WARNING - OLD ARCHIVES

This is an archived copy of the Xen.org mailing list, which we have preserved to ensure that existing links to archives are not broken. The live archive, which contains the latest emails, can be found at http://lists.xen.org/
   
 
 
Xen 
 
Home Products Support Community News
 
   
 

xen-devel

Re : Re : Re : Re : [Xen-devel] Re: Patches for VGA-Passthrough XEN 4.2

To: JavMV <javier.manzano@xxxxxxxxxx>, "xen-devel@xxxxxxxxxxxxxxxxxxx" <xen-devel@xxxxxxxxxxxxxxxxxxx>
Subject: Re : Re : Re : Re : [Xen-devel] Re: Patches for VGA-Passthrough XEN 4.2 unstable
From: David TECHER <davidtecher@xxxxxxxx>
Date: Wed, 7 Sep 2011 17:25:36 +0100 (BST)
Cc:
Delivery-date: Wed, 07 Sep 2011 09:26:14 -0700
Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=yahoo.fr; s=s1024; t=1315412736; bh=qJXeoB4Db/F3v5gB71ZooR0vJ3/QDpDqu/RRb3ANNbg=; h=X-YMail-OSG:Received:X-Mailer:References:Message-ID:Date:From:Reply-To:Subject:To:In-Reply-To:MIME-Version:Content-Type; b=afY0fJsRdEaW5pfpEkl/BWtDAWJgQBdlo9faYqSyirmEZnH9nJp5YlzMRXkFRtarYRDKiuCVs6u9Y0ktRwSl/TPz9VBQDBvcGJaJE5bvNMaNMycaE2VtK2NE7KZ4KWTl1swOQutIcCwebDM8e5+L7uZrymkhehs/7xOqdBHazrE=
Domainkey-signature: a=rsa-sha1; q=dns; c=nofws; s=s1024; d=yahoo.fr; h=X-YMail-OSG:Received:X-Mailer:References:Message-ID:Date:From:Reply-To:Subject:To:In-Reply-To:MIME-Version:Content-Type; b=YZKqTwA6mgcAZnhxAzQpNt8PViuDl7noTY6PZM+fmEsi+sYYLqiTGGbhoGZXxCnMc4fiGeeckNoIIBOJ1D9aL/RXomETqLIC/6wC8UWhJHC9mVkCIxG0rMUBuwcQ7N/W5due2aBXhnKv+/zZ7Beyk65Itgkp5q7OopXjRlsLal8=;
Envelope-to: www-data@xxxxxxxxxxxxxxxxxxx
In-reply-to: <1315408931610-4778925.post@xxxxxxxxxxxxx>
List-help: <mailto:xen-devel-request@lists.xensource.com?subject=help>
List-id: Xen developer discussion <xen-devel.lists.xensource.com>
List-post: <mailto:xen-devel@lists.xensource.com>
List-subscribe: <http://lists.xensource.com/mailman/listinfo/xen-devel>, <mailto:xen-devel-request@lists.xensource.com?subject=subscribe>
List-unsubscribe: <http://lists.xensource.com/mailman/listinfo/xen-devel>, <mailto:xen-devel-request@lists.xensource.com?subject=unsubscribe>
References: <1315231341254-4770602.post@xxxxxxxxxxxxx> <1315310225039-4774097.post@xxxxxxxxxxxxx> <1315343879.61837.YahooMailNeo@xxxxxxxxxxxxxxxxxxxxxxxxxxx> <1315382742268-4777689.post@xxxxxxxxxxxxx> <1315389138.56652.YahooMailNeo@xxxxxxxxxxxxxxxxxxxxxxxxxxx> <1315398360680-4778339.post@xxxxxxxxxxxxx> <1315401901083-4778494.post@xxxxxxxxxxxxx> <1315403644525-4778586.post@xxxxxxxxxxxxx> <1315405372.87224.YahooMailNeo@xxxxxxxxxxxxxxxxxxxxxxxxxxx> <1315405921936-4778727.post@xxxxxxxxxxxxx> <1315408931610-4778925.post@xxxxxxxxxxxxx>
Reply-to: David TECHER <davidtecher@xxxxxxxx>
Sender: xen-devel-bounces@xxxxxxxxxxxxxxxxxxx
Javier

Since you can test before the end of the month, please be informed to find the "all" patches here

Hope this helps.

