linux-kernel - Re: [PATCH v2 1/8] tty: add a poll() callback in struct tty

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Date:   Thu, 15 Jun 2017 17:44:03 +0300
From:   Tal Shorer <tal.shorer@...il.com>
To:     Alan Cox <gnomes@...rguk.ukuu.org.uk>
Cc:     linux-doc@...r.kernel.org,
        "<linux-kernel@...r.kernel.org>" <linux-kernel@...r.kernel.org>,
        USB list <linux-usb@...r.kernel.org>,
        "<gregkh@...uxfoundation.org>" <gregkh@...uxfoundation.org>,
        Felipe Balbi <balbi@...nel.org>,
        Jonathan Corbet <corbet@....net>
Subject: Re: [PATCH v2 1/8] tty: add a poll() callback in struct tty_operations

On Wed, Jun 14, 2017 at 4:33 PM, Alan Cox <gnomes@...rguk.ukuu.org.uk> wrote:
>> That would cut it, but TIOCPKT is too coupled with having a linked tty.
>> I could make acm behave like a pty (accept TIOCPKT and issue the
>> ctrl_status bits), but for that I need n_tty to know that packet does
>> not always mean a linked tty is present, and that in case it isn't we
>> take our own ctrl_status bits instead of the link's. I could write a
>> small (inline?) function to fetch the correct ctrl_status bits and put
>> that in n_tty. Does that make sense?
>
> I think that makes sense, and I would do the job properly rather than do
> a hack with tty->link. Those hacks in the long term never work out the
> best approach.
>
> Alan
Ok, so I'm doing that and everything is great until I got to actually
modifying tty->termios.
I need to modify it from interrupt context (the usb_request's
complete() callback), but modifying the termios requires a
rw_semaphore I can't take.
I could queue_work() to do it, but then I'd have to flush the work
from another non-sleepable context in acm_disconnect() (which runs
under a spinlock).
I can't change the semaphore to a spinlock because some drivers that
use it actually wanna sleep while holding it.
Any ideas?