Driver Problems? Questions? Issues?
Put OSR's experience to work for you! Contact us for assistance with:
  • Creating the right design for your requirements
  • Reviewing your existing driver code
  • Analyzing driver reliability/performance issues
  • Custom training mixed with consulting and focused directly on your specific areas of interest/concern.
Check us out. OSR, the Windows driver experts.

Monthly Seminars at OSR Headquarters

East Coast USA
Windows Internals and SW Drivers, Dulles (Sterling) VA, 13 November 2017

Kernel Debugging & Crash Analysis for Windows, Nashua (Amherst) NH, 4 December 2017

Writing WDF Drivers I: Core Concepts, Nashua (Amherst) NH, 8 January 2018

WDF Drivers II: Advanced Implementation Techniques, Nashua (Amherst) NH, 15 January 2018


Go Back   OSR Online Lists > ntdev
Welcome, Guest
You must login to post to this list
  Message 1 of 71  
05 Oct 17 10:30
Peter Viscarola (OSR)
xxxxxx@osr.com
Join Date:
Posts To This List: 5949
List Moderator
Driver Developer Conference

We REALLY think it's time that Microsoft held a Driver Developer Conference. What do YOU think? PLEASE TAKE 30 SECONDS TO VOTE in our (anonymous) 6 question online poll: <https://www.osr.com/microsoft-hold-conference-driver-developers/> You can read a bit more about what we're thinking on this topic here: <https://www.osr.com/blog/2017/10/03/want-driver-developer-conference/> Peter OSR @OSRDrivers
  Message 2 of 71  
05 Oct 17 14:06
Tim Roberts
xxxxxx@probo.com
Join Date: 28 Jan 2005
Posts To This List: 11622
Driver Developer Conference

xxxxx@osr.com wrote: > We REALLY think it's time that Microsoft held a Driver Developer Conference. > > What do YOU think? There are two fundamentally incompatible needs at work here. Twenty years ago, I went to Driver Developer Conferences because there was so much I didn't know.  I needed to learn basics.  There's still a need for that, and I strongly suspect that's what those East Asian DDCs were all about, but those are sessions that I would not attend today. Today, I basically want to know what has changed.  It seems like a trivial point, but I would be happy if I could find a huge "Release Notes" document for each release of Windows and each release of the WDK that enumerated what has changed since the last release.  In the 20th Century, release notes were shipped with the Windows CD.  Today, I cannot find such a thing.  You can go to the Linux kernel web sites and find a document that summarizes every change in every release of the kernel.  That's just incredibly useful. -- Tim Roberts, xxxxx@probo.com Providenza & Boekelheide, Inc.
  Message 3 of 71  
05 Oct 17 15:14
Peter Viscarola (OSR)
xxxxxx@osr.com
Join Date:
Posts To This List: 5949
List Moderator
Driver Developer Conference

<quote> I basically want to know what has changed. It seems like a trivial point, but I would be happy if I could find a huge "Release Notes" document for each release of Windows and each release of the WDK that enumerated what has changed since the last release. </quote> I certainly agree. But I think the complexity is beyond the level of "release notes" -- I mean, what will they put in the Release Notes about Universal Drivers that'll get the point across (something akin to what I wrote in this issue of The NT Insider on this topic)? How would they POSSIBLY communicate the level of architectural changes we've seen in the OS of interest to driver devs since Windows 8 in Release Notes? PoFx, for example. Or, how drivers can get resources from multiple separate controllers (none of which may be the bus driver that enumerated them). Or all the stuff about connected/modern standby. Then there's the "inside info" shit that you care about. That SDV stopped catching many errors in VS 2015, but there's a new analysis engine that's active in RS3. THAT sort of thing. Yeah... I just don't feel like release notes would could come close to doing it. Even assuming I can "read the source." Peter OSR @OSRDrivers
  Message 4 of 71  
05 Oct 17 15:25
Don Burn
xxxxxx@windrvr.com
Join Date: 23 Feb 2011
Posts To This List: 1349
Driver Developer Conference

Having gone to all the Driver Developer Conferences and around 10 WinHEC's a lot of the time I ended up going to a presentation since it was in an area of interest. What made it good was in most cases the presenter had some nugget of information I missed (from the examples and the docs), or had a different approach to developing/coding a driver for some case that I had not thought of. To me these small items made the trip worthwhile. Of course also being able to interact with other driver dev's and the folks from Redmond was important on the conference. I don't know about others, but about half my queries are from firms that are looking to withdraw software or in some cases software and hardware development from their current outsourced approach in Asia. I'm fine with developing a driver for the older platforms, but these folks want the full breath from Windows 7 to a universal driver, and I know I could use some more data. Don Burn Windows Driver Consulting Website: http://www.windrvr.com -----Original Message----- From: xxxxx@lists.osr.com [mailto:xxxxx@lists.osr.com] On Behalf Of xxxxx@osr.com Sent: Thursday, October 05, 2017 3:15 PM To: Windows System Software Devs Interest List <xxxxx@lists.osr.com> Subject: RE:[ntdev] Driver Developer Conference <quote> I basically want to know what has changed. It seems like a trivial point, but I would be happy if I could find a huge "Release Notes" document for each release of Windows and each release of the WDK that enumerated what has changed since the last release. </quote> I certainly agree. But I think the complexity is beyond the level of "release notes" -- I mean, what will they put in the Release Notes about Universal Drivers that'll get the point across (something akin to what I wrote in this issue of The NT Insider on this topic)? How would they POSSIBLY communicate the level of architectural changes we've seen in the OS of interest to driver devs since Windows 8 in Release Notes? PoFx, for example. Or, how drivers can get resources from multiple separate controllers (none of which may be the bus driver that enumerated them). Or all the stuff about connected/modern standby. Then there's the "inside info" shit that you care about. That SDV stopped catching many errors in VS 2015, but there's a new analysis engine that's active in RS3. THAT sort of thing. Yeah... I just don't feel like release notes would could come close to doing it. Even assuming I can "read the source." Peter OSR @OSRDrivers --- NTDEV is sponsored by OSR Visit the list online at: <http://www.osronline.com/showlists.cfm?list=ntdev> MONTHLY seminars on crash dump analysis, WDF, Windows internals and software drivers! Details at <http://www.osr.com/seminars> To unsubscribe, visit the List Server section of OSR Online at <http://www.osronline.com/page.cfm?name=ListServer>
  Message 5 of 71  
05 Oct 17 15:35
R0b0t1
xxxxxx@gmail.com
Join Date: 24 Mar 2017
Posts To This List: 53
Driver Developer Conference

> What made it good was in most cases the presenter had some > nugget of information I missed (from the examples and the docs), or had a > different approach to developing/coding a driver for some case that I had > not thought of. I am trying to develop kernel mode software for Windows, but lots of the information I need is "hidden." It seems like some users here got their start by talking to other developers. I also tried looking for release notes and never found any. The one problem I have with such conferences is they serve as gates to information. In some cases what was discussed never gets put up anywhere else. It is unlikely I would ever be able to attend a conference; companies typically avoid the Midwest and I can not afford travel. Cheers, R0b0t1 On Thu, Oct 5, 2017 at 2:25 PM, xxxxx@windrvr.com <xxxxx@lists.osr.com> wrote: > Having gone to all the Driver Developer Conferences and around 10 WinHEC's a > lot of the time I ended up going to a presentation since it was in an area > of interest. What made it good was in most cases the presenter had some > nugget of information I missed (from the examples and the docs), or had a > different approach to developing/coding a driver for some case that I had > not thought of. To me these small items made the trip worthwhile. Of > course also being able to interact with other driver dev's and the folks > from Redmond was important on the conference. > > I don't know about others, but about half my queries are from firms that are <...excess quoted lines suppressed...>
  Message 6 of 71  
05 Oct 17 15:40
Don Burn
xxxxxx@windrvr.com
Join Date: 23 Feb 2011
Posts To This List: 1349
Driver Developer Conference

