must. resist. urge. to. pull. on. thread.

(we appear to call sd_check_vpd_page_support() which issues a SCSI inquiry multiple times during attach.. and doesn't seem like we do anything that would necessitate doing that)

i saw Tribblix had a recent update, but OmniOS hasn't and the last 'bloody' update was a considerable time ago... is it still updated?

darwin - there was an omnios update a couple of days ago, with the openssl security fixes omnios.org/rn/r46

and openjdk8, apparently

and bloody has openssl 3.1.2, updated on the first of August:

% pkg list -nvg pkg.omnios.org/bloody/core openssl-3

pkg://omnios/library/security/openssl-3⊙3:20230801T171832Z ---

[illumos-gate] 15818 loader.efi: cstyle cleanup after 15781 -- Toomas Soome <tsoome⊙mc>

does illumos have any plans to expand development to the risc-v architecture?

[illumos-gate] 15821 Tidy up the cfgadm manual pages -- Peter Tribble <peter.tribble⊙gc>

hrm... how committeed is the sata interface (in terms of sata hba drivers using it?)

sata_get_rdwr_pmult_pkt takes a uint8_t for 'regn'.. that should (AFAICT) actually be uint16_t

I think any int type (e.g. uint8_t) < 'unsigned int' gets converted, so accepting uint16_t _should_ be ok

but i don't feel very confident in that either..

jbk: that's likely in the "probably safe but not guaranteed by anyone" category. the code I see generated in godbolt only reads part of the argument register in the callee but writes the whole thing in the caller (perhaps because partial-register writes are reportedly more expensive due to having a dependency on the old value of the register)

jbk: It's specifically consolidation private, so we can change it as required.

I would not rely / assume anything about changing the types there being safe. But again, this is basically private to us, so just change it as required.

I don't know if any older still in use closed bins would rely on port multiplier framework. Probably not.

if you're feeling excruciatingly careful, you can rename the function when you change it to be sure we'll notice if someone's using it. there's only one caller in illumos-gate (in ahci.c)

squidgy: my sense is that it's been talked about as a possible target but I see more the energy right now invested in finishing the ARM port (and getting the arm port integrated will make integrating a third active architecture easier)

ok, i'll include that then -- I was just cleaning up sata.c (making it smatch/warning clean) and that was something smatch pointed out

(it was doing 'uint8_t >> 8' basically, but the spec I found implies the register field is actually 16 bits

I wouldn't bother with the renaming.

This is private.

Let's just take advantage of the fact that it's private?

yeah, i'll just include the uint8_t -> uint16_t change along with the rest of the fixes