Title: INTRQ from a different angle

The requirement is stated in the following section:

Section 7.9.6 of revision 6:

nIEN is the enable bit for the device Assertion of INTRQ to the host. When the nIEN bit is cleared to

zero, and the device is selected, INTRQ shall be enabled through a tri-state buffer and shall be

asserted or negated by the device as appropriate. When the nIEN bit is set to one, or the device is

not selected, the INTRQ signal shall be in a high impedance state.

Note the use of the word "shall" which indicates a mandatory requirement placed on the device (which is driving INTRQ) whenever eIEN is cleared to 0.  The host can do that whenever DMACK- is not asserted and the device is selected.  No matter what the application environment, the device cannot prevent the host from clearing eIEN.  There is no way for the host to detect that a write to device control register failed to have the intended effect (you cannot even read the register - it is write only).  And there is no provision for the device to notify the host via other means (e.g. IDENTIFY DEVICE) that this capability (interrupts) is not supported.

So you cannot make a device that does not support INTRQ and have it pass a compliance test.  However, since it is the host that enables or disables this feature, a host can be made that does not implement interrupts but is still compliant (with of course some limitations others have already pointed out).

Jim

PS note that this is not a new requirement - the exact same text appears in ATA-4.  Earlier versions did not have the phrase "and shall be asserted or negated by the device as appropriate," which might have given some folks wiggle room for ATA-, ATA-2, and ATA-3.

-----Original Message-----
From: Curtis Stevens [mailto:[EMAIL PROTECTED]]
Sent: Wednesday, January 23, 2002 4:41 PM
To: [EMAIL PROTECTED]
Subject: [t13] INTRQ from a different angle

I believe that ATA/ATAPI-6 makes it mandatory for a device to provide the interrupt (I did not actually find the part that makes the interrupt mandatory), I do not think that any T13 specification makes it madatory for the host interface to propigate the interrupt.

-----------------------
Curtis E. Stevens
Pacific Digital Corp.
2052 Alton Parkway
Irvine, CA 92606

Phone (949) 477-5713
Fax (949) 252-9397

E-Mail: [EMAIL PROTECTED]
WEB: www.PacificDigital.com

Reply via email to