Message-ID: <4714D700.7060008@rtr.ca>
Date: Tue, 16 Oct 2007 11:21:36 -0400
From: Mark Lord <lkml@rtr.ca>
User-Agent: Thunderbird 2.0.0.6 (X11/20070728)
MIME-Version: 1.0
To: kristen.c.accardi@intel.com, pcihpd-discuss@lists.sourceforge.net,
       Linux Kernel <linux-kernel@vger.kernel.org>
Subject: Re: PCIe Hotplug:  NFG unless I boot with card already inserted.
References: <4714C0A6.1030204@rtr.ca>
In-Reply-To: <4714C0A6.1030204@rtr.ca>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 2418
Lines: 55

Mark Lord wrote:
> I have a Dell notebook with an PCIe ExpressCard slot.
> I also have a PCIe ExpressCard SATA controller (uses sata_sil24 driver).
> 
> I would like to be able to hot plug/unplug the controller card at will.
> But alas, Linux doesn't cope with it *unless* I boot the kernel with
> the card initially inserted.
> 
> 1. Booting Linux kernel (latest 2.6.23) without the card inserted
> means that the card will never be detected, regardless of how many
> times subsequently the card is inserted/removed/whatever.
> 
> 2. Booting Linux kernel *with* the card inserted means that it is
> detected and used, and can be unplugged/replugged as I please,
> with intervening suspend/resume (RAM or disk) cycles not interfering.
> 
> 3. Booting Linux kernel without the card inserted, and then doing
> a suspend-to-disk poweroff, inserting the card, and powering on again,
> the card's BIOS extension runs as normal.  But on resume from the
> suspend-to-disk, the running kernel again never sees the card,
> even after removing/reinserting/whatever.
> 
> 4. All of this leads me to believe that the kernel must be doing some
> kind of once-only scan of hardware at boot time, and never repeating
> it afterwards.  Loading/unloading all of the PCI/PCIe hotplug stuff
> has no effect on this, so it must be broken elsewhere.
> 
> 5. It is not likely to be a BIOS thing, because it still fails on
> power-on (with card inserted) after a suspend-to-disk, which appears
> to the BIOS exactly the same as any other power-on.
> 
> 6. But it's probably a "kernel relies on BIOS data structure read
> at boot time" issue, based on the observations above.

Actually, I must now take back some of that.

Most of these tests were done a month or two ago.
With 2.6.23.1 running, I just now redid all of the tests.

Now it seems that pciehp fails to notice a newly inserted card
only after a suspend/resume cycle with the slot empty.

I can now get it to work again by just doing:
	1. remove the card, so the slot is empty.
	2. rmmod pciehp; modprobe pciehp
	3. insert the card again -- it works!

So we just need to fix an issue or two with suspend/resume (RAM) in pciehp.

Cheers
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/