On Tue, Sep 14, 2021 at 09:38:58AM -0700, Linus Torvalds wrote:
On Tue, Sep 14, 2021 at 9:28 AM Al Viro <viro@xxxxxxxxxxxxxxxxxx> wrote:
See https://lkml.kernel.org/lkml/YP2dTQPm1wGPWFgD@xxxxxxxxxxxxxxxxxxxxx/T/
for original thread.
Ok, thanks for the pointer.
I'm actually surprised we have that "multiple nested signals" thing
still. I had this memory of us having removed that, and just returning
to user space with the one signal frame set up.
Too much headache and too many things to keep track of - for example,
you have to cope with fun like raising (and handling) SIGTRAP on
singlestepping into signal handler on some architectures. That would
need to be special-cased and so would SIGSEGV on failure to set sigframe
up, etc.
Al, back to digging through io_uring and hating every minute of that...
