pcmcia: patch to fix pccard_store_cis

pcmcia: patch to fix pccard_store_cis

[Pete Zaitcev]

The ``ret'' obviously cannot be zero here, because it's initialized to the
write count and not zero.

For the complete description, see:
 https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=207910

Author: Fabrice Bellet <fabrice@bellet.info>
Signed-off-by: Pete Zaitcev <zaitcev@redhat.com>

--- linux-2.6.17.i686/drivers/pcmcia/socket_sysfs.c	2006-06-18 03:49:35.000000000 +0200
+++ /tmp/socket_sysfs.c	2006-10-01 19:30:09.000000000 +0200
@@ -321,7 +321,7 @@
 
 	kfree(cis);
 
-	if (!ret) {
+	if (ret == count) {
 		mutex_lock(&s->skt_mutex);
 		if ((s->callback) && (s->state & SOCKET_PRESENT) &&
 		    !(s->state & SOCKET_CARDBUS)) {

Re: pcmcia: patch to fix pccard_store_cis

[Dominik Brodowski]

Hi,

On Sun, Oct 01, 2006 at 12:21:07PM -0700, Pete Zaitcev wrote:
> The ``ret'' obviously cannot be zero here, because it's initialized to the
> write count and not zero.

Thanks -- Linus was faster, though, and already applied his patch to the
linux-2.6 git tree. Regarding the other issue seen in RH bug# 207910, I'll
try to take a look at it soon.

	Dominik

Re: pcmcia: patch to fix pccard_store_cis

[Romano Giannetti]

On Mon, 2006-10-02 at 02:31 +0200, Dominik Brodowski wrote:
> Hi,
> 
> On Sun, Oct 01, 2006 at 12:21:07PM -0700, Pete Zaitcev wrote:
> > The ``ret'' obviously cannot be zero here, because it's initialized to the
> > write count and not zero.
> 
> Thanks -- Linus was faster, though, and already applied his patch to the
> linux-2.6 git tree. Regarding the other issue seen in RH bug# 207910, I'll
> try to take a look at it soon.

BTW: I had the same problem, reported here: 

https://launchpad.net/distros/ubuntu/+source/pcmciautils/+bug/52510

and here: 

http://lists.infradead.org/pipermail/linux-pcmcia/2006-August/003893.html

and my modem did work without IRQ problems after I got rid of .cis and
started (obsolete) cardmgr. Just as a data point more... 

Romano 

-- 
Romano Giannetti <romano.giannetti@gmail.com>

Re: pcmcia: patch to fix pccard_store_cis

[Dominik Brodowski]

Hi,

On Mon, Oct 02, 2006 at 03:01:34PM +0200, Romano Giannetti wrote:
> On Mon, 2006-10-02 at 02:31 +0200, Dominik Brodowski wrote:
> > Hi,
> > 
> > On Sun, Oct 01, 2006 at 12:21:07PM -0700, Pete Zaitcev wrote:
> > > The ``ret'' obviously cannot be zero here, because it's initialized to the
> > > write count and not zero.
> > 
> > Thanks -- Linus was faster, though, and already applied his patch to the
> > linux-2.6 git tree. Regarding the other issue seen in RH bug# 207910, I'll
> > try to take a look at it soon.
> 
> BTW: I had the same problem, reported here: 
> 
> https://launchpad.net/distros/ubuntu/+source/pcmciautils/+bug/52510
> 
> and here: 
> 
> http://lists.infradead.org/pipermail/linux-pcmcia/2006-August/003893.html
> 
> and my modem did work without IRQ problems after I got rid of .cis and
> started (obsolete) cardmgr. Just as a data point more... 

Does it work again (after re-copying the cis file to /lib/firmware) when
you use this patch?