Actually, Microsoft did put the slides up for everyone. Also, there = was a decent amount of discussion of things on this list when something = new came out, along with the NT Insider having articles. Don Burn Windows Driver Consulting Website: http://www.windrvr.com=20 -----Original Message----- From: xxxxx@lists.osr.com = [mailto:xxxxx@lists.osr.com] On Behalf Of = xxxxx@gmail.com Sent: Thursday, October 05, 2017 3:34 PM To: Windows System Software Devs Interest List <xxxxx@lists.osr.com> Subject: Re: [ntdev] Driver Developer Conference > What made it good was in most cases the presenter had some nugget of=20 > information I missed (from the examples and the docs), or had a=20 > different approach to developing/coding a driver for some case that I=20 > had not thought of. I am trying to develop kernel mode software for Windows, but lots of the = information I need is "hidden." It seems like some users here got their = start by talking to other developers. I also tried looking for release = notes and never found any. The one problem I have with such conferences is they serve as gates to = information. In some cases what was discussed never gets put up anywhere = else. It is unlikely I would ever be able to attend a conference; = companies typically avoid the Midwest and I can not afford travel. Cheers, R0b0t1 On Thu, Oct 5, 2017 at 2:25 PM, xxxxx@windrvr.com <xxxxx@lists.osr.com> = wrote: > Having gone to all the Driver Developer Conferences and around 10=20 > WinHEC's a lot of the time I ended up going to a presentation since it = > was in an area of interest. What made it good was in most cases the=20 > presenter had some nugget of information I missed (from the examples=20 > and the docs), or had a different approach to developing/coding a = driver for some case that I had > not thought of. To me these small items made the trip worthwhile. = Of > course also being able to interact with other driver dev's and the=20 > folks from Redmond was important on the conference. > > I don't know about others, but about half my queries are from firms=20 > that are looking to withdraw software or in some cases software and = hardware > development from their current outsourced approach in Asia. I'm fine = with > developing a driver for the older platforms, but these folks want the=20 > full breath from Windows 7 to a universal driver, and I know I could=20 > use some more data. > > > Don Burn > Windows Driver Consulting > Website: http://www.windrvr.com > > <...excess quoted lines suppressed...> release. > </quote> > > I certainly agree. > > But I think the complexity is beyond the level of "release notes" -- I = > mean, what will they put in the Release Notes about Universal Drivers=20 > that'll get the point across (something akin to what I wrote in this=20 > issue of The NT Insider on this topic)? How would they POSSIBLY=20 > communicate the level of architectural changes we've seen in the OS of = > interest to driver devs since Windows 8 in Release Notes? PoFx, for=20 software drivers! > Details at <http://www.osr.com/seminars> > > To unsubscribe, visit the List Server section of OSR Online at=20 > <http://www.osronline.com/page.cfm?name=3DListServer> --- NTDEV is sponsored by OSR Visit the list online at: = <http://www.osronline.com/showlists.cfm?list=3Dntdev> MONTHLY seminars on crash dump analysis, WDF, Windows internals and = software drivers! Details at <http://www.osr.com/seminars> To unsubscribe, visit the List Server section of OSR Online at = <http://www.osronline.com/page.cfm?name=3DListServer>
  Message 7 of 71  
05 Oct 17 16:00
Peter Viscarola (OSR)
xxxxxx@osr.com
Join Date:
Posts To This List: 5949
List Moderator
Driver Developer Conference

The slides (and videos) of the presentations from AT LEAST the past two WinHECs are available online. That your company won't pay for you to travel to (let's just say Seattle) to WinHEC is pretty sad. The airfare is about $200 from Chicago to SEA. There are hotels that *I* would stay in (not PREFER to stay in, mind you) that run about $100/night. So, for the TRAVEL we're talking pretty short money... certainly less than $1K. I suspect the cost of actually attending the conference would be higher than that, sadly. Peter OSR @OSRDrivers
  Message 8 of 71  
05 Oct 17 17:00
Tim Roberts
xxxxxx@probo.com
Join Date: 28 Jan 2005
Posts To This List: 11622
Driver Developer Conference

xxxxx@windrvr.com wrote: > ...I'm fine with > developing a driver for the older platforms, but these folks want the full > breath from Windows 7 to a universal driver, and I know I could use some > more data. That's interesting; I have never had a single client ask me about universal driver support. -- Tim Roberts, xxxxx@probo.com Providenza & Boekelheide, Inc.
  Message 9 of 71  
05 Oct 17 18:22
Peter Viscarola (OSR)
xxxxxx@osr.com
Join Date:
Posts To This List: 5949
List Moderator
Driver Developer Conference

<quote> I have never had a single client ask me about universal driver support. </quote> We?ve had several misguided requests, where the client referred to Google to write their requirements and ?Universal? sounded like something good. We?ve had one, genuine, proper, request to make a driver universal. It as for an SMB device that legitimately can appear on tablets, phones, IoT devices or laptops. It?s a rare thing, though. Peter OSR @OSRDrivers
  Message 10 of 71  
05 Oct 17 20:11
R0b0t1
xxxxxx@gmail.com
Join Date: 24 Mar 2017
Posts To This List: 53
Driver Developer Conference

To you an Don: I will go look for those. I was more referring to the unpresented material, and being able to talk to people in the field. For the work I am trying to do I am not associated with a company. On Thu, Oct 5, 2017 at 3:01 PM, xxxxx@osr.com <xxxxx@lists.osr.com> wrote= : > The slides (and videos) of the presentations from AT LEAST the past two W= inHECs are available online. > > That your company won't pay for you to travel to (let's just say Seattle)= to WinHEC is pretty sad. The airfare is about $200 from Chicago to SEA. = There are hotels that *I* would stay in (not PREFER to stay in, mind you) t= hat run about $100/night. So, for the TRAVEL we're talking pretty short mo= ney... certainly less than $1K. I suspect the cost of actually attending t= he conference would be higher than that, sadly. > > Peter > OSR > @OSRDrivers > > > --- > NTDEV is sponsored by OSR > > Visit the list online at: <http://www.osronline.com/showlists.cfm?list=3D= ntdev> > > MONTHLY seminars on crash dump analysis, WDF, Windows internals and softw= are drivers! > Details at <http://www.osr.com/seminars> > > To unsubscribe, visit the List Server section of OSR Online at <http://ww= w.osronline.com/page.cfm?name=3DListServer>
  Message 11 of 71  
05 Oct 17 20:58
Don Burn
xxxxxx@windrvr.com
Join Date: 23 Feb 2011
Posts To This List: 1349
Driver Developer Conference

