On Sep 26, 2013, at 6:59 AM, DavidHalko <[email protected]> wrote:

> An attempt at migration of capability before EOF seems to be reasonable 
> planning.

That would be true *if* we knew of any active use of them.  I don't.

> 
> Are there USB interfaces on known affected platforms where USB flash drive 
> support is functional enough to provide equivalent storage capability? If so, 
> then this cares for storage concerns.
> 
> Do the systems adversely affected by the PCMCIA purge, which leverage older 
> WiFi cards, offer a USB slot where there is an upgrade path to leverage 
> working WiFi usb sticks with supported Illumos drivers for them? If WiFi USB 
> is working under Illumos, then this cares for this concern.

Generally, yes.  ZyDAS devices supported by "zyd", Atheros AR5523 supported by 
"uath", RealTek 8187 supported by "urtw", Ralink devices supported by "ural" 
and "rum", Atmel wifi supported by "atu".  So there are quite a few USB WiFi 
options.

Additionally, USB Ethernet is supported by the usbecm for compliant devices.  I 
don't know which manufacturers support the USB CDC ethernet spec offhand, but 
it is a standard spec.


> 
> A lot of portable equipment without Ethernet often have PCMCIA & CardBus 
> interfaces - so it is probably a good idea to give those systems an upgrade 
> path. I use Intel laptops with such card support running Solaris 10, but 
> don't use the slots because I could never make them functional.

Again, generally USB is better.  Cardbus 32 systems can use "afe" to get access 
to Cardbus ethernet.  The Microsoft cardbus ethernet devices are supported by 
this driver.  There may be other brands as well, I don't have a complete list 
handy, but this particular chip (ADMtek Centaur-C) is the most popular cardbus 
ethernet controller.  Its only 100 Mbps of course.   I'm not sure about gigE 
options, but anyway the PCMCIA bus maxes out at 20 Mbps (total, so 10 Mbps full 
duplex), so anyone with such an ethernet will hardly be disappointed by an 
upgrade to 100 Mbps Cardbus. :-)

> 
> I have some working SPARC platforms with an Illumos based kernel now - I 
> would be willing to buy some WiFi USB sticks to test with, if there would be 
> some support from the list. I think this is a reasonable approach 

I wouldn't worry about it *unless* you need the migration for your own path.  
As you're not using these devices, I'm not worried about it.

        - Garrett

> 
> Sent from my iPhone
> 
> On Sep 26, 2013, at 1:35 AM, Garrett D'Amore <[email protected]> wrote:
> 
>> Forwarding this to discuss -- if you care about PCMCIA (16-bit!) speak up.  
>> I'm of the opinion we'll be better off without it, but this gives anyone 
>> actually *using* this stuff a chance to make a case for saving it...
>> 
>>      - Garrett
>> 
>> Begin forwarded message:
>> 
>>> From: Garrett D'Amore <[email protected]>
>>> Subject: Proposal: EOF pcmcia bits
>>> Date: September 25, 2013 11:21:54 AM PDT
>>> To: "[email protected] Developer" <[email protected]>
>>> Cc: Bart Coddens <[email protected]>
>>> 
>>> I would like to propose (and maybe this was done previously) the final EOF 
>>> of the PCMCIA bits.  Its important to note that this proposal *only* 
>>> affects 16-bit PCMCIA.  The 32-bit Cardbus support (while also crufty) is 
>>> not included in this proposal.
>>> 
>>> The proposal specifically is for the removal of:
>>> 
>>>     * pcan - Cisco Aeronet 802.11b driver. 
>>>     * pcwl - Intersil Prism 802.11b driver
>>>     * pcata/pcide and friends
>>>     * sys/pccard.h
>>>     * csx_ card CardServices kernel routines
>>>     * pcs place holder driver
>>> 
>>> Further, it would be impossible to support 16-bit PCMCIA on illumos after 
>>> this proposal integrates.
>>> 
>>> A bit more detail.
>>> 
>>> 1. PCMCIA is 16-bit only
>>> 2. PCMCIA does not support DMA, so all these devices use incredibly slow PIO
>>> 3. pcan in theory also supports a PCI card, but I've never seen one in the 
>>> wild.  It was never a very popular device, because they were pretty 
>>> expensive compared to the Prism cards.
>>> 4. pcwl devices *were* popular (probably the most common 802.11b PC card.) 
>>> Additionally the driver "supports" a miniPCI variant (PRISM 2.5).  However, 
>>> the miniPCI variant has a nasty erratum which can hard hang easily, and the 
>>> pcwl driver lacks the fix for this.  (I have some information on fixing it, 
>>> but I never got around to getting this work done for the open pcwl driver 
>>> as I didn't get approval from Intersil or Tadpole to open source the 
>>> equivalent work I did for the Tadpole driver for the same chip.)
>>> 5. Neither pcan nor pcwl support modern WPA (never mind WPA2).  This makes 
>>> them mostly useless for anything except wide open wireless networks.
>>> 6. pcata and friends are for PCMCIA memory cards, the most common variant 
>>> of which was CF.  However, all modern systems that still have CF support do 
>>> so over much faster IDE, SATA, or USB interfaces.   The only exceptions to 
>>> these are certain embedded platforms (e.g. MIPS Alchemy boards) that have 
>>> no chance of ever running illumos.
>>> 
>>> The main motivation for this change is general housekeeping -- this code is 
>>> thoroughly bitrotten and I don't know of anyone who's used any of this 
>>> stuff in the last several years.  (I was probably one of the last users on 
>>> certain SPARC hardware, but that was about 5 years ago now, I guess, and I 
>>> removed the SPARCLE platform support from illumos years ago.)  The other 
>>> thing is that these interfaces -- the pcmcia nexus in particular, make use 
>>> of some other legacy kernel interfaces, which we would like to clean up.  
>>> Removing the pcmcia nexus support will allow the existing cardbus stack to 
>>> be cleaned up, and will allow some of the underlying legacy interfaces to 
>>> also be cleaned up and simplified.  (In particular, the interfaces most 
>>> directly benefiting from this will be the legacy power management 
>>> interfaces;   see illumos bug 680.
>>> 
>>> So, with that goal stated, please indicate any specific reasons for 
>>> objecting to this change, with supporting details.  Such reasons might 
>>> indicate current or planned use of these interfaces (seems unlikely), or 
>>> any reason(s) that anyone knows of specific risks.
>>> 
>>> As a follow up, I'd also like to hear (more as a matter of a survey than 
>>> anything else) if anyone out there is using CardBus.  My guess is that even 
>>> that has fallen by the way side, having been thoroughly supplanted by 
>>> ExpressCard and USB.  If it turns out that even CardBus is of little or not 
>>> use, I'd be happy to nuke that as well -- frankly it would make the task 
>>> easier if I don't have to extricate cardbus and could just nuke the entire 
>>> pcic nexus. :-)
>>> 
>>> Thanks.
>>> 
>>>     - Garrett
>>> 
>>> 
>> 
>> illumos-discuss | Archives  | Modify Your Subscription        
> 
> illumos-discuss | Archives  | Modify Your Subscription         




-------------------------------------------
illumos-discuss
Archives: https://www.listbox.com/member/archive/182180/=now
RSS Feed: https://www.listbox.com/member/archive/rss/182180/21175430-2e6923be
Modify Your Subscription: 
https://www.listbox.com/member/?member_id=21175430&id_secret=21175430-6a77cda4
Powered by Listbox: http://www.listbox.com

Reply via email to