Thanks,
	Dominik

>From 4bb59569454f09e8bfc3a0f7bbdef46ccc7a51e0 Mon Sep 17 00:00:00 2001
From: Dominik Brodowski <linux@dominikbrodowski.net>
Date: Fri, 3 Nov 2006 10:54:00 -0500
Subject: [PATCH] pcmcia: start over after CIS override

When overriding the CIS, re-start the configuration of the card from
scratch. Reported and debugged by Fabrice Bellet <fabrice@bellet.info>

Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net>
---
 drivers/pcmcia/cs_internal.h  |    2 +-
 drivers/pcmcia/ds.c           |   12 ++++++++----
 drivers/pcmcia/socket_sysfs.c |    4 ++--
 3 files changed, 11 insertions(+), 7 deletions(-)

diff --git a/drivers/pcmcia/cs_internal.h b/drivers/pcmcia/cs_internal.h
index d6164cd..f573ea0 100644
--- a/drivers/pcmcia/cs_internal.h
+++ b/drivers/pcmcia/cs_internal.h
@@ -135,7 +135,7 @@ int pccard_get_status(struct pcmcia_sock
 struct pcmcia_callback{
 	struct module	*owner;
 	int		(*event) (struct pcmcia_socket *s, event_t event, int priority);
-	void		(*requery) (struct pcmcia_socket *s);
+	void		(*requery) (struct pcmcia_socket *s, int new_cis);
 	int		(*suspend) (struct pcmcia_socket *s);
 	int		(*resume) (struct pcmcia_socket *s);
 };
diff --git a/drivers/pcmcia/ds.c b/drivers/pcmcia/ds.c
index 0f70192..3fe4d31 100644
--- a/drivers/pcmcia/ds.c
+++ b/drivers/pcmcia/ds.c
@@ -714,22 +714,26 @@ static int pcmcia_requery(struct device
 	return 0;
 }
 
-static void pcmcia_bus_rescan(struct pcmcia_socket *skt)
+static void pcmcia_bus_rescan(struct pcmcia_socket *skt, int new_cis)
 {
-	int no_devices=0;
+	int no_devices = 0;
 	int ret = 0;
 	unsigned long flags;
 
 	/* must be called with skt_mutex held */
 	spin_lock_irqsave(&pcmcia_dev_list_lock, flags);
 	if (list_empty(&skt->devices_list))
-		no_devices=1;
+		no_devices = 1;
 	spin_unlock_irqrestore(&pcmcia_dev_list_lock, flags);
 
+	/* If this is because of a CIS override, start over */
+	if (new_cis && !no_devices)
+		pcmcia_card_remove(skt, NULL);
+
 	/* if no devices were added for this socket yet because of
 	 * missing resource information or other trouble, we need to
 	 * do this now. */
-	if (no_devices) {
+	if (no_devices || new_cis) {
 		ret = pcmcia_card_add(skt);
 		if (ret)
 			return;
diff --git a/drivers/pcmcia/socket_sysfs.c b/drivers/pcmcia/socket_sysfs.c
index 933cd86..b005602 100644
--- a/drivers/pcmcia/socket_sysfs.c
+++ b/drivers/pcmcia/socket_sysfs.c
@@ -188,7 +188,7 @@ static ssize_t pccard_store_resource(str
 	    (s->state & SOCKET_PRESENT) &&
 	    !(s->state & SOCKET_CARDBUS)) {
 		if (try_module_get(s->callback->owner)) {
-			s->callback->requery(s);
+			s->callback->requery(s, 0);
 			module_put(s->callback->owner);
 		}
 	}
@@ -325,7 +325,7 @@ static ssize_t pccard_store_cis(struct k
 	if ((s->callback) && (s->state & SOCKET_PRESENT) &&
 	    !(s->state & SOCKET_CARDBUS)) {
 		if (try_module_get(s->callback->owner)) {
-			s->callback->requery(s);
+			s->callback->requery(s, 1);
 			module_put(s->callback->owner);
 		}
 	}
-- 
1.4.3.3

