xfs
[Top] [All Lists]

ADD 800061 - HIGHMEM support is broken

To: nb@xxxxxxx
Subject: ADD 800061 - HIGHMEM support is broken
From: pv@xxxxxxxxxxxxx (chait@xxxxxxxxxxxx)
Date: Thu, 24 Aug 2000 11:12:58 -0700 (PDT)
Cc: linux-xfs@xxxxxxxxxxx
Reply-to: sgi.bugs.xfs@xxxxxxxxxxxxxxxxx
Sender: owner-linux-xfs@xxxxxxxxxxx
Webexec: webpvupdate,pvincident
Webpv: getafix.engr.sgi.com
View Incident: 
http://co-op.engr.sgi.com/BugWorks/code/bwxquery.cgi?search=Search&wlong=1&view_type=Bug&wi=800061

 Status : open                         Priority : 2                         
 Assigned Engineer : nb                Submitter : ananth                   
*Modified User : chait                *Modified User Domain : engr          
*Description :
There are two problems in handling highmem with XFS:

1. consistent use of kmap/kunmap to get at the address
   of a page.

2. supporting highmem with kiobufs.

I'll check-in a short fix that will check, issue warning, etc.
if highmem is turned on.


.....


==========================
ADDITIONAL INFORMATION (ADD)
From: chait@engr (BugWorks)
Date: Aug 24 2000 11:12:57AM
==========================

FYI:

With respect to kiobuf and highmem, the only available solution
is to use bounce buffering. However, there are heated debates 
going on about where exactly this bouncing should be done:
        - I/O queuing layer
        - scsi/ide/lvm/... midlayers
        - device driver layers
        - pci-dma mapping layers

I'm working on a quick'n'dirty fix which should work at the I/O
queuing layer __for now__. This is to conform with the existing
fix for highmem boucing done for buffer-head based I/Os.

This quick'n'dirty fix is just a forward port of sct's original
patch for kiobufs in the 2.2 series of kernels.

-Chait.

<Prev in Thread] Current Thread [Next in Thread>
  • ADD 800061 - HIGHMEM support is broken, chait@xxxxxxxxxxxx <=