A lot of my clients are doing industrial like things, for example laser = marking, stress testing, various digital controls, etc. Most of them = are thinking along the lines of moving their platform from a full PC, to = an IOT type device. They aren't sure they are going there, but that is = why they ask about universal. Don Burn Windows Driver Consulting Website: http://www.windrvr.com=20 -----Original Message----- From: xxxxx@lists.osr.com = [mailto:xxxxx@lists.osr.com] On Behalf Of xxxxx@probo.com Sent: Thursday, October 05, 2017 5:00 PM To: Windows System Software Devs Interest List <xxxxx@lists.osr.com> Subject: Re: [ntdev] Driver Developer Conference xxxxx@windrvr.com wrote: > ...I'm fine with > developing a driver for the older platforms, but these folks want the=20 > full breath from Windows 7 to a universal driver, and I know I could=20 > use some more data. That's interesting; I have never had a single client ask me about = universal driver support. -- Tim Roberts, xxxxx@probo.com Providenza & Boekelheide, Inc. --- NTDEV is sponsored by OSR Visit the list online at: = <http://www.osronline.com/showlists.cfm?list=3Dntdev> MONTHLY seminars on crash dump analysis, WDF, Windows internals and = software drivers! Details at <http://www.osr.com/seminars> To unsubscribe, visit the List Server section of OSR Online at = <http://www.osronline.com/page.cfm?name=3DListServer>
  Message 12 of 71  
06 Oct 17 06:03
Shaarang Tyagi
xxxxxx@gmail.com
Join Date: 08 Mar 2017
Posts To This List: 6
Driver Developer Conference

Conferences are fine but I was wondering whats the future of painstakingly learning such a complex thing when world is moving very fast to the mobile space and most of the things are now online web apps, I guess only some hardware creators would need to write drivers in the future, I am just guessing I don't know how much business the driver development consulting companies right now have? And has the business also started seeing a dip? Or it is sure to see one!? On Fri, Oct 6, 2017 at 6:27 AM, xxxxx@windrvr.com <xxxxx@lists.osr.com> wrote: > A lot of my clients are doing industrial like things, for example laser > marking, stress testing, various digital controls, etc. Most of them are > thinking along the lines of moving their platform from a full PC, to an IOT > type device. They aren't sure they are going there, but that is why they > ask about universal. > > > Don Burn > Windows Driver Consulting > Website: http://www.windrvr.com <...excess quoted lines suppressed...> --
  Message 13 of 71  
06 Oct 17 07:41
ashish kohli
xxxxxx@hotmail.com
Join Date: 29 Nov 2014
Posts To This List: 42
Driver Developer Conference

Ya that is my question as well. If someone can throw some light. what will be future of windows device drivers like file system filters and network filtering along with Win32.Will it remain relevant for say next 10 years.Is new software being created using these technologies. My concern is: --> World moving to mobile space --> Desktop and server world also seeing microsoft share being taken by mac/linux
  Message 14 of 71  
06 Oct 17 08:45
Don Burn
xxxxxx@windrvr.com
Join Date: 23 Feb 2011
Posts To This List: 1349
Driver Developer Conference

I don't think anyone can say for sure on this issue. I do know that the same questions you asked were being asked 10 years ago. I can also say, that I've dealt with a fair number of startups that developed a driver for Linux only to have their venture capitalist backers say "but you need to support Windows". I think a lot of this depends on where Microsoft goes with Windows. I do a lot of weird devices that don't fit any normal device class, for me the question is how will Windows evolve? I have clients who absolutely will not use Windows 10, because the release cycle means they find an expensive failure in their product with every Win10 release (since something changed in Windows). By expensive I mean situations where parts costing $50,000 to $250,000 a piece are wasted with the new release, and if they are lucky only one! I have clients that reject the universal driver model, because they need a separate device class. The example above was told by Microsoft that they should just make their device a "printer". Can you imagine the costs if a secretary choose the wrong selection in a print dialog and spewed out a few dozen $250,000 parts badly, while trying to print a letter? The Microsoft "experts" did not get it! So to me the question can't be answered since we don't know what Redmond is thinking. This is another reason for a conference, 15 years ago Microsoft really interacted with the developer base. They would tell us where they planned to go, and we explained problems we had. Not saying things were perfect, but at least there was good interaction, something that best happens face to face. Don Burn Windows Driver Consulting Website: http://www.windrvr.com -----Original Message----- From: xxxxx@lists.osr.com [mailto:xxxxx@lists.osr.com] On Behalf Of xxxxx@hotmail.com Sent: Friday, October 06, 2017 7:43 AM To: Windows System Software Devs Interest List <xxxxx@lists.osr.com> Subject: RE:[ntdev] Driver Developer Conference Ya that is my question as well. If someone can throw some light. what will be future of windows device drivers like file system filters and network filtering along with Win32.Will it remain relevant for say next 10 years.Is new software being created using these technologies. My concern is: --> World moving to mobile space --> Desktop and server world also seeing microsoft share being taken by mac/linux --- NTDEV is sponsored by OSR Visit the list online at: <http://www.osronline.com/showlists.cfm?list=ntdev> MONTHLY seminars on crash dump analysis, WDF, Windows internals and software drivers! Details at <http://www.osr.com/seminars> To unsubscribe, visit the List Server section of OSR Online at <http://www.osronline.com/page.cfm?name=ListServer>
  Message 15 of 71  
06 Oct 17 09:47
Peter Viscarola (OSR)
xxxxxx@osr.com
Join Date:
Posts To This List: 5949
List Moderator
Driver Developer Conference

<quote> moving their platform from a full PC, to an IOT type device. They aren't sure they are going there, but that is why they ask about universal. </quote> I'll venture they're not moving from a PC to IoT Core. In that case, they'll be moving to IoT Enterprise, which is the "desktop" driver model. <quote> what will be future of windows device drivers like file system filters and network filtering along with Win32? Will it remain relevant for say next 10 years. Is new software being created using these technologies. </quote> So... let's see: Yes, there is new software, and lots of it, being created using these technologies. And I will personally guarantee you that Windows will remain "relevant" for the next 10 years. There's just too much corporate line-of-business use, too much inertia in the installed base, for Windows to disappear in that time-frame. Most Corporate systems are still Windows. Most big industrial control systems are based on Windows. Most medical systems are based on Windows. So, yeah... There'll be lots of drivers written for new devices, lots of file system filters, lots of "stuff" done on Windows in the next 10 years. But would I *completely and exclusively* bet on Windows, the way I did in, say 1999? No, no way. For proof, see: <https://www.osr.com/fesf-linux/> The *real* question (and what I think you're actually asking) is "Will Windows remain a vital, evolving, developing force in computing with a growing user base for the next 10 years? And if not... what is?" And, that, sir is the multi-billion dollar question. If you could answer it, you'd stand the chance to become very wealthy indeed. <quote> I have clients that reject the universal driver model, because they need a separate device class. </quote> This times 100. The fact that the Universal INF model does not allow ClassInstall32 sections is a BIG fucking problem. I can't BELIEVE our MSFT friends do not see this. I *get* that they want "declarative" INF files. But there's no reason that having a ClassInstall32 section that INCLUDES the Class GUID can't be "declarative." What annoys me about these little things is that they are PRECISELY the types of problems that close interaction with the third party driver development community would avoid. If they had a conference and discussed this plan, we would have been able to give them feedback that not allowing custom device classes was a VERY bad idea. Peter OSR @OSRDrivers
  Message 16 of 71  
06 Oct 17 11:02
Mark Roddy
xxxxxx@gmail.com
Join Date: 25 Feb 2000
Posts To This List: 4012
Driver Developer Conference

