[dpdk-dev] [PATCH v2] bus/pci: forbid VA as IOVA mode if IOMMU address width too small

[Zhang, Qi Z]

> -----Original Message-----
> From: Tan, Jianfeng
> Sent: Friday, January 12, 2018 11:01 AM
> To: 'Maxime Coquelin' <maxime.coquelin at redhat.com>; dev at dpdk.org;
> stable at dpdk.org; santosh.shukla at caviumnetworks.com; Burakov, Anatoly
> <anatoly.burakov at intel.com>; thomas at monjalon.net;
> stephen at networkplumber.org
> Cc: peterx at redhat.com; Zhang, Qi Z <qi.z.zhang at intel.com>
> Subject: RE: [PATCH v2] bus/pci: forbid VA as IOVA mode if IOMMU address
> width too small
> 
> 
> 
> > -----Original Message-----
> > From: Maxime Coquelin [mailto:maxime.coquelin at redhat.com]
> > Sent: Tuesday, January 9, 2018 9:18 PM
> > To: dev at dpdk.org; stable at dpdk.org; Tan, Jianfeng;
> > santosh.shukla at caviumnetworks.com; Burakov, Anatoly;
> > thomas at monjalon.net; stephen at networkplumber.org
> > Cc: peterx at redhat.com; Maxime Coquelin
> > Subject: [PATCH v2] bus/pci: forbid VA as IOVA mode if IOMMU address
> > width too small
> >
> > Intel VT-d supports different address widths for the IOVAs, from
> > 39 bits to 56 bits.
> >
> > While recent processors support at least 48 bits, VT-d emulation
> > currently only supports 39 bits. It makes DMA mapping to fail in this
> > case when using VA as IOVA mode, as user-space virtual addresses uses
> > up to 47 bits (see kernel's Documentation/x86/x86_64/mm.txt).
> >
> > This patch parses VT-d CAP register value available in sysfs, and
> > forbid VA as IOVA mode if the GAW is 39 bits or unknown.
> >
> > Fixes: f37dfab21c98 ("drivers/net: enable IOVA mode for Intel PMDs")
> >
> > Cc: stable at dpdk.org
> > Signed-off-by: Maxime Coquelin <maxime.coquelin at redhat.com>
> 
> I don't have strong objection on this patch. Plus, this patch has been verified
> with --no-huge, and Intel low-end processors (with the help of Zhang Qi).

Sorry, my bad, I just found it is tested with igb_uio.

re-tested it with vfio, seems VA width is no decoded correctly

I have sagaw = 4, which means 48b VA address, but the real device is 39, so IOVA is not disabled, still see that failure.

Thanks
Qi

> 
> Tested-by: Jianfeng Tan <jianfeng.tan at intel.com>
> 
> > ---
> >
> > Changes in v2:
> > ==============
> > - Rework pci_one_device_iommu_support_va #ifdefery (Stephen)
> > - Don't inline introduced functions (Stephen)
> >
> >  drivers/bus/pci/linux/pci.c | 108
> > ++++++++++++++++++++++++++++++++++++++++----
> >  1 file changed, 99 insertions(+), 9 deletions(-)
> >
> > diff --git a/drivers/bus/pci/linux/pci.c b/drivers/bus/pci/linux/pci.c
> > index 25f907e04..0a43c4b89 100644
> > --- a/drivers/bus/pci/linux/pci.c
> > +++ b/drivers/bus/pci/linux/pci.c
> > @@ -547,6 +547,100 @@ pci_one_device_has_iova_va(void)
> >  	return 0;
> >  }
> >
> > +#if defined(RTE_ARCH_X86)
> > +static bool
> > +pci_one_device_iommu_support_va(struct rte_pci_device *dev) {
> > +#define VTD_CAP_SAGAW_SHIFT         8
> > +#define VTD_CAP_SAGAW_MASK          (0x1fULL <<
> > VTD_CAP_SAGAW_SHIFT)
> > +#define X86_VA_WIDTH 47 /* From Documentation/x86/x86_64/mm.txt
> */
> > +	struct rte_pci_addr *addr = &dev->addr;
> > +	char filename[PATH_MAX];
> > +	FILE *fp;
> > +	uint64_t sagaw, vtd_cap_reg = 0;
> > +	int guest_addr_width = 0;
> > +
> > +	snprintf(filename, sizeof(filename),
> > +		 "%s/" PCI_PRI_FMT "/iommu/intel-iommu/cap",
> > +		 rte_pci_get_sysfs_path(), addr->domain, addr->bus, addr-
> > >devid,
> > +		 addr->function);
> > +	if (access(filename, F_OK) == -1) {
> > +		/* We don't have an Intel IOMMU, assume VA supported*/
> 
> A nitpick: missed a space between "supported" and "*/"
> 
> Thanks,
> Jianfeng