Date: Mon, 14 Dec 2009 12:45:26 +0100 From: Ivan Voras <ivoras@freebsd.org> To: Alexander Motin <mav@freebsd.org> Cc: FreeBSD-Current <freebsd-current@freebsd.org> Subject: Re: How do I use NCQ of Intel X25-E(SSD) on ahci(4)? Message-ID: <9bbcef730912140345u6dada7f4g3b445b18c531ed63@mail.gmail.com> In-Reply-To: <4B262119.6090904@FreeBSD.org> References: <1258575782.00184876.1258562404@10.7.7.3> <mailpost.1258563606.7337474.35089.mailing.freebsd.current@FreeBSD.cs.nctu.edu.tw> <1258579386.00184908.1258567802@10.7.7.3> <1258590193.00184983.1258579804@10.7.7.3> <4B262119.6090904@FreeBSD.org>
next in thread | previous in thread | raw e-mail | index | archive | help
2009/12/14 Alexander Motin <mav@freebsd.org>: > Ivan Voras wrote: >> 2009/11/18 Alexander Motin <mav@freebsd.org>: >>> Ivan Voras wrote: >> >>>> I know next to nothing about AHCI and drivers so this might be obviously >>>> wrong but wouldn't a quick (i.e. MFC-able) obvious temporary fix be to say >>>> >>>> numslots = min(get_minimum_tags_of_all_drives(), ...) >>>> ? >>> Problem is that SIM driver has no idea about devices capabilities, and >>> also doesn't have method to resize queue after attach. In SCSI case, >>> tags are random and only simultaneous number of request is limited, and >>> this is handled fine by CAM. SATA NCQ is more restrictive, allowing to >>> use only tags 0..(N-1). I am planning to make XPT inform SIM about >>> supported tags for each device, to allow SIM to use that information >>> while scheduling requests. I didn't do it yet, just because most of >>> devices able to handle all 32 tags possible on SATA. This Intel SSD is >>> one of rare exceptions. >> >> Ok, (still thinking about something that could be MFC-able in an >> emergency), how about adding a loader tunable integer instead of "32"? > > Just for note, it is already fixed and merged. Thank you!
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?9bbcef730912140345u6dada7f4g3b445b18c531ed63>