On 5/20/21 4:05 PM, Christoph Hellwig wrote: > This fix is pretty gross. Adding pages to bios can fail for all kinds > of reasons, so the fix is to use bio_add_page and check its return > value, and if it needs another bio keep looping and chaining more bios. > OK, I will try this idea and avoid to access BIO_MAX_VECS directly. > And maybe capping the readahead to some sane upper bound still makes > sense, but it should never look at BIO_MAX_VECS for that. > Thanks for the hint. Coly Li
