Skip to contentSkip to conversation

broken ide-tape

DISCUSSION Jiri Slaby 2 participants 2 messages May 31, 2009 libata

lore
Share & tools
lore

Jiri Slaby

author
#0 lore raw

Hi,

ide-tape is broken since commit 346331f82823682fb5c2df6e827db25d95cc2e92 (ide-tape: remove atomic test/set macros for packet commands) from Apr 18 2008.

The commit introduced flags which are defined as (1 << nr) and uses them in {clear,set,test}_bit. And hence it must oops on each do_request.

Is it worth fixing (something like ilog2 added to each bit operation) or is a complete removal of the driver better "fix"?

#1 lore raw
Quoted 8 lines · Jiri Slaby

Hi,

ide-tape is broken since commit 346331f82823682fb5c2df6e827db25d95cc2e92 (ide-tape: remove atomic test/set macros for packet commands) from Apr 18 2008.

The commit introduced flags which are defined as (1 << nr) and uses them in {clear,set,test}_bit. And hence it must oops on each do_request.

/me ducks behind the couch.

Ouch, that _is_ dumb, sorry. This is because we wanted to keep the atomic ops for the pipeline handling (which got removed) and use simple flags for the rest. And so the flags got mixed up and the mess unleashed...

Quoted 2 lines

Is it worth fixing (something like ilog2 added to each bit operation) or is a complete removal of the driver better "fix"?

Actually, the driver needs to be reaudited and fixed properly - we still have some users left. I'm looking into it.

Thanks for catching that.

Signature · 2 lines
Regards/Gruss,
    Boris.