Re: pcmcia: patch to fix pccard_store_cis

[Romano Giannetti]

On Fri, 2006-11-03 at 11:02 -0500, Dominik Brodowski wrote:
> Hi,
> 
> On Mon, Oct 02, 2006 at 03:01:34PM +0200, Romano Giannetti wrote:

> > 
> > https://launchpad.net/distros/ubuntu/+source/pcmciautils/+bug/52510
> > 
> > and here: 
> > 
> > http://lists.infradead.org/pipermail/linux-pcmcia/2006-August/003893.html
> > 
> > and my modem did work without IRQ problems after I got rid of .cis and
> > started (obsolete) cardmgr. Just as a data point more... 
> 
> Does it work again (after re-copying the cis file to /lib/firmware) when
> you use this patch?
> 

Will try... I have upgraded to ubuntu edgy yesterday, where I think some
thing changed (it's a 2.16.17-based kernel). I had no time to test it
more thoroughly, but the change is that now (standard edgy kernel): 
- only the first function is recognized
- if i start cardmgr, without the .cis, the two function are recognized
but now I *do* have the irq problem, and the modem is all-time-busy.

If tomorrow my little one sleeps a bit more, I will try to test your
patch. I hope I'll find how to apply it to that kernel, I do not want to
change it (for the first time my vaio suspend to ram and my  wifi works
without ndiswrapper!).

Romano 
 



-- 
Romano Giannetti <romano.giannetti@gmail.com>

Re: pcmcia: patch to fix pccard_store_cis

[Romano Giannetti]

(Sorry if you receive this message two times. Problems with the stupid
autodisclaimer signature of my institution).

>
> Does it work again (after re-copying the cis file to /lib/firmware) when
> you use this patch?
>
> Thanks,
>         Dominik
>
> From 4bb59569454f09e8bfc3a0f7bbdef46ccc7a51e0 Mon Sep 17 00:00:00 2001
> From: Dominik Brodowski <linux@dominikbrodowski.net>
> Date: Fri, 3 Nov 2006 10:54:00 -0500
> Subject: [PATCH] pcmcia: start over after CIS override
>
> When overriding the CIS, re-start the configuration of the card from
> scratch. Reported and debugged by Fabrice Bellet <fabrice@bellet.info>
>

Mr Dominik, all,

                I tried your patch on top of the kernel 2.6.17.13 that
came with ubuntu
        edgy (and had sufficient trouble with it, I do not know if I'll be able
        to try with a recent kernel.). This message is in copy to akpm which
        asked me a report on the same problem.

        I applied your patch and this one:
        http://lkml.org/lkml/2006/10/1/179

        but the kernel continues to ignore the second function of my card.
        Moreover, now the modem does not work even with the dear old cardmgr:

        [17179741.804000] pcmcia: registering new device pcmcia1.1
        [17179741.844000] 1.1: ttyS1 at I/O 0x2f8 (irq = 3) is a 16550A
        [17179741.992000] setup_irq: irq handler mismatch
        [17179741.992000]  <c0139e82> setup_irq+0x102/0x110
<c0223250> serial8250_interrupt+0x0/0xf0
        [17179741.992000]  <c0139f29> request_irq+0x99/0xb0
<c0223195> serial8250_startup+0x415/0x440
        [17179741.992000]  <c021ec58> uart_startup+0x48/0x130
<c021fa3f> uart_open+0xbf/0x410
        [17179741.992000]  <c0160a13> cdev_get+0x23/0x60  <c020eba1>
tty_open+0x161/0x310
        [17179741.992000]  <c0160ace> chrdev_open+0x6e/0x140
<c0160a60> chrdev_open+0x0/0x140
        [17179741.992000]  <c0156924> __dentry_open+0xb4/0x1e0
<c0156b05> nameidata_to_filp+0x35/0x40
        [17179741.992000]  <c0156b60> do_filp_open+0x50/0x60
<c0156805> get_unused_fd+0x45/0xb0
        [17179741.992000]  <c0156bba> do_sys_open+0x4a/0xe0
<c0156c8c> sys_open+0x1c/0x20
        [17179741.992000]  <c0102dbb> sysenter_past_esp+0x54/0x79

        Mr Morton, do you still think I should open a bug in bugzilla? Even if I
        have not tested a newer kernel? It seems to me that nothing has changed
        over there, and this _is_ a regression (from 2.6.13, even).

