| Message ID | 1384940196-32514-1-git-send-email-shangw@linux.vnet.ibm.com (mailing list archive) |
|---|---|
| State | Superseded |
| Headers | show |
On Wed, 2013-11-20 at 17:36 +0800, Gavin Shan wrote: > While we detect frozen PE on PHB3, it's always meaningful to have > the dumped diag-data for further diagnosis and analysis. Don't we trip that during PCI probing ? For example if we probe behind a PCI-X bridge (which can exist on an adapter) we'll trip EEH on every non-existing device won't we ? Cheers, Ben. > Signed-off-by: Gavin Shan <shangw@linux.vnet.ibm.com> > --- > arch/powerpc/platforms/powernv/eeh-ioda.c | 3 +++ > 1 file changed, 3 insertions(+) > > diff --git a/arch/powerpc/platforms/powernv/eeh-ioda.c b/arch/powerpc/platforms/powernv/eeh-ioda.c > index 02245ce..481528d 100644 > --- a/arch/powerpc/platforms/powernv/eeh-ioda.c > +++ b/arch/powerpc/platforms/powernv/eeh-ioda.c > @@ -994,8 +994,11 @@ static int ioda_eeh_next_error(struct eeh_pe **pe) > if (ioda_eeh_get_pe(hose, frozen_pe_no, pe)) > break; > > + /* It would be always indicative to have PHB diag-data */ > pr_err("EEH: Frozen PE#%x on PHB#%x detected\n", > (*pe)->addr, (*pe)->phb->global_number); > + ioda_eeh_phb_diag(hose); > + > ret = 1; > goto out; > }
On Wed, Nov 20, 2013 at 08:38:48PM +1100, Benjamin Herrenschmidt wrote: >On Wed, 2013-11-20 at 17:36 +0800, Gavin Shan wrote: >> While we detect frozen PE on PHB3, it's always meaningful to have >> the dumped diag-data for further diagnosis and analysis. > >Don't we trip that during PCI probing ? For example if we probe behind >a PCI-X bridge (which can exist on an adapter) we'll trip EEH on every >non-existing device won't we ? > Yes, we already had the dumped PHB diag-data when detecting frozen PE during PCI probing. After PCI probing is completed, the EEH takes over and we won't dump PHB diag-data during PCI config cycles. Took a close look on what we have in the code. Those functions to dump PHB (P7IOC & PHB3) needs a bit rework or refactoring since we're dumping same PHB diag-data in pci.c and eeh-ioda.c at the same time. Besides, I think the appropriate place to dump PHB diag-data (for EEH core itself) would be ioda_eeh_get_log(), which is the indirect backend of eeh_ops::get_log, instead of the function ioda_eeh_next_error(). Ben, please drop this one for now and I'll send the revised one :-) Thanks, Gavin >> Signed-off-by: Gavin Shan <shangw@linux.vnet.ibm.com> >> --- >> arch/powerpc/platforms/powernv/eeh-ioda.c | 3 +++ >> 1 file changed, 3 insertions(+) >> >> diff --git a/arch/powerpc/platforms/powernv/eeh-ioda.c b/arch/powerpc/platforms/powernv/eeh-ioda.c >> index 02245ce..481528d 100644 >> --- a/arch/powerpc/platforms/powernv/eeh-ioda.c >> +++ b/arch/powerpc/platforms/powernv/eeh-ioda.c >> @@ -994,8 +994,11 @@ static int ioda_eeh_next_error(struct eeh_pe **pe) >> if (ioda_eeh_get_pe(hose, frozen_pe_no, pe)) >> break; >> >> + /* It would be always indicative to have PHB diag-data */ >> pr_err("EEH: Frozen PE#%x on PHB#%x detected\n", >> (*pe)->addr, (*pe)->phb->global_number); >> + ioda_eeh_phb_diag(hose); >> + >> ret = 1; >> goto out; >> } > >
diff --git a/arch/powerpc/platforms/powernv/eeh-ioda.c b/arch/powerpc/platforms/powernv/eeh-ioda.c index 02245ce..481528d 100644 --- a/arch/powerpc/platforms/powernv/eeh-ioda.c +++ b/arch/powerpc/platforms/powernv/eeh-ioda.c @@ -994,8 +994,11 @@ static int ioda_eeh_next_error(struct eeh_pe **pe) if (ioda_eeh_get_pe(hose, frozen_pe_no, pe)) break; + /* It would be always indicative to have PHB diag-data */ pr_err("EEH: Frozen PE#%x on PHB#%x detected\n", (*pe)->addr, (*pe)->phb->global_number); + ioda_eeh_phb_diag(hose); + ret = 1; goto out; }
While we detect frozen PE on PHB3, it's always meaningful to have the dumped diag-data for further diagnosis and analysis. Signed-off-by: Gavin Shan <shangw@linux.vnet.ibm.com> --- arch/powerpc/platforms/powernv/eeh-ioda.c | 3 +++ 1 file changed, 3 insertions(+)