Re: IDE interrupt masking and serial ports (was Re: Lockout during

Stephen C. Tweedie (sct@redhat.com)
Wed, 25 Nov 1998 15:57:27 GMT


Hi,

On Wed, 18 Nov 1998 00:13:36 +0000, Mark Lord <mlord@pobox.com> said:

> "Benjamin C.R. LaHaise" wrote:
> ...
>> Dumping PPP packets in the default configuration is just a really *bad*
>> thing to impress users with. If unmask irq and dma are going to be
>> disabled by default, then the driver should only transfer 1 sector or do
>> whatever is *necessary* to ensure that irq latency isn't as ridiculous as
>> it is presently.

> The default is to transfer a single sector per-interrupt,
> UNLESS the user has requested otherwise by enabling
> "BLOCK MODE" (multi-sector mode) in the BIOS.

Unfortunately this is different from 2.0. With precisely the same
config on both boots, after a fresh boot into 2.0, hdparm -a gives me:

/dev/hda:
multcount = 0 (off)

And 2.1 current:

/dev/hda:
multcount = 16 (on)

If we keep unmaskirq to be 0 by default, then the plain fact is that 2.1
defaults to doing more IO and therefore disabling interrupts for longer
than 2.0 does.

--Stephen

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/