        Romano

        PS: can anyone point me to a tutorial on howto install a new kernael
        easily on ubuntu? The procedure they point to (creating deb package,
        installing, etc) are quite cumbersome. Anyway, I will try.

Re: pcmcia: patch to fix pccard_store_cis

[Romano Giannetti]

On Mon, 2006-11-13 at 11:12 +0100, Romano Giannetti wrote:
>                 I tried your patch on top of the kernel 2.6.17.13 that
> came with ubuntu
>         edgy (and had sufficient trouble with it, I do not know if I'll be able
>         to try with a recent kernel.). This message is in copy to akpm which
>         asked me a report on the same problem.

I tried again, applying your patch after applying Linus' one (commit 
933cd864a5c95c296844493b65d868b7cf7548aa in Linus git). Nothing. Only
the first function of the card is discovered. 

I was trying to read the code, and I cannot find where the code in
socket_sysfs.c is supposed to loop over the various functions of the
card. Maybe there could be some problem there?

I would really like to help on this. My modem is dead from 1.6.13...

Romano

Re: pcmcia: patch to fix pccard_store_cis Was: [SOLUTION/HACK/PUZZLED] pcmcia modem only works with cardmgr in recent 2.6.15 kernel.

[Andrew Morton]

On Mon, 13 Nov 2006 11:02:39 +0100
Romano Giannetti <romanol@upcomillas.es> wrote:

> 
> On Fri, 2006-11-03 at 11:02 -0500, Dominik Brodowski wrote:
> > Does it work again (after re-copying the cis file to /lib/firmware)
> > when
> > you use this patch?
> > 
> > Thanks,
> >         Dominik
> > 
> > >From 4bb59569454f09e8bfc3a0f7bbdef46ccc7a51e0 Mon Sep 17 00:00:00
> > 2001
> > From: Dominik Brodowski <linux@dominikbrodowski.net>
> > Date: Fri, 3 Nov 2006 10:54:00 -0500
> > Subject: [PATCH] pcmcia: start over after CIS override
> > 
> > When overriding the CIS, re-start the configuration of the card from
> > scratch. Reported and debugged by Fabrice Bellet
> > <fabrice@bellet.info> 
> 
> Mr Dominik, all,
> 
> 	I tried your patch on top of the kernel 2.6.17.13 that came with ubuntu
> edgy (and had sufficient trouble with it, I do not know if I'll be able
> to try with a recent kernel.). This message is in copy to akpm which
> asked me a report on the same problem.
> 
> I applied your patch and this one: 
> http://lkml.org/lkml/2006/10/1/179
> 
> but the kernel continues to ignore the second function of my card.
> Moreover, now the modem does not work even with the dear old cardmgr:
> 
> [17179741.804000] pcmcia: registering new device pcmcia1.1
> [17179741.844000] 1.1: ttyS1 at I/O 0x2f8 (irq = 3) is a 16550A
> [17179741.992000] setup_irq: irq handler mismatch
> [17179741.992000]  <c0139e82> setup_irq+0x102/0x110  <c0223250> serial8250_interrupt+0x0/0xf0
> [17179741.992000]  <c0139f29> request_irq+0x99/0xb0  <c0223195> serial8250_startup+0x415/0x440
> [17179741.992000]  <c021ec58> uart_startup+0x48/0x130  <c021fa3f> uart_open+0xbf/0x410
> [17179741.992000]  <c0160a13> cdev_get+0x23/0x60  <c020eba1> tty_open+0x161/0x310
> [17179741.992000]  <c0160ace> chrdev_open+0x6e/0x140  <c0160a60> chrdev_open+0x0/0x140
> [17179741.992000]  <c0156924> __dentry_open+0xb4/0x1e0  <c0156b05> nameidata_to_filp+0x35/0x40
> [17179741.992000]  <c0156b60> do_filp_open+0x50/0x60  <c0156805> get_unused_fd+0x45/0xb0
> [17179741.992000]  <c0156bba> do_sys_open+0x4a/0xe0  <c0156c8c> sys_open+0x1c/0x20
> [17179741.992000]  <c0102dbb> sysenter_past_esp+0x54/0x79
> 
> Mr Morton, do you still think I should open a bug in bugzilla? Even if I
> have not tested a newer kernel? It seems to me that nothing has changed
> over there, and this _is_ a regression (from 2.6.13, even).
> 

