *tgtmap* something in its symbols? :)

By using nm and nothing else jbk?

that should work..

I would probably look for scsi_hba_iport_register.

the problem with scsi_hba_iport_register is that mpt_sas uses it while not really being scsav3

so may be both?

Ah, right. My mistake.

I think there are some drivers that are "scsav3" but don't really use tgtmaps either.

I guess it depends on what does it really mean to be SCSAv3, which I can't really answer.

I would assume complex addressing and iports is the true thing?

and letting framework configure targets instead of manually rolling out the ndi_devi_online/ndi_devi_offline and stuff

yeah

[illumos-gate] 15781 loader.efi: implement efi_get_protocol_handles() -- Toomas Soome <tsoome⊙mc>

that reminds me.. i need to file a ticket for adding more support for identifying various EFI GUIDs to loader

yes, I actually have some related work in progress there, but it is a bit of trap too

tsoome: github.com/illumos/illumos-gate/com…..jasonbking:illumos-gate:efi-guids is what I have

it looked like some of the files might have originally come from intel

but i could never find an upstream that might have been more appropriate as a source

I have been using edk2/MdePkg/Include

ahh

i don't know if that stuff will overlap with it or not

there is good chance they are good.

sadly, the uefi code on my home server doesn't support the smartcard protocol :(

would have been nice for a project

the data structure behind lsefi has issue that GUID's are not pointer to value, so we have some wasted space there.

though i was able to see some GUIDs get translated with that change that otherwise weren't on my home server

yes, that would be the expected result:)

rmustacc: I feel like the biggest SCSAv3 thing is not doing the dumb packet construction thing from the old interface

viz., tran_init_pkt(9E) being the terrible old thing, and tran_setup_pkt(9E) being the substantially less terrible new thing

I suppose that might technically be SCSAv2 or whatever