| Message ID | 1408072477-10008-1-git-send-email-wangyijing@huawei.com |
|---|---|
| State | New |
| Headers | show |
On Fri, Aug 15, 2014 at 11:14:37AM +0800, Yijing Wang wrote: > We had cached the MSI msg in __write_msi_msg(), > so we should retrieve the cached MSI msg directly, > not to access the MSI address/data hardware registers > again. Is this fixing a particular problem? Could you kindly explain to me when you do the __write_msi_msg before doing an read_msi_msg? The order of operation in this function is read, and then write. > > Signed-off-by: Yijing Wang <wangyijing@huawei.com> > CC: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> > CC: xen-devel@lists.xenproject.org > --- > arch/x86/pci/xen.c | 2 +- > 1 files changed, 1 insertions(+), 1 deletions(-) > > diff --git a/arch/x86/pci/xen.c b/arch/x86/pci/xen.c > index 4656099..7a74404 100644 > --- a/arch/x86/pci/xen.c > +++ b/arch/x86/pci/xen.c > @@ -228,7 +228,7 @@ static int xen_hvm_setup_msi_irqs(struct pci_dev *dev, int nvec, int type) > return 1; > > list_for_each_entry(msidesc, &dev->msi_list, list) { > - read_msi_msg(msidesc, &msg); > + get_cached_msi_msg(msidesc, &msg); > pirq = MSI_ADDR_EXT_DEST_ID(msg.address_hi) | > ((msg.address_lo >> MSI_ADDR_DEST_ID_SHIFT) & 0xff); > if (msg.data != XEN_PIRQ_MSI_DATA || > -- > 1.7.1 >
On 2014/8/15 11:51, Konrad Rzeszutek Wilk wrote: > On Fri, Aug 15, 2014 at 11:14:37AM +0800, Yijing Wang wrote: >> We had cached the MSI msg in __write_msi_msg(), >> so we should retrieve the cached MSI msg directly, >> not to access the MSI address/data hardware registers >> again. > > Is this fixing a particular problem? > > Could you kindly explain to me when you do the __write_msi_msg > before doing an read_msi_msg? The order of operation in this > function is read, and then write. Hi Konrad, Thanks for your review and comments. It looks like I made a mistake. Before I thought xen_hvm_setup_msi_irqs() here worked like restore MSI irq, but it seems not in real. Since every time we call the setup MSI irq function, the msi_desc always new allocated, so here should use read_msi_msg(), I'm sorry for my mistake. :( Bjorn, please drop this one. Thanks! Yijing. > >> >> Signed-off-by: Yijing Wang <wangyijing@huawei.com> >> CC: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> >> CC: xen-devel@lists.xenproject.org >> --- >> arch/x86/pci/xen.c | 2 +- >> 1 files changed, 1 insertions(+), 1 deletions(-) >> >> diff --git a/arch/x86/pci/xen.c b/arch/x86/pci/xen.c >> index 4656099..7a74404 100644 >> --- a/arch/x86/pci/xen.c >> +++ b/arch/x86/pci/xen.c >> @@ -228,7 +228,7 @@ static int xen_hvm_setup_msi_irqs(struct pci_dev *dev, int nvec, int type) >> return 1; >> >> list_for_each_entry(msidesc, &dev->msi_list, list) { >> - read_msi_msg(msidesc, &msg); >> + get_cached_msi_msg(msidesc, &msg); >> pirq = MSI_ADDR_EXT_DEST_ID(msg.address_hi) | >> ((msg.address_lo >> MSI_ADDR_DEST_ID_SHIFT) & 0xff); >> if (msg.data != XEN_PIRQ_MSI_DATA || >> -- >> 1.7.1 >> > >
diff --git a/arch/x86/pci/xen.c b/arch/x86/pci/xen.c index 4656099..7a74404 100644 --- a/arch/x86/pci/xen.c +++ b/arch/x86/pci/xen.c @@ -228,7 +228,7 @@ static int xen_hvm_setup_msi_irqs(struct pci_dev *dev, int nvec, int type) return 1; list_for_each_entry(msidesc, &dev->msi_list, list) { - read_msi_msg(msidesc, &msg); + get_cached_msi_msg(msidesc, &msg); pirq = MSI_ADDR_EXT_DEST_ID(msg.address_hi) | ((msg.address_lo >> MSI_ADDR_DEST_ID_SHIFT) & 0xff); if (msg.data != XEN_PIRQ_MSI_DATA ||
We had cached the MSI msg in __write_msi_msg(), so we should retrieve the cached MSI msg directly, not to access the MSI address/data hardware registers again. Signed-off-by: Yijing Wang <wangyijing@huawei.com> CC: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> CC: xen-devel@lists.xenproject.org --- arch/x86/pci/xen.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-)