It doesn't sound like these problems will be fixed in the short term so
yes, please let's get these into bugzilla so at least they don't get
forgotten about and we have contact information filed away against
particular problems.  One record per bug, please.

I do have a debugging patch here which will give us better info about the
above IRQ-handler conflict.  I'll probably push that into 2.6.19 - we seem
to be getting a few of these lately.


> 
> PS: can anyone point me to a tutorial on howto install a new kernael
> easily on ubuntu? The procedure they point to (creating deb package,
> installing, etc) are quite cumbersome. Anyway, I will try. 

Well it sounds like fixing these bugs is a long-term project.  So it would
be fine if you were to wait until Ubuntu have a 2.6.19-based kernel
available.

But otoh I don't think Ubuntu release bleeding-edge kernel packages, so
that may be quite some time in the future.  So if you're able to work out
how to build and install a kernel.org kernel then that would help things
along a bit.  

Re: pcmcia: patch to fix pccard_store_cis Was: [SOLUTION/HACK/PUZZLED] pcmcia modem only works with cardmgr in recent 2.6.15 kernel.

[Romano Giannetti]

On Mon, 2006-11-13 at 13:54 -0800, Andrew Morton wrote:

> It doesn't sound like these problems will be fixed in the short term so
> yes, please let's get these into bugzilla so at least they don't get
> forgotten about and we have contact information filed away against
> particular problems.  One record per bug, please.

All right. Tomorrow I will try to find the time to file a bug. 

> I do have a debugging patch here which will give us better info about the
> above IRQ-handler conflict.  I'll probably push that into 2.6.19 - we seem
> to be getting a few of these lately.

But wait, I do not think this is really a relevant bug. It happens only
if I tart cardmgr, which is supposed it's not needed for newer kernels.
The problem here (as I stated in another mail in this thread) is that it
seems that the new pccardctl infrastructure miss the second function of
my card, registering and initialiazng just the first one. 


> > 
> > PS: can anyone point me to a tutorial on howto install a new kernael
> > easily on ubuntu? The procedure they point to (creating deb package,
> > installing, etc) are quite cumbersome. Anyway, I will try. 
> 
> Well it sounds like fixing these bugs is a long-term project.  So it would
> be fine if you were to wait until Ubuntu have a 2.6.19-based kernel
> available.

Uf. This will be at least six month. I really need my modem before than
that :-). I have an old external Robotics one, but well, it's a bit, you
know, big... I will try a newer kernel, I think that my problem is
simply that in ubuntu kernel all is modular, and make install do not
generate a initrd by default. 

> But otoh I don't think Ubuntu release bleeding-edge kernel packages, so
> that may be quite some time in the future.  So if you're able to work out
> how to build and install a kernel.org kernel then that would help things
> along a bit.  

Will try. If anyone can suggest a recipe before I test zillions of
reboots, it will be very welcome.

Thanks,
	Romano