On Tue, Feb 13, 2024 at 01:01:51PM -0800, Bart Van Assche wrote:
On 2/12/24 11:28, Bart Van Assche wrote:
On 2/9/24 10:12, Jens Axboe wrote:
Greg, can you elaborate on how useful cancel is for gadgets? Is it one of those things that was wired up "just because", or does it have actually useful cases?
I found two use cases in the Android Open Source Project and have submitted CLs that request to remove the io_cancel() calls from that code. Although I think I understand why these calls were added, the race conditions that these io_cancel() calls try to address cannot be addressed completely by calling io_cancel().
(replying to my own e-mail) The adb daemon (adbd) maintainers asked me to preserve the I/O cancellation code in adbd because it was introduced recently in that code to fix an important bug. Does everyone agree with the approach of the untested patches below?
But I mean, the cancellation code has seemingly been broken since forever according to your patch description. So their fix which relies on it actually fixes nothing? And they seemingly didn't notice until you told them that it's broken. Can we get a link to that stuff, please?
Btw, you should probably provide that context in your patch series you sent that Christoph and I responded too. Because I just stumbled upon this here and it provided context I was missing.