Re: PCI arbiter crash on last qemu image

[Samuel Thibault]

Samuel Thibault, le dim. 06 sept. 2020 15:17:51 +0200, a ecrit:
> Samuel Thibault, le dim. 06 sept. 2020 15:14:27 +0200, a ecrit:
> > Thanks for working on this!
> > 
> > I have uploaded libpciaccess_0.16-1+hurd.5 with the latest upstream
> > version.
> 
> One issue remains, however: Xorg's vesa driver produces
> 
> [1669282.478] (II) VESA(0): initializing int10
> [1669282.478] (EE) VESA(0): Cannot read int vect
> 
> which comes from hw/xfree86/int10/generic.c xf86ExtendedInitInt10:
> 
>     if (!sysMem)
>         pci_device_map_legacy(pInt->dev, V_BIOS, BIOS_SIZE + SYS_BIOS - 
> V_BIOS,
>                               PCI_DEV_MAP_FLAG_WRITABLE, &sysMem);
>     INTPriv(pInt)->sysMem = sysMem;
> 
>     if (!readIntVec(pInt->dev, base, LOW_PAGE_SIZE)) {
>         xf86DrvMsg(pScrn->scrnIndex, X_ERROR, "Cannot read int vect\n");
>         goto error1;
>     }
> 
> where
> 
> #define V_BIOS    0xC0000
> #define BIOS_SIZE 0x10000
> #define SYS_BIOS  0xF0000
> 
> static Bool
> readIntVec(struct pci_device *dev, unsigned char *buf, int len)
> {
>     void *map;
> 
>     if (pci_device_map_legacy(dev, 0, len, 0, &map))
>         return FALSE;
> 
>     memcpy(buf, map, len);
>     pci_device_unmap_legacy(dev, map, len);
> 
>     return TRUE;
> }
> 
> Linux uses its own implementation, so it could very well be a bug in the
> code above, please check according to what you have understood of the
> libpciaccess mapping practices.

(or possibly it's simply a pci-arbiter bug?)

Samuel