Maybe this should be billed as The Last Driver Dev Con Ever. Then five years later we can have another one. Mark Roddy On Fri, Oct 6, 2017 at 9:47 AM, xxxxx@osr.com <xxxxx@lists.osr.com> wrote: > <quote> > moving their platform from a full PC, to an IOT type device. > They aren't sure they are going there, but that is why they ask about > universal. > </quote> > > I'll venture they're not moving from a PC to IoT Core. In that case, > they'll be moving to IoT Enterprise, which is the "desktop" driver model. > > <quote> <...excess quoted lines suppressed...> --
  Message 17 of 71  
06 Oct 17 11:48
Shaarang Tyagi
xxxxxx@gmail.com
Join Date: 08 Mar 2017
Posts To This List: 6
Driver Developer Conference

Well, the truth is also that I have never seen any OS as robust as windows! Certainly there maybe problems but for normal end users or admins I think that windows is best and it's really hard to beat it ! This is true for desktop but for mobile it made entry a bit too late! On Oct 6, 2017 8:31 PM, "xxxxx@gmail.com" <xxxxx@lists.osr.com> wrote: > Maybe this should be billed as The Last Driver Dev Con Ever. > > Then five years later we can have another one. > > Mark Roddy > > On Fri, Oct 6, 2017 at 9:47 AM, xxxxx@osr.com <xxxxx@lists.osr.com> > wrote: > >> <quote> <...excess quoted lines suppressed...> --
  Message 18 of 71  
06 Oct 17 12:21
Don Burn
xxxxxx@windrvr.com
Join Date: 23 Feb 2011
Posts To This List: 1349
Driver Developer Conference

" Well, the truth is also that I have never seen any OS as robust as = windows! " That is a sad commentary on the industry. Windows and Linux both are = pretty good operating systems, but a large number of the mini-computer = systems and a fair number of Unix systems could beat them on = reliability. That does not even consider, the fault tolerant computer = firms like Tandem, Stratus and Sequoia (I and Mark Roddy both worked for = Sequoia) who measured up time in years.=20 My favorite situation was about 10 years ago, when I got a call asking = if I remember how to reboot a Sequoia computer. It turns out a = customer of Sequoia installed a system in 1990, and it had not gone down = in over 15 years. The firm lost their lease and needed help shutting = down and rebooting the computer! Don Burn Windows Driver Consulting Website: http://www.windrvr.com=20
  Message 19 of 71  
06 Oct 17 12:59
Gregory G. Dyess
xxxxxx@pdq.net
Join Date:
Posts To This List: 356
Driver Developer Conference

That sounds very much like VMS and VAXClusters. Even major OS upgrades could be done without every completely shutting down the system. The nodes in the cluster could be separated by many miles and provide full fault-tolerant failover capabilities. Windows and Linux still don't even come close to this level of fault tolerance. Don't get me wrong, I've developed drivers and applications for all three and like (and hate) all three. As for the relevance of MS Windows in 10 years, who knows? Microslop could decide to create an all new OS and try to force people to move to it. Some say Windows 10 was exactly that with it's forced update policy. Anyone who thinks Windows won't abandon a platform should just look to Windows CE/Windows Embedded Compact. People who used to make a living as a consultant for WinCE/WEC are all horrified to see Microsoft completely abandon that platform in favor of the "Unified Driver Model" which is completely inappropriate for the Embedded/RealTime world. They basically gave away that market to the much mode configurable and lighter-weight Linux for embedded work. I finally gave up and went to work for Arm. Greg --- xxxxx@lists.osr.com wrote: From: "xxxxx@windrvr.com" <xxxxx@lists.osr.com> To: "Windows System Software Devs Interest List" <xxxxx@lists.osr.com> Subject: RE: [ntdev] Driver Developer Conference Date: Fri, 6 Oct 2017 12:20:43 -0400 " Well, the truth is also that I have never seen any OS as robust as windows! " That is a sad commentary on the industry. Windows and Linux both are pretty good operating systems, but a large number of the mini-computer systems and a fair number of Unix systems could beat them on reliability. That does not even consider, the fault tolerant computer firms like Tandem, Stratus and Sequoia (I and Mark Roddy both worked for Sequoia) who measured up time in years. My favorite situation was about 10 years ago, when I got a call asking if I remember how to reboot a Sequoia computer. It turns out a customer of Sequoia installed a system in 1990, and it had not gone down in over 15 years. The firm lost their lease and needed help shutting down and rebooting the computer! Don Burn Windows Driver Consulting Website: http://www.windrvr.com --- NTDEV is sponsored by OSR Visit the list online at: <http://www.osronline.com/showlists.cfm?list=ntdev> MONTHLY seminars on crash dump analysis, WDF, Windows internals and software drivers! Details at <http://www.osr.com/seminars> To unsubscribe, visit the List Server section of OSR Online at <http://www.osronline.com/page.cfm?name=ListServer>
  Message 20 of 71  
06 Oct 17 13:07
Don Burn
xxxxxx@windrvr.com
Join Date: 23 Feb 2011
Posts To This List: 1349
Driver Developer Conference

