Tom Lane wrote:
> Zdenek Kotala <Zdenek.Kotala@Sun.COM> writes:
>> I would like to make following modification:
>
>> 1) Add ReadBuffer_noerror (recommend me better name) function which will
>> accept damaged page without Error. This page will be marked as corrupted
>> and when ReadBuffer will touch this page then it will be handled in
>> standard way.
>
> This seems like a pretty horrid idea. Bad pages shouldn't be allowed to
> get into shared buffers in the first place. Why not have the checking
> logic operate outside shared buffers?
It currently works outside the shared buffers, but I afraid about
collision due to parallel read and write access on one block. I'm not
sure if parallel write(8k) and read(8k) is synchronized by kernel/fs or
not. If not it should generates false positive results. If yes than I'm
happy :-) with outside processing.
>> 3) Add PageHeaderIsValid check also for write operation
>
>> In production it should catch problem with memory or software bugs. In
>> development it should catch memory overwriting.
>
> Is there any evidence whatsoever to demonstrate that this is worth the
> cycles it will eat?
Alex from clickware tries this modification to catch their problem with
random damaged database. But, I don't have any result yet.
Zdenek