Hi Xiaogang, Alex, threading note: my clarification after Alex's ack is archived on lore, but not attached to this patch instance there. Repeating only the essential point here because this patch carries my Tested-by tag.
My Tested-by covers the reproduced split-tail fault on RX 7600 XT. It does not validate the head-side boundary condition in this exact patch. This version still uses head->last on the head side. In svm_range_split_head(), the generated head range is [old_start, new_start - 1], so the split boundary is new_start, i.e. head->last + 1. Technical details and boundary examples: https://lore.kernel.org/all/[email protected]/ Please either use an exclusive-boundary version for the head-side condition, e.g. new_start as in v2, or drop my Tested-by from this exact version. v2: https://lore.kernel.org/all/[email protected]/ Thanks, Gerhard