Actually, VAXClusters were a lot less robust than a full fault tolerant = system. I remember a couple of times the DEC sales guy's thought they = could bid for contracts that the FT firms were going for, and then = complained that no one could meet the requirements. Most of the time = all the FT firms could. Unfortunately, Microsoft and others believed the cluster claims of = reliability, while clusters are good, they are not at the same level of = reliability. Don Burn Windows Driver Consulting Website: http://www.windrvr.com=20 -----Original Message----- From: xxxxx@lists.osr.com = [mailto:xxxxx@lists.osr.com] On Behalf Of xxxxx@pdq.net Sent: Friday, October 06, 2017 12:59 PM To: Windows System Software Devs Interest List <xxxxx@lists.osr.com> Cc: xxxxx@lists.osr.com Subject: RE: [ntdev] Driver Developer Conference That sounds very much like VMS and VAXClusters. Even major OS upgrades = could be done without every completely shutting down the system. The = nodes in the cluster could be separated by many miles and provide full = fault-tolerant failover capabilities. =20 Windows and Linux still don't even come close to this level of fault = tolerance. Don't get me wrong, I've developed drivers and applications = for all three and like (and hate) all three. As for the relevance of MS Windows in 10 years, who knows? Microslop = could decide to create an all new OS and try to force people to move to = it. Some say Windows 10 was exactly that with it's forced update = policy. Anyone who thinks Windows won't abandon a platform should just = look to Windows CE/Windows Embedded Compact. People who used to make a = living as a consultant for WinCE/WEC are all horrified to see Microsoft = completely abandon that platform in favor of the "Unified Driver Model" = which is completely inappropriate for the Embedded/RealTime world. They = basically gave away that market to the much mode configurable and = lighter-weight Linux for embedded work. I finally gave up and went to work for Arm. =20 Greg --- xxxxx@lists.osr.com wrote: From: "xxxxx@windrvr.com" <xxxxx@lists.osr.com> To: "Windows System Software Devs Interest List" <xxxxx@lists.osr.com> Subject: RE: [ntdev] Driver Developer Conference Date: Fri, 6 Oct 2017 12:20:43 -0400 " Well, the truth is also that I have never seen any OS as robust as = windows! " That is a sad commentary on the industry. Windows and Linux both are = pretty good operating systems, but a large number of the mini-computer = systems and a fair number of Unix systems could beat them on = reliability. That does not even consider, the fault tolerant computer = firms like Tandem, Stratus and Sequoia (I and Mark Roddy both worked for = Sequoia) who measured up time in years.=20 My favorite situation was about 10 years ago, when I got a call asking = if I remember how to reboot a Sequoia computer. It turns out a = customer of Sequoia installed a system in 1990, and it had not gone down = in over 15 years. The firm lost their lease and needed help shutting = down and rebooting the computer! Don Burn Windows Driver Consulting Website: http://www.windrvr.com=20 --- NTDEV is sponsored by OSR Visit the list online at: = <http://www.osronline.com/showlists.cfm?list=3Dntdev> MONTHLY seminars on crash dump analysis, WDF, Windows internals and = software drivers! Details at <http://www.osr.com/seminars> To unsubscribe, visit the List Server section of OSR Online at = <http://www.osronline.com/page.cfm?name=3DListServer> --- NTDEV is sponsored by OSR Visit the list online at: = <http://www.osronline.com/showlists.cfm?list=3Dntdev> MONTHLY seminars on crash dump analysis, WDF, Windows internals and = software drivers! Details at <http://www.osr.com/seminars> To unsubscribe, visit the List Server section of OSR Online at = <http://www.osronline.com/page.cfm?name=3DListServer>
  Message 21 of 71  
06 Oct 17 13:12
Erik Dabrowsky
xxxxxx@laservault.com
Join Date: 29 Aug 2016
Posts To This List: 8
Driver Developer Conference

About two years ago we had a customer call us wanting to upgrade their system. They were running our legacy report archiving software on a Windows 95 machine that had been installed in the late 90s! I couldn't believe the hardware had even lasted that long. It was one of those "If it's not broke don't fix it!" situations. Erik
  Message 22 of 71  
06 Oct 17 14:00
Peter Viscarola (OSR)
xxxxxx@osr.com
Join Date:
Posts To This List: 5949
List Moderator
Driver Developer Conference

<quote> Anyone who thinks Windows won't abandon a platform should just look to Windows CE/Windows Embedded Compact. </quote> Seriously? You can't equate support for a niche operating system with support for one that runs more than 90% of the computers in the world (that's over a BILLION computers). Windows CE was a good operating system. There were some very impressive people designing and maintaining it. But it was entirely separate and apart from Windows. And there aren't that many companies that want to pay for developing and maintaining TWO operating systems. <quote> "Unified Driver Model" which is completely inappropriate for the Embedded/RealTime world. </quote> Inappropriate for the Real-Time world, yes. For the Embedded world... Windows is probably entirely appropriate. It's been used (NOT the CE or "Embedded Compact" flavor) for a long time that way. Over the past few years, we've seen a LOT of XP-Embedded systems needing to be revised to run Windows 7 Embedded (or later). As I wrote earlier: Industrial automation, medical instrumentation... these are Windows systems and not Windows CE systems. Peter OSR @OSRDrivers
  Message 23 of 71  
07 Oct 17 00:46
anton bassov
xxxxxx@hotmail.com
Join Date: 16 Jul 2006
Posts To This List: 4384
Driver Developer Conference

<quote> Well, the truth is also that I have never seen any OS as robust as windows! Certainly there maybe problems but for normal end users or admins I think that windows is best and it's really hard to beat it ! </quote> Wow!!!! It looks like we have finally found a candidate for the "Official Windows fanboy of NTDEV" position that has been vacant for around a year or so (i.e since Max stopped posting). Just in case if someone is bothered about my opinion, I would vote for this geezer with both hands. Let's face it - a statement like this one seems too be a way too bold even by Max's standards..... Anton Bassov
  Message 24 of 71  
07 Oct 17 01:26
anton bassov
xxxxxx@hotmail.com
Join Date: 16 Jul 2006
Posts To This List: 4384
Driver Developer Conference

> The *real* question (and what I think you're actually asking) is "Will Windows remain > a vital, evolving, developing force in computing with a growing user base for the next 10 years? The answer seems to be pretty obvious, don't you think..... > And if not... what is?" And, that, sir is the multi-billion dollar question. If you could answer it, > you'd stand the chance to become very wealthy indeed. Well, I can answer this question right on the spot - it is going to be something that is built on top of Linux kernel. Certainly,it is not going to be GNU/Linux (i.e. the OS that people actually refer to when they use the term "Linux"). It is going to be something of Android/Tizen/etc -style system,i.e something that is based upon the Linux kernel while,at he same time, being probably as un-Unix-like as one may possibly imagine. What makes me so certain? Pure logical reasoning - building your system around the Linux kernel automatically gives you an access to the widest range of devices and hardware platforms that you may possibly imagine, and giving you this access without actually having to worry about this part and,instead, delegating the whole thing to kernel.org. Therefore, unless we are speaking about some "a-la Apple" hardware vendor who wants to design a system strictly for its own hardware platform, this is the easiest and cheapest way to go. I can assure you that in a NG that is related to the system in question (whatever it may be like) a thread like this particular one will be simply infeasible. Concerning the particular way the system in question may be implemented, the question remains open - it may be a "Linux kernel - custom userland" model; it may be a hypervisor-based system with Linux kernel running as a privileged guest that has an access to the physical hardware; or it may be any combination thereof. The "only" problem is that I don't see how such a "clairvoyance" may possibly make me "very wealthy". If you don't mind,could you please make some suggestions - I am going to give you a fair share of the potential wealth if you show me any possibilities in this area..... Anton Bassov
  Message 25 of 71  
07 Oct 17 02:25
Prokash Sinha
xxxxxx@gmail.com
Join Date: 28 Jun 2012
Posts To This List: 99
Driver Developer Conference

