Attachment 144287 Details for Bug 216169 – Patch posted to virtualist on 12/19

[patch] Patch posted to virtualist on 12/19

ia64-pvfb.patch (text/plain), 5.84 KB, created by Chris Lalancette on 2006-12-22 15:15:10 UTC

(hide)

Description:

Filename:

MIME Type:

Creator: Chris Lalancette

Created: 2006-12-22 15:15:10 UTC

Size: 5.84 KB

patch

obsolete

>Return-Path: <virtualist-admin@redhat.com>
>Received: from pobox.corp.redhat.com ([unix socket])
>	 by pobox.corp.redhat.com (Cyrus v2.2.12-Invoca-RPM-2.2.12-3.RHEL4.1) with LMTPA;
>	 Tue, 19 Dec 2006 16:41:07 -0500
>X-Sieve: CMU Sieve 2.2
>Received: from int-mx1.corp.redhat.com (int-mx1.corp.redhat.com [172.16.52.254])
>	by pobox.corp.redhat.com (8.13.1/8.12.8) with ESMTP id kBJLf67Y006380;
>	Tue, 19 Dec 2006 16:41:06 -0500
>Received: from post-office.corp.redhat.com (post-office.corp.redhat.com [172.16.52.227])
>	by int-mx1.corp.redhat.com (8.13.1/8.13.1) with ESMTP id kBJLf2sS013346;
>	Tue, 19 Dec 2006 16:41:02 -0500
>Received: from post-office.corp.redhat.com (localhost.localdomain [127.0.0.1])
>	by post-office.corp.redhat.com (8.12.11.20060308/8.11.6) with ESMTP id kBJLf2PI013742;
>	Tue, 19 Dec 2006 16:41:02 -0500
>Received: from int-mx1.corp.redhat.com (int-mx1.corp.redhat.com [172.16.52.254])
>	by post-office.corp.redhat.com (8.12.11.20060308/8.11.6) with ESMTP id kBJLe5fx013554
>	for <virtualist@post-office.corp.redhat.com>; Tue, 19 Dec 2006 16:40:05 -0500
>Received: from pike.pond.sub.org (vpn-4-11.str.redhat.com [10.32.4.11])
>	by int-mx1.corp.redhat.com (8.13.1/8.13.1) with ESMTP id kBJLe4Rs013020
>	for <virtualist@redhat.com>; Tue, 19 Dec 2006 16:40:04 -0500
>Received: by pike.pond.sub.org (Postfix, from userid 1000)
>	id 368596769B; Tue, 19 Dec 2006 22:40:04 +0100 (CET)
>To: virtualist@redhat.com
>Subject: [PATCH RHEL5] BZ 216169: vncfb does not run on IA64
>From: Markus Armbruster <armbru@redhat.com>
>Message-ID: <87irg7muob.fsf@pike.pond.sub.org>
>User-Agent: Gnus/5.1008 (Gnus v5.10.8) Emacs/21.4 (gnu/linux)
>MIME-Version: 1.0
>Content-Type: text/plain; charset=us-ascii
>X-loop: virtualist@redhat.com
>Sender: virtualist-admin@redhat.com
>Errors-To: virtualist-admin@redhat.com
>X-BeenThere: virtualist@redhat.com
>X-Mailman-Version: 2.0.13
>Precedence: bulk
>Reply-To: virtualist@redhat.com
>List-Help: <mailto:virtualist-request@redhat.com?subject=help>
>List-Post: <mailto:virtualist@redhat.com>
>List-Subscribe: <http://post-office.corp.redhat.com/mailman/listinfo/virtualist>,
>	<mailto:virtualist-request@redhat.com?subject=subscribe>
>List-Id: Virtualisation and Resource Management <virtualist.redhat.com>
>List-Unsubscribe: <http://post-office.corp.redhat.com/mailman/listinfo/virtualist>,
>	<mailto:virtualist-request@redhat.com?subject=unsubscribe>
>List-Archive: <http://post-office.corp.redhat.com/archives/virtualist/>
>Date: Tue, 19 Dec 2006 22:40:04 +0100
>
>The new pvfb works on IA-64 in xen-unstable.  The backport to RHEL-5
>doesn't.  That's because xen-unstable has more general functions
>xc_map_foreign_range() and xc_map_foreign_batch() in libxc.  The
>actual smarts of that change are buried in the privcmd device driver
>and/or the hypervisor.  Backporting improvements there to fix this bug
>seems way too risky to me.  I didn't bother to track them down.
