At Wed, 19 Feb 2020 13:28:04 +0900, Michael Paquier <michael@paquier.xyz> wrote in > On Wed, Feb 19, 2020 at 01:07:36PM +0900, Kyotaro Horiguchi wrote: > > If we also verify checksum in md layer, callback is overkill since the > > immediate caller consumes the event immediately. We can signal the > > error by somehow returning a file tag. > > FWIW, I am wondering if there is any need for a change here and > complicate more the code. If you know the block number, the page size > and the segment file size you can immediately guess where is the > damaged block. The first information is already part of the error
I have had support requests related to broken block several times, and (I think) most of *them* had hard time to locate the broken block or even broken file. I don't think it is useles at all, but I'm not sure it is worth the additional complexity.
> damaged block. The first information is already part of the error > message, and the two other ones are constants defined at > compile-time.
May you have misread the snippet?
What Hubert proposed is:
"invalid page in block %u of relation file %s; zeroing out page", blkno, <filename>
The second format in my messages just before is: "invalid page in block %u in relation %u, file \"%s\"", blockNum, smgr->smgr_rnode.node.relNode, smgrfname()
All of them are not compile-time constant at all.
I like your error message, the block number is relation level not file level.
I 'll change the error message to
"invalid page in block %u of relation %u, file %s"