David

------------------------------------------------------------

--- tools/firmware/hvmloader/acpi/dsdt.asl      2011-08-13 18:43:39.000000000 +0200
+++ tools/firmware/hvmloader/acpi/dsdt.asl      2011-08-13 18:58:25.000000000 +0200
@@ -173,15 +173,43 @@ DefinitionBlock ("DSDT.aml", "DSDT", 2,
                         0x00000000,
                         0x00020000)

+                    /* reserve MMIO BARs of gfx for 1:1 mapping */
                     DWordMemory(
                         ResourceProducer, PosDecode, MinFixed, MaxFixed,
                         Cacheable, ReadWrite,
                         0x00000000,
-                        0xF0000000,
-                        0xF4FFFFFF,
+                        0xFA000000,
+                        0xFAFFFFFF,
                         0x00000000,
-                        0x05000000,
-                        ,, _Y01)
+                        0x01000000)
+
+                    DWordMemory(
+                        ResourceProducer, PosDecode, MinFixed, MaxFixed,
+                        NonCacheable, ReadWrite,
+                        0x00000000,
+                        0xC0000000,
+                        0xCFFFFFFF,
+                        0x00000000,
+                        0x10000000)
+
+                    DWordMemory(
+                        ResourceProducer, PosDecode, MinFixed, MaxFixed,
+                        Cacheable, ReadWrite,
+                        0x00000000,
+                        0xD0000000,
+                        0xD1FFFFFF,
+                        0x00000000,
+                        0x02000000)
+
+                    DWordMemory(
+                        ResourceProducer, PosDecode, MinFixed, MaxFixed,
+                        Cacheable, ReadWrite,
+                        0x00000000,
+                        0xFB000000,
+                        0xFB07FFFF,
+                        0x00000000,
+                        0x00080000,
+                       ,, _Y01)
                 })

                 CreateDWordField(PRT0, \_SB.PCI0._CRS._Y01._MIN, MMIN)
--- tools/firmware/hvmloader/hvmloader.c        2011-08-13 18:43:39.000000000 +0200
+++ tools/firmware/hvmloader/hvmloader.c        2011-08-13 19:16:21.000000000 +0200
@@ -31,6 +31,7 @@
 #include <xen/hvm/params.h>

 #define ROM_INCLUDE_VGABIOS
+#define ROM_INCLUDE_PTVGABIOS
 #define ROM_INCLUDE_ETHERBOOT
 #include "roms.inc"

@@ -113,6 +114,9 @@ asm (

 unsigned long scratch_start = SCRATCH_PHYSICAL_ADDRESS;

+/* virtual BDF of pass-throughed gfx */
+uint8_t gfx_bdf;
+
 static void init_hypercalls(void)
 {
     uint32_t eax, ebx, ecx, edx;
@@ -310,6 +314,20 @@ static int pci_load_option_roms(unsigned
     return rom_phys_addr - rom_base_addr;
 }

+static void init_vm86_tss(void)
+{
+    void *tss;
+    struct xen_hvm_param p;
+
+    tss = mem_alloc(128, 128);
+    memset(tss, 0, 128);
+    p.domid = DOMID_SELF;
+    p.index = HVM_PARAM_VM86_TSS;
+    p.value = virt_to_phys(tss);
+    hypercall_hvm_op(HVMOP_set_param, &p);
+    printf("vm86 TSS at %08lx\n", virt_to_phys(tss));
+}
+
 /* Replace possibly erroneous memory-size CMOS fields with correct values. */
 static void cmos_write_memory_size(void)
 {
@@ -336,25 +354,6 @@ static void cmos_write_memory_size(void)
     cmos_outb(0x35, (uint8_t)( alt_mem >> 8));
 }