I typed up my answers to say that - probably yes I will attend a Driver/Kernel Conference if it is within US, preferably in the west coast. How much would be useful learning is another question. Lately I got involved in Windows driver, and the first part was to get something out of WFP. In part, it is good that things are out of TDI, NDIS ( in some way), the bad part is the documentations ( as usual, nothing new here). But what surprised me is the examples they wrote. I'm not sure if MS engineers coded those examples. If so, that is a clear indication of the milage. On the Windows 10, the security and other things are on my way to test out drivers. Not like on Windows 7. From the user point of view, Windows 10 is still convoluted, trying to do too many thing. Edge is really bad. Finally MS never understood user's time. In the early days it was geeky and okay. Not now, not anymore. Just get a desktop/laptop/tablet and use it. And that goes to Apple. Linux will go to embedded, and will go to farm, not on desktop/laptop/... As I said in 2009/10 that Mac seems to be getting popular, it is now getting almost everywhere when it comes down to a desktop. Server side is still lot of Windows perhaps because exchange, database etc. So another 10 years is probably a good estimate. MS main thing now is Cloud, and consolidate servers under VM, and HW farm. And who knows, if business wants local servers, Apple servers can appear. They have money, they have the product line, they are just not concerned about its marketability, since busy with other products. -pro On Fri, Oct 6, 2017 at 10:27 PM, xxxxx@hotmail.com <xxxxx@lists.osr.com> wrote: >> The *real* question (and what I think you're actually asking) is "Will W= indows remain >> a vital, evolving, developing force in computing with a growing user ba= se for the next 10 years? > > > The answer seems to be pretty obvious, don't you think..... > > >> And if not... what is?" And, that, sir is the multi-billion dollar que= stion. If you could answer it, >> you'd stand the chance to become very wealthy indeed. > > Well, I can answer this question right on the spot - it is going to be so= mething that is built on top of Linux kernel. Certainly,it is not going to = be GNU/Linux (i.e. the OS that people actually refer to when they use the t= erm "Linux"). It is going to be something of Android/Tizen/etc -style syste= m,i.e something that is based upon the Linux kernel while,at he same time, = being probably as un-Unix-like as one may possibly imagine. > > > What makes me so certain? Pure logical reasoning - building your system = around the Linux kernel > automatically gives you an access to the widest range of devices and hard= ware platforms that you may possibly imagine, and giving you this access wi= thout actually having to worry about this part and,instead, delegating the= whole thing to kernel.org. Therefore, unless we are speaking about some "= a-la Apple" hardware vendor who wants to design a system strictly for its o= wn hardware platform, this is the easiest and cheapest way to go. I can ass= ure you that in a NG that is related to the system in question (whatever it= may be like) a thread like this particular one will be simply infeasible. > > > > > Concerning the particular way the system in question may be implemented, = the question remains open - it may be a "Linux kernel - custom userland" mo= del; it may be a hypervisor-based system with Linux kernel running as a pri= vileged guest that has an access to the physical hardware; or it may be any= combination thereof. > > > > The "only" problem is that I don't see how such a "clairvoyance" may poss= ibly make me > "very wealthy". If you don't mind,could you please make some suggestions = - I am going to give you a fair share of the potential wealth if you show m= e any possibilities in this area..... > > > Anton Bassov > > > > > --- > NTDEV is sponsored by OSR > <...excess quoted lines suppressed...> ntdev> > > MONTHLY seminars on crash dump analysis, WDF, Windows internals and softw= are drivers! > Details at <http://www.osr.com/seminars> > > To unsubscribe, visit the List Server section of OSR Online at <http://ww= w.osronline.com/page.cfm?name=3DListServer>
  Message 26 of 71  
07 Oct 17 02:47
anton bassov
xxxxxx@hotmail.com
Join Date: 16 Jul 2006
Posts To This List: 4384
Driver Developer Conference

> And who knows, if business wants local servers, Apple servers can appear. They have money, >they have the product line, The "only" thing that they haven't got is the OS that is suitable for this purpose...... Let's face it - no matter how you look at it, OSX is simply unsuitable for the server-grade systems. It is based upon the very old version of FreeBSD that, in turn, is built on top of Mach, with both running in the kernel address space. The whole thing is designed for a single-CPU system. In order to turn it into the OS that may take the full advantage of, say, even 64 logical CPUs (which is not THAT much by the server-grade system's standards) one needs almost a complete rewrite of its kernel.... Anton Bassov
  Message 27 of 71  
07 Oct 17 08:48
Prokash Sinha
xxxxxx@gmail.com
Join Date: 28 Jun 2012
Posts To This List: 99
Driver Developer Conference

SMP support is there for long time though . On Fri, Oct 6, 2017 at 11:47 PM, xxxxx@hotmail.com <xxxxx@lists.osr.com> wrote: >> And who knows, if business wants local servers, Apple servers can appear= . They have money, >>they have the product line, > > > > The "only" thing that they haven't got is the OS that is suitable for thi= s purpose...... > > > Let's face it - no matter how you look at it, OSX is simply unsuitable fo= r the server-grade systems. > It is based upon the very old version of FreeBSD that, in turn, is built= on top of Mach, with both running in the kernel address space. The whole t= hing is designed for a single-CPU system. In order to turn it into the OS t= hat may take the full advantage of, say, even 64 logical CPUs (which is not= THAT much by the server-grade system's standards) one needs almost a compl= ete rewrite of its kernel.... > > > Anton Bassov > > > > --- > NTDEV is sponsored by OSR > > Visit the list online at: <http://www.osronline.com/showlists.cfm?list=3D= ntdev> > > MONTHLY seminars on crash dump analysis, WDF, Windows internals and softw= are drivers! > Details at <http://www.osr.com/seminars> > > To unsubscribe, visit the List Server section of OSR Online at <http://ww= w.osronline.com/page.cfm?name=3DListServer>
  Message 28 of 71  
07 Oct 17 10:03
anton bassov
xxxxxx@hotmail.com
Join Date: 16 Jul 2006
Posts To This List: 4384
Driver Developer Conference

> SMP support is there for long time though . You seem to be confusing "SMP support" with "the OS designed for multi-processing"...... The former means that the system is SMP-safe - it does not necessarily .imply that the OS in question is capable of taking the full advantage of it. For example, if you use a single dispatcher spinlock for all threads, events, mutexes and other kernel objects in the system, you can still say "I support SMP", no matter how inefficient this approach may be. However, can you really say something to the effect of "this is the OS designed for multi-processing and is capable of taking the full advantage of it"? This is a philosophical question, but, on my books, "the OS designed for multi-processing" is the one that is designed in such way that it spends as little time on idle spinning as only possible. If you ask me for examples, I would immediately say "OpenSolaris". This is, indeed,an example of the server-grade OS that was designed for multiprocessing almost since its very conception. Its main problem is that it is heavily optimised for SPARC, which may negatively affect its performance on x86_64-based systems (hence the derogative term "Slowaris" that you can sometimes hear in the Linux community). As a result, the whole thing may become obsolete if SPARCvlines are shut down by Oracle http://www.zdnet.com/article/sun-set-oracle-closes-down-last-sun-product-lines/ Anton Bassov
  Message 29 of 71  
07 Oct 17 10:18
Don Burn
xxxxxx@windrvr.com
Join Date: 23 Feb 2011
Posts To This List: 1349
Driver Developer Conference

Anton stated: >> You seem to be confusing "SMP support" with "the OS designed for multi-processing"...... While I won't dispute that it is easier and better to have an OS designed for multi-processing from the beginning, but that is pretty rare. Linux could barely get out of its way when it was put on a dual processor machine in the 1990's, but it fixed its overall problems. In Windows the memory manager has been re-architected multiple times reducing the lock contention by an order of magnitude or more each time, the presentations by Landy Wang at the Driver Developer Conferences or WinHEC about how they achieved this strained the room capacity. There may be an OS out there that was completely designed for multi-processing, but most of them are constrained by API's that evolved from systems that were pre-MP. To truly get the best performance even the API's need to consider the architecture (such as both CPU, multi-processor, and how distributed the model is) from the beginning. I was part of an architecture team for a couple of OS'es in the time that multi-processing and distributed systems were early in development and it was the backwards compatibility that caused a lot of the problems. Don Burn Windows Driver Consulting Website: http://www.windrvr.com
  Message 30 of 71  
07 Oct 17 11:48
Volodymyr M. Shcherbyna
xxxxxx@shcherbyna.com
Join Date: 07 Oct 2010
Posts To This List: 164
Driver Developer Conference