>
>Instead, I'm proposing to put the smarts into the vfb backend.
>
>The appended patch tries to filter mfns through
>xc_domain_translate_gpfn_list() before feeding them to
>xc_map_foreign_FOO().  If the guest is in shadow translate mode (the
>case that's broken right now), xc_domain_translate_gpfn_list()
>translates them.  Else, it fails with EINVAL, leaving the mfns alone.
>
>Scratch rpms are at /mnt/redhat/brewroot/scratch/armbru/task_477555
>Jarod Wilson reports they fix IA-64 and is verifying that they don't
>break x86_64.
>
>Please ACK.
>
>
>diff -rup a/tools/xenfb/xenfb.c b/tools/xenfb/xenfb.c
>--- a/tools/xenfb/xenfb.c	2006-12-19 19:40:21.000000000 +0100
>+++ b/tools/xenfb/xenfb.c	2006-12-19 19:37:53.000000000 +0100
>@@ -378,6 +378,48 @@ static int xenfb_wait_for_frontend_initi
> 	return 0;
> }
> 
>+static void *xenfb_map_foreign_range(int xc_handle, uint32_t dom,
>+				     int size, int prot,
>+				     unsigned long mfn)
>+{
>+	int rc;
>+
>+	rc = xc_domain_translate_gpfn_list(xc_handle, dom, 1, &mfn, &mfn);
>+	if (rc < 0 && errno != EINVAL)
>+		return NULL;
>+	return xc_map_foreign_range(xc_handle, dom, size, prot, mfn);
>+}
>+
>+static void *xenfb_map_foreign_batch(int xc_handle, uint32_t dom, int prot,
>+				     xen_pfn_t *arr, int num)
>+{
>+	xen_pfn_t *buf;
>+	int rc;
>+	void *ret;
>+
>+	/* make a copy to avoid clobbering arr[] */
>+	buf = malloc(num * sizeof(*buf));
>+	if (!buf)
>+		return NULL;
>+	memcpy(buf, arr, num * sizeof(*buf));
>+
>+	rc = xc_domain_translate_gpfn_list(xc_handle, dom, num, buf, buf);
>+	if (rc < 0 && errno != EINVAL) {
>+		free(buf);
>+		return NULL;
>+	}
>+
>+	/*
>+	 * Bug alert: xc_map_foreign_batch() can fail partly and
>+	 * return a non-null value.  This is a design flaw.  When it
>+	 * happens, we happily continue here, and later crash on
>+	 * access.
>+	 */
>+	ret = xc_map_foreign_batch(xc_handle, dom, prot, buf, num);
>+	free(buf);
>+	return ret;
>+}
>+
> static int xenfb_map_fb(struct xenfb_private *xenfb, int domid)
> {
> 	struct xenfb_page *page = xenfb->fb.page;
>@@ -389,18 +431,12 @@ static int xenfb_map_fb(struct xenfb_pri
> 	n_fbdirs = n_fbmfns * sizeof(unsigned long);
> 	n_fbdirs = (n_fbdirs + (XC_PAGE_SIZE - 1)) / XC_PAGE_SIZE;
> 
>-	/*
>-	 * Bug alert: xc_map_foreign_batch() can fail partly and
>-	 * return a non-null value.  This is a design flaw.  When it
>-	 * happens, we happily continue here, and later crash on
>-	 * access.
>-	 */
>-	fbmfns = xc_map_foreign_batch(xenfb->xc, domid,
>+	fbmfns = xenfb_map_foreign_batch(xenfb->xc, domid,
> 			PROT_READ, page->pd, n_fbdirs);
> 	if (fbmfns == NULL)
> 		return -1;
> 
>-	xenfb->pub.pixels = xc_map_foreign_batch(xenfb->xc, domid,
>+	xenfb->pub.pixels = xenfb_map_foreign_batch(xenfb->xc, domid,
> 				PROT_READ | PROT_WRITE, fbmfns, n_fbmfns);
> 	if (xenfb->pub.pixels == NULL) {
> 		munmap(fbmfns, n_fbdirs * XC_PAGE_SIZE);
>@@ -428,7 +464,7 @@ static int xenfb_bind(struct xenfb_devic
> 	if (dev->port == -1)
> 		return -1;
> 
>-	dev->page = xc_map_foreign_range(xenfb->xc, dev->otherend_id,
>+	dev->page = xenfb_map_foreign_range(xenfb->xc, dev->otherend_id,
> 			XC_PAGE_SIZE, PROT_READ | PROT_WRITE, mfn);
> 	if (dev->page == NULL)
> 		return -1;
>

Actions: View | Diff

Attachments on bug 216169: 144287 | 144308