On Thu, Oct 05, 2017 at 09:28:34AM +0100, Bruce Richardson wrote:
> On Thu, Oct 05, 2017 at 01:06:41AM +0100, Ferruh Yigit wrote:
> > On 6/3/2017 11:57 PM, Changpeng Liu wrote:
> > > For PCI prefetchable resources, Linux will create a
> > > write combined file as well, the library will try
> > > to map resourceX_wc file first, if the file does
> > > not exist, then it will map resourceX as usual.
> > 
> > Hi Changpeng,
> > 
> > Code part looks OK, but can you please describe more why we should try
> > write combined resource file first, what is the benefit of using it _wc
> > file?
> > 
> > Thanks,
> > ferruh
> > 
> Also, if we use a write combining resource file, I believe we will use
s/will use/will lose/

> correct ordering of instructions within our drivers. Does applying this
> patch not also mean that we would need memory barriers inside all the
> drivers, so as to ensure that we don't have a queue doorbell write
> reordered with the descriptor writes? I don't think it's safe to apply
> this change on it's own, without driver changes, since all PMDs assume
> strong ordering on IA.
> 
> Regards,
> /Bruce
> 
> > 
> > > 
> > > Signed-off-by: Changpeng Liu <[email protected]>
> > > ---
> > >  lib/librte_eal/linuxapp/eal/eal_pci_uio.c | 19 ++++++++++++++-----
> > >  1 file changed, 14 insertions(+), 5 deletions(-)
> > > 
> > > diff --git a/lib/librte_eal/linuxapp/eal/eal_pci_uio.c 
> > > b/lib/librte_eal/linuxapp/eal/eal_pci_uio.c
> > > index fa10329..d9fc20a 100644
> > > --- a/lib/librte_eal/linuxapp/eal/eal_pci_uio.c
> > > +++ b/lib/librte_eal/linuxapp/eal/eal_pci_uio.c
> > > @@ -321,7 +321,7 @@
> > >  
> > >   /* update devname for mmap  */
> > >   snprintf(devname, sizeof(devname),
> > > -                 "%s/" PCI_PRI_FMT "/resource%d",
> > > +                 "%s/" PCI_PRI_FMT "/resource%d_wc",
> > >                   pci_get_sysfs_path(),
> > >                   loc->domain, loc->bus, loc->devid,
> > >                   loc->function, res_idx);
> > > @@ -335,13 +335,22 @@
> > >   }
> > >  
> > >   /*
> > > -  * open resource file, to mmap it
> > > +  * open prefetchable resource file first, try to mmap it
> > >    */
> > >   fd = open(devname, O_RDWR);
> > >   if (fd < 0) {
> > > -         RTE_LOG(ERR, EAL, "Cannot open %s: %s\n",
> > > -                         devname, strerror(errno));
> > > -         goto error;
> > > +         snprintf(devname, sizeof(devname),
> > > +                         "%s/" PCI_PRI_FMT "/resource%d",
> > > +                         pci_get_sysfs_path(),
> > > +                         loc->domain, loc->bus, loc->devid,
> > > +                         loc->function, res_idx);
> > > +         /* then try to map resource file */
> > > +         fd = open(devname, O_RDWR);
> > > +         if (fd < 0) {
> > > +                 RTE_LOG(ERR, EAL, "Cannot open %s: %s\n",
> > > +                                 devname, strerror(errno));
> > > +                 goto error;
> > > +         }
> > >   }
> > >  
> > >   /* try mapping somewhere close to the end of hugepages */
> > > 
> > 

Reply via email to