Hello, On 10/07/2017 07:27 AM, xxxxx@hotmail.com wrote: > Concerning the particular way the system in question may be implemented, the question remains open - it may be a "Linux kernel - custom userland" model; it may be a hypervisor-based system with Linux kernel running as a privileged guest that has an access to the physical hardware; or it may be any combination thereof. > > > > The "only" problem is that I don't see how such a "clairvoyance" may possibly make me > "very wealthy". If you don't mind,could you please make some suggestions - I am going to give you a fair share of the potential wealth if you show me any possibilities in this area..... > > > Anton Bassov Overall it feels like the kernel development under Windows is getting less and less popular, I may conclude this on the observations of the last 10 years: - community around OSR is shrinking - MS dropping those DDK MVP awards - more & more strict rules on how to sign & certify drivers - those wars between MS & AV vendors (https://www.theverge.com/2017/6/20/15836208/microsoft-kaspersky-eu-anti-virus-co mplaint-response) most likely leading towards making windows a sealed OS like iOS or Windows 10 S? It seems like Windows kernel development becomes a niche segment and if this tendency will continue further it will just die, or it will solely be done by OEMs for support of some particular devices. Something similar may happen also with macOS kernel development - especially taking into account the recent move of Apple to prevent *signed* kext's being loaded by default unless user goes to some UI dialog and manually allows kernel driver. This is new in High Sierra and the tendency is worrying. My bet is that this will only worsen, so I constantly communicate to "venture capitalists" of my company that we need to diversify and avoid having business solely based on kernel software ... -- with best regards, Volodymyr.
  Message 31 of 71  
07 Oct 17 12:26
Prokash Sinha
xxxxxx@gmail.com
Join Date: 28 Jun 2012
Posts To This List: 99
Driver Developer Conference

Yeah, that is known that BSD 4.4 initially not designed for multiprocess support. But both BSD line and OSX's BSD part are modified to support multi-processor. Sever grade BSDs are there, I know for sure for DATACENTER parallel computing platforms. NT was designed to support that from ground up too. And I've seen 64 processor machine 10 years back. How efficiently OSX is implemented for such multi-processing is not known. But the point is -- SUN server business on downhill. Open system is a good idea, but keeping up with changes from all over is a big problem. Even for FreeBSD, the guy who wrote the book, and contributed to its development from almost the beginning with Bill Joy talked about same problem. Business people try to have definite help from Vendor, and they could not , should not wait and hope. So to me, if Apple gets into this, they will have a direct vendor that they can go to incase of trouble. Now I think, I'm going to a live lock situation with you, Anton. So lets other to talk. I'm done now :) -pro On Sat, Oct 7, 2017 at 7:04 AM, xxxxx@hotmail.com <xxxxx@lists.osr.com> wrote: > >> SMP support is there for long time though . > > > You seem to be confusing "SMP support" with "the OS designed for multi-p= rocessing"...... > > > The former means that the system is SMP-safe - it does not necessarily .i= mply that the OS in question is capable of taking the full advantage of it.= For example, if you use a single dispatcher spinlock for all threads, even= ts, mutexes and other kernel objects in the system, you can still say "I su= pport SMP", no matter how inefficient this approach may be. However, can y= ou really say something to the effect of "this is the OS designed for mult= i-processing and is capable of taking the full advantage of it"? > > > This is a philosophical question, but, on my books, "the OS designed fo= r multi-processing" is the one that is designed in such way that it spends = as little time on idle spinning as only possible. > If you ask me for examples, I would immediately say "OpenSolaris". This i= s, indeed,an example of the server-grade OS that was designed for multiproc= essing almost since its very conception. Its main problem is that it is hea= vily optimised for SPARC, which may negatively affect its performance on x8= 6_64-based systems (hence the derogative term "Slowaris" that you can somet= imes hear in the Linux community). As a result, the whole thing may become = obsolete if > SPARCvlines are shut down by Oracle > > http://www.zdnet.com/article/sun-set-oracle-closes-down-last-sun-product-= lines/ > > > Anton Bassov > > > --- > NTDEV is sponsored by OSR > > Visit the list online at: <http://www.osronline.com/showlists.cfm?list=3D= ntdev> > > MONTHLY seminars on crash dump analysis, WDF, Windows internals and softw= are drivers! > Details at <http://www.osr.com/seminars> > > To unsubscribe, visit the List Server section of OSR Online at <http://ww= w.osronline.com/page.cfm?name=3DListServer>
  Message 32 of 71  
07 Oct 17 13:39
Shaarang Tyagi
xxxxxx@gmail.com
Join Date: 08 Mar 2017
Posts To This List: 6
Driver Developer Conference

soviet_bloke: I don't really find any "wow" in linux on which after getting a driver as a normal user I am being forced to finally-at-last after trying a lot of things to make it work (and trying a hell lot of things\workarounds is one of the USP of "so called" OS linux), compile and build a driver and then you can be sure that you have done it right and sometimes it is only way left to actually use a driver, so maybe linux is only for developers and for people who know a lot of commands but for end user like people, i guess windows is far better, it may not be "wow" but not a linux like "wow" either :P On Sat, Oct 7, 2017 at 7:34 PM, xxxxx@hotmail.com < xxxxx@lists.osr.com> wrote: > > > SMP support is there for long time though . > > > You seem to be confusing "SMP support" with "the OS designed for > multi-processing"...... > > > The former means that the system is SMP-safe - it does not necessarily > .imply that the OS in question is capable of taking the full advantage of <...excess quoted lines suppressed...> --
  Message 33 of 71  
07 Oct 17 13:58
Volodymyr M. Shcherbyna
xxxxxx@shcherbyna.com
Join Date: 07 Oct 2010
Posts To This List: 164
Driver Developer Conference

shaarang tyagi, On 10/07/2017 07:38 PM, shaarang tyagi wrote: > soviet_bloke: > > I don't really find any "wow" in linux on which after getting a driver as > a normal user I am being forced to finally-at-last after trying a lot of > things to make it work (and trying a hell lot of things\workarounds is one > of the USP of "so called" OS linux), compile and build a driver and then > you can be sure that you have done it right and sometimes it is only way > left to actually use a driver, so maybe linux is only for developers and > for people who know a lot of commands but for end user like people, i guess > windows is far better, it may not be "wow" but not a linux like "wow" <...excess quoted lines suppressed...> It really depends on distribution and the vendor support of your device. If you are taking about business grade systems, usually those ship RHEL, and an installation of a driver is just a matter of downloading an RPM for your device and installing it via single command line. If you are talking about home users, yes, you may struggle with certain distros, but there are also user friendly ones: Linux Mint, Ubuntu ... -- with best regards, Volodymyr.
  Message 34 of 71  
07 Oct 17 14:04
Shaarang Tyagi
xxxxxx@gmail.com
Join Date: 08 Mar 2017
Posts To This List: 6
Driver Developer Conference

