On Sun, Feb 26, 2023 at 02:58:44PM -0500, Theodore Ts'o wrote: > On Sat, Feb 25, 2023 at 06:14:11PM -0800, Linus Torvalds wrote: > > On Sat, Feb 25, 2023 at 3:40 PM Theodore Ts'o <tytso@xxxxxxx> wrote: > > > > > > For what it's worth, in the VM world (e.g., qemu, AWS, GCP, Azure, > > > Linode, etc.) serial consoles are quite common way of debugging VM's, > > > and as an emergency login path when the networking has been screwed up > > > for some reason.... > > > > Everybody seems to be missing the point. > > > > We don't make new drivers "default y" (or, in this case, "default SERIAL_8250". > > > > It does not matter ONE WHIT if you have a serial device in your > > machine. If your old driver was enabled and worked for you and you > > used it daily, that is ENTIRELY IMMATERIAL to a new driver, even if > > that new driver then happens to use some of the same infrastructure as > > the old one did. > > Oh, agreed, I wasn't responding to that part of your message. New > serial drivers should never be enabled by default. +1 here. I don't know how I missed that during review. Some of the "new" (not really, the split of the 8250_pci) drivers I made in the past inherited that so user won't see the change (sudden disappearance of the console w/o touching defconfig). -- With Best Regards, Andy Shevchenko