-/*
- * Set up an empty TSS area for virtual 8086 mode to use.
- * The only important thing is that it musn't have any bits set
- * in the interrupt redirection bitmap, so all zeros will do.
- */
-static void init_vm86_tss(void)
-{
-    void *tss;
-    struct xen_hvm_param p;
-
-    tss = mem_alloc(128, 128);
-    memset(tss, 0, 128);
-    p.domid = DOMID_SELF;
-    p.index = HVM_PARAM_VM86_TSS;
-    p.value = virt_to_phys(tss);
-    hypercall_hvm_op(HVMOP_set_param, &p);
-    printf("vm86 TSS at %08lx\n", virt_to_phys(tss));
-}
-
 static void apic_setup(void)
 {
     /* Set the IOAPIC ID to the static value used in the MP/ACPI tables. */
@@ -487,8 +486,10 @@ int main(void)
             break;
         case VGA_pt:
             printf("Loading VGABIOS of passthroughed gfx ...\n");
-            vgabios_sz = round_option_rom(
-                (*(uint8_t *)(VGABIOS_PHYSICAL_ADDRESS+2)) * 512);
+        memcpy((void *)VGABIOS_PHYSICAL_ADDRESS,
+               vgabios_pt, sizeof(vgabios_pt));
+        *(uint8_t *)(VGABIOS_PHYSICAL_ADDRESS + sizeof(vgabios_pt)) = gfx_bdf;
+        vgabios_sz = round_option_rom(sizeof(vgabios_pt) + 1);
             break;
         default:
             printf("No emulated VGA adaptor ...\n");
@@ -525,7 +526,7 @@ int main(void)
         hypercall_hvm_op(HVMOP_set_param, &p);
     }

-    init_vm86_tss();
+   init_vm86_tss();

     cmos_write_memory_size();

--- tools/firmware/hvmloader/Makefile   2011-08-13 18:43:39.000000000 +0200
+++ tools/firmware/hvmloader/Makefile   2011-08-13 18:58:36.000000000 +0200
@@ -56,6 +56,7 @@ CIRRUSVGA_ROM := ../vgabios/VGABIOS-lgpl
 else
 CIRRUSVGA_ROM := ../vgabios/VGABIOS-lgpl-latest.cirrus.bin
 endif
+PTVGA_ROM     := ../vgabios/vgabios-pt.bin

 .PHONY: all
 all: subdirs-all
@@ -94,6 +95,11 @@ ifneq ($(CIRRUSVGA_ROM),)
        sh ./mkhex vgabios_cirrusvga $(CIRRUSVGA_ROM) >> $@.new
        echo "#endif" >> $@.new
 endif
+ifneq ($(PTVGA_ROM),)
+       echo "#ifdef ROM_INCLUDE_PTVGABIOS" >> $@.new
+       sh ./mkhex vgabios_pt $(PTVGA_ROM) >> $@.new
+       echo "#endif" >> $@.new
+endif

        echo "#ifdef ROM_INCLUDE_ETHERBOOT" >> $@.new
        cat ../etherboot/eb-roms.h >> $@.new
--- tools/ioemu-remote/hw/pass-through.c        2011-08-13 23:15:13.000000000 +0200
+++ tools/ioemu-remote/hw/pass-through.c        2011-08-13 18:58:41.000000000 +0200
@@ -3243,6 +3243,8 @@ static int pt_cmd_reg_read(struct pt_dev
 }

 /* read BAR */
+static int gfx_first_read_BAR[7] = {1, 1, 1, 1, 1, 1, 1};
+
 static int pt_bar_reg_read(struct pt_dev *ptdev,
         struct pt_reg_tbl *cfg_entry,
         uint32_t *value, uint32_t valid_mask)
@@ -3265,6 +3267,17 @@ static int pt_bar_reg_read(struct pt_dev
     /* use fixed-up value from kernel sysfs */
     *value = ptdev->pci_dev->base_addr[index];