Even on user friendly ones you struggle.. On Oct 7, 2017 11:29 PM, "xxxxx@shcherbyna.com" <xxxxx@lists.osr.com> wrote: > shaarang tyagi, > > On 10/07/2017 07:38 PM, shaarang tyagi wrote: > > soviet_bloke: > > > > I don't really find any "wow" in linux on which after getting a driver > as > > a normal user I am being forced to finally-at-last after trying a lot of > > things to make it work (and trying a hell lot of things\workarounds is > one <...excess quoted lines suppressed...> --
  Message 35 of 71  
07 Oct 17 19:09
anton bassov
xxxxxx@hotmail.com
Join Date: 16 Jul 2006
Posts To This List: 4384
Driver Developer Conference

Don, Another point to consider is that the distinction between "SMP support" and "the OS designed for multi-processing" becomes obvious only when the sufficiently large number of logical CPUs are around. Otherwise, the whole thing may be pretty pointless in a sense that it gives you a marginal performance improvement on average while GREATLY increasing code complexity. For example, consider the OS with the following kernel features: 1. Adaptive kernel mutexes with support for priority inheritance 2. A mutex owner thread is allowed to block infinitely or with a timeout on semaphore or event only as long as its current (i.e. adapter) priority does not exceed its basic one - otherwise a blocking call has to return E_SIGNAL 3. Every kernel object in existence (i.e. a thread, a semaphore, a mutex, an event,a runqueue,etc) has its own spinlock - there is no such thing as a single dispatcher spinlock The question is "Can such a system be implemented?" The answer is positive, but in order to implement a system like that one needs quite complex locking hierarchies and protocols. It does not really makes sense to bother yourself with something like that if we are speaking about the system with,say, 2- 4 logical CPUs - the whole thing is simply not worth an effort. However, a system with 64-128 logical CPUs may, indeed, get a very significant and noticeable improvement from this model Anton Bassov
  Message 36 of 71  
07 Oct 17 19:28
anton bassov
xxxxxx@hotmail.com
Join Date: 16 Jul 2006
Posts To This List: 4384
Driver Developer Conference

> It really depends on distribution and the vendor support of your device. Well, in some cases our "New Windows Fanboy's" statement may,indeed,hold true. This normally happens when you use a brand-new piece of hardware with a distro that relies upon a relatively outdated kernel. It happened with me 7 years ago when I installed "Lenny" Debian on a new laptop with Tigon3 NIC. I had to install a new kernel, and then to do a bit of debugging in order to find why tg3.ko did not work properly (it was happening because it depended on another driver for PHY so that I had to load it before tg3). However, in most cases everything works just flawlessly.... Anton Bassov
  Message 37 of 71  
07 Oct 17 20:07
W. D.
xxxxxx@gmail.com
Join Date: 12 Oct 2017
Posts To This List: 40
Driver Developer Conference

Windows was not performing well on "big" machines (machines with 256 cpus). Microsoft's customers realized that, on these machines, about 15%, if I recall well, of the CPU cycles was lost spinning for the Dispatcher Lock. The Dispatcher Lock problem was solved (eliminated) with Windows 7: https://channel9.msdn.com/Shows/Going+Deep/Arun-Kishan-Farewell-to-the-Windows-Ke rnel-Dispatcher-Lock >I had to install a new kernel, and then to do a bit of debugging in order to find why tg3.ko did not work properly (it was happening because it depended on another driver for PHY so that I had to load it before tg3). The 'insmod tg3.ko' command should fail in that case and the missing symbols should be printed on the console ? The depmod command resolves modules dependencies and the modprobe command can be used to load modules with their dependencies.
  Message 38 of 71  
08 Oct 17 00:23
Shaarang Tyagi
xxxxxx@gmail.com
Join Date: 08 Mar 2017
Posts To This List: 6
Driver Developer Conference

> However, in most cases everything works just flawlessly.... Don't you provide a "Linux Fanboy" title as well? On Sun, Oct 8, 2017 at 4:59 AM, xxxxx@hotmail.com < xxxxx@lists.osr.com> wrote: > > It really depends on distribution and the vendor support of your device. > > Well, in some cases our "New Windows Fanboy's" statement may,indeed,hold > true. This normally happens when you use a brand-new piece of hardware with > a distro that relies upon a relatively outdated kernel. It happened with me > 7 years ago when I installed "Lenny" Debian on a new laptop with Tigon3 > NIC. I had to install a new kernel, and then to do a bit of debugging in > order to find why tg3.ko did not work properly (it was happening because it > depended on another driver for PHY so that I had to load it before tg3). > However, in most cases everything works just flawlessly.... <...excess quoted lines suppressed...> --
  Message 39 of 71  
08 Oct 17 01:06
anton bassov
xxxxxx@hotmail.com
Join Date: 16 Jul 2006
Posts To This List: 4384
Driver Developer Conference

> The 'insmod tg3.ko' command should fail in that case and the missing symbols should > be printed on the console ? It was not the question of unresolved dependency - it was the question of a function call returning a NULL pointer instead of the the one to platform initialisation function. As a result ,tg3.ko's probe() was returning an error, instead of performing a full device initialisation. Therefore, I had to add some printk()s to tg3.ko's probe() in order to find out at which exact point it was failing. Taking into consideration how heavily Linux relies upon calling functions by their addresses that get returned in a structure, rather than compile-time names, I think that above mentioned problem is, probably, not THAT uncommon..... Anton Bassov
  Message 40 of 71  
09 Oct 17 10:41
Gregory G. Dyess
xxxxxx@pdq.net
Join Date:
Posts To This List: 356
Driver Developer Conference

SMP, yes. Multi-socket with fault-tolerant pluggable hardware (CPU, RAM), no. --- xxxxx@lists.osr.com wrote: From: "xxxxx@gmail.com" <xxxxx@lists.osr.com> To: "Windows System Software Devs Interest List" <xxxxx@lists.osr.com> Subject: Re: [ntdev] Driver Developer Conference Date: Sat, 7 Oct 2017 05:47:42 -0700 SMP support is there for long time though . On Fri, Oct 6, 2017 at 11:47 PM, xxxxx@hotmail.com <xxxxx@lists.osr.com> wrote: >> And who knows, if business wants local servers, Apple servers can appear. They have money, >>they have the product line, > > > > The "only" thing that they haven't got is the OS that is suitable for this purpose...... > > > Let's face it - no matter how you look at it, OSX is simply unsuitable for the server-grade systems. > It is based upon the very old version of FreeBSD that, in turn, is built on top of Mach, with both running in the kernel address space. The whole thing is designed for a single-CPU system. In order to turn it into the OS that may take the full advantage of, say, even 64 logical CPUs (which is not THAT much by the server-grade system's standards) one needs almost a complete rewrite of its kernel.... <...excess quoted lines suppressed...> --- NTDEV is sponsored by OSR Visit the list online at: <http://www.osronline.com/showlists.cfm?list=ntdev> MONTHLY seminars on crash dump analysis, WDF, Windows internals and software drivers! Details at <http://www.osr.com/seminars> To unsubscribe, visit the List Server section of OSR Online at <http://www.osronline.com/page.cfm?name=ListServer>
  Message 41 of 71  
09 Oct 17 10:52
Gregory G. Dyess
xxxxxx@pdq.net
Join Date:
Posts To This List: 356
Driver Developer Conference

This is why Linux OWNS the entire HPC (High Performance Computing) domain. When you're dealing with local scheduling of 