2007-10-24 14:32:32

by Rodolfo Giometti

Subject: PCMCIA insert on resume

Hello,

I don't understand why, on resume, the function socket_resume() (in
drivers/pcmcia/cs.c) do:

if (!(skt->state & SOCKET_PRESENT)) {
skt->state &= ~SOCKET_SUSPEND;
return socket_insert(skt);
}

This cause a powered off socket before suspend to be powered on during
resume.

To avoid such (erroneous?) behaviour I propose this patch:

diff --git a/drivers/pcmcia/cs.c b/drivers/pcmcia/cs.c
index 729e37d..d69de74 100644
--- a/drivers/pcmcia/cs.c
+++ b/drivers/pcmcia/cs.c
@@ -562,10 +562,8 @@ static int socket_resume(struct pcmcia_socket *skt)
skt->ops->init(skt);
skt->ops->set_socket(skt, &skt->socket);

- if (!(skt->state & SOCKET_PRESENT)) {
- skt->state &= ~SOCKET_SUSPEND;
- return socket_insert(skt);
- }
+ if (!(skt->state & SOCKET_PRESENT))
+ goto resume_exit;

ret = socket_setup(skt, SS_COMA, resume_delay);
if (ret == CS_SUCCESS) {
@@ -599,6 +597,7 @@ static int socket_resume(struct pcmcia_socket *skt)
socket_shutdown(skt);
}

+resume_exit:
skt->state &= ~SOCKET_SUSPEND;

return CS_SUCCESS;

With this patch if the socket was powered down before the suspend, at
resume time it's not powered on and you need a "pccardctl insert" to
activate the socket.

Ciao,

Rodolfo

--

GNU/Linux Solutions e-mail: [email protected]
Linux Device Driver [email protected]
Embedded Systems [email protected]
UNIX programming phone: +39 349 2432127

2007-10-28 18:12:59

by Pavel Machek

Subject: Re: PCMCIA insert on resume

Hi!

> I don't understand why, on resume, the function socket_resume() (in
> drivers/pcmcia/cs.c) do:
>
> if (!(skt->state & SOCKET_PRESENT)) {
> skt->state &= ~SOCKET_SUSPEND;
> return socket_insert(skt);
> }
>
> This cause a powered off socket before suspend to be powered on during
> resume.
>
> To avoid such (erroneous?) behaviour I propose this patch:
>
> diff --git a/drivers/pcmcia/cs.c b/drivers/pcmcia/cs.c
> index 729e37d..d69de74 100644
> --- a/drivers/pcmcia/cs.c
> +++ b/drivers/pcmcia/cs.c
> @@ -562,10 +562,8 @@ static int socket_resume(struct pcmcia_socket *skt)
> skt->ops->init(skt);
> skt->ops->set_socket(skt, &skt->socket);
>
> - if (!(skt->state & SOCKET_PRESENT)) {
> - skt->state &= ~SOCKET_SUSPEND;
> - return socket_insert(skt);
> - }
> + if (!(skt->state & SOCKET_PRESENT))
> + goto resume_exit;
>
> ret = socket_setup(skt, SS_COMA, resume_delay);
> if (ret == CS_SUCCESS) {
> @@ -599,6 +597,7 @@ static int socket_resume(struct pcmcia_socket *skt)
> socket_shutdown(skt);
> }
>
> +resume_exit:
> skt->state &= ~SOCKET_SUSPEND;
>
> return CS_SUCCESS;
>
> With this patch if the socket was powered down before the suspend, at
> resume time it's not powered on and you need a "pccardctl insert" to
> activate the socket.

Seems okay to me, but I do not know much about pcmcia. Add
signed-off-by, changelog at the beggining, cc me, rafael and akpm... ?
Pavel
--
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

2007-10-28 18:39:34

by Rafael J. Wysocki

Subject: Re: PCMCIA insert on resume

On Saturday, 27 October 2007 09:45, Pavel Machek wrote:
> Hi!
>
> > I don't understand why, on resume, the function socket_resume() (in
> > drivers/pcmcia/cs.c) do:
> >
> > if (!(skt->state & SOCKET_PRESENT)) {
> > skt->state &= ~SOCKET_SUSPEND;
> > return socket_insert(skt);
> > }
> >
> > This cause a powered off socket before suspend to be powered on during
> > resume.
> >
> > To avoid such (erroneous?) behaviour I propose this patch:
> >
> > diff --git a/drivers/pcmcia/cs.c b/drivers/pcmcia/cs.c
> > index 729e37d..d69de74 100644
> > --- a/drivers/pcmcia/cs.c
> > +++ b/drivers/pcmcia/cs.c
> > @@ -562,10 +562,8 @@ static int socket_resume(struct pcmcia_socket *skt)
> > skt->ops->init(skt);
> > skt->ops->set_socket(skt, &skt->socket);
> >
> > - if (!(skt->state & SOCKET_PRESENT)) {
> > - skt->state &= ~SOCKET_SUSPEND;
> > - return socket_insert(skt);
> > - }
> > + if (!(skt->state & SOCKET_PRESENT))
> > + goto resume_exit;
> >
> > ret = socket_setup(skt, SS_COMA, resume_delay);
> > if (ret == CS_SUCCESS) {
> > @@ -599,6 +597,7 @@ static int socket_resume(struct pcmcia_socket *skt)
> > socket_shutdown(skt);
> > }
> >
> > +resume_exit:
> > skt->state &= ~SOCKET_SUSPEND;
> >
> > return CS_SUCCESS;
> >
> > With this patch if the socket was powered down before the suspend, at
> > resume time it's not powered on and you need a "pccardctl insert" to
> > activate the socket.
>
> Seems okay to me, but I do not know much about pcmcia. Add
> signed-off-by, changelog at the beggining, cc me, rafael and akpm... ?

Actually, there are some objections wrt this patch from Russell King.

Greetings,
Rafael