+    if ( ptdev->pci_dev->device_class == 0x300 )
+    {
+        if ( gfx_first_read_BAR[index] == 1 )
+        {
+            gfx_first_read_BAR[index] = 0;
+            PT_LOG("first read BARs of gfx\n");
+            return 0;
+        }
+    }
+
+
     /* set emulate mask depend on BAR flag */
     switch (ptdev->bases[index].bar_flag)
     {
--- tools/firmware/hvmloader/pci.c      2011-08-13 18:43:39.000000000 +0200
+++ tools/firmware/hvmloader/pci.c      2011-08-13 18:58:53.000000000 +0200
@@ -33,6 +33,10 @@ unsigned long pci_mem_end = PCI_MEM_END;

 enum virtual_vga virtual_vga = VGA_none;

+/* virtual BDF of pass-throughed gfx declared in hvmloader.c*/
+extern uint8_t gfx_bdf;
+
+
 void pci_setup(void)
 {
     uint32_t base, devfn, bar_reg, bar_data, bar_sz, cmd, mmio_total = 0;
@@ -93,9 +97,44 @@ void pci_setup(void)
             }
             else if ( virtual_vga == VGA_none )
             {
-                vga_devfn = devfn;
-                virtual_vga = VGA_pt;
+                 virtual_vga = VGA_pt;
+                 gfx_bdf = devfn;
+
+                 /* Make vBAR=pBAR */
+                 printf("Make vBAR = pBAR of assigned gfx\n");
+                 for ( bar = 0; bar < 7; bar++ )
+                 {
+                     bar_reg = PCI_BASE_ADDRESS_0 + 4*bar;
+                     if ( bar == 6 )
+                             bar_reg = PCI_ROM_ADDRESS;
+                     /* When first time read, it will return physical address */
+                     bar_data = pci_readl(devfn, bar_reg);
+                     pci_writel(devfn, bar_reg, bar_data);
+
+                     /* Now enable the memory or I/O mapping. */
+                     cmd = pci_readw(devfn, PCI_COMMAND);
+                     if ( (bar_reg == PCI_ROM_ADDRESS) ||
+                              ((bar_data & PCI_BASE_ADDRESS_SPACE) ==
+                               PCI_BASE_ADDRESS_SPACE_MEMORY) )
+                           cmd |= PCI_COMMAND_MEMORY;
+                     else
+                           cmd |= PCI_COMMAND_IO;
+                     cmd |= PCI_COMMAND_MASTER;
+                     pci_writew(devfn, PCI_COMMAND, cmd);
+                }
+
+                 /* Map the interrupt. */
+                 pin = pci_readb(devfn, PCI_INTERRUPT_PIN);
+                 if ( pin != 0 )
+                 {
+                     /* This is the barber's pole mapping used by Xen. */
+                     link = ((pin - 1) + (devfn >> 3)) & 3;
+                     isa_irq = pci_readb(PCI_ISA_DEVFN, 0x60 + link);
+                     pci_writeb(devfn, PCI_INTERRUPT_LINE, isa_irq);
+                 }
+                 continue;
             }
+
             break;
         case 0x0680:
             /* PIIX4 ACPI PM. Special device with special PCI config space. */
-------------------------------------------------------------------------------


De : JavMV <javier.manzano@xxxxxxxxxx>
À : xen-devel@xxxxxxxxxxxxxxxxxxx
Envoyé le : Mercredi 7 Septembre 2011 17h22
Objet : Re: Re : Re : Re : [Xen-devel] Re: Patches for VGA-Passthrough XEN 4.2 unstable

I suppose you have modified the dsdt.asl file right? If you did it and it
crashed (If you didn't try it with this let me know because I have to
document the process of adapt the Tobias patches (Actually Han,Weidong
patches)) then there is something that Xen devels have added between those
changeset to the code that make the David's adaptation works (because David
says he didn't add code to the patches).

I was working on a specific computer with XEN at university and I can't try
this patchset now, that is why I have to wait to end of month.. Im
impatient!!

Well thanks for the answers.

Javier Manzano.

-----
JavMV
--
View this message in context: http://xen.1045712.n5.nabble.com/Patches-for-VGA-Passthrough-XEN-4-2-unstable-tp4406265p4778925.html
Sent from the Xen - Dev mailing list archive at Nabble.com.

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel
<Prev in Thread] Current Thread [Next in Thread>