> -----Original Message----- > From: Martin Wilck <martin.wilck@xxxxxxxx> > Sent: Tuesday, November 3, 2020 3:00 AM > To: bmarzins@xxxxxxxxxx > Cc: dm-devel@xxxxxxxxxx; ng-eseries-upstream-maintainers <ng-eseries-upstream-maintainers@xxxxxxxxxx>; > christophe.varoqui@xxxxxxxxxxx; Schremmer, Steven <Steve.Schremmer@xxxxxxxxxx> > Subject: Re: [PATCH 4/5] libmultipath: only read 0xc9 vpd page for devices with rdac checker > > On Mon, 2020-11-02 at 19:11 -0600, Benjamin Marzinski wrote: > > On Fri, Oct 30, 2020 at 09:12:46PM +0000, Martin Wilck wrote: > > > > > > Suggestion: we could try to retrieve VPD 0 (supported VPDs) before > > > checking VPD 0xc9. That would avoid the errors on non-Netapp > > > hardware, > > > while still allowing us to autodetect RDAC systems that are missing > > > in > > > the hwtable. > > > > When this idea was brought up in discussions about how to solve this, > > there was a worry that older rdac devices might not not support vpd > > page > > 0 correctly. I'm not sure how valid that worry is. > > These old devices should be in the hwtable. > > The algorithm would be something like this: > > A) if hwtable says "rdac", goto D) > B) if hwtable says nothing, goto E) > C) otherwise use hwtable settings, END > D) try VPD 0xc9 > 1) if it works as expected, use RDAC, END > 2) otherwise, goto F) > E) try VPD 0 > 1) if it fails, goto F) > 2) if it works and page c9 is listed, goto D) > 3) if it works and c9 is not listed, goto F) > F) use defaults, END > > I suppose "hwtable" means all built-in and user configuration here, > including "overrides". But I haven't thought that through yet. > > Cheers, > Martin I think Martin's algorithm would work, but it would also be safe to use VPD 0 in all cases as rdac systems have supported page 0 for well over a decade. I think having "hwtable" include user configuration would be a good choice, so that new rdac devices could be specified in multipath.conf. Thanks, Steve -- dm-devel mailing list dm-devel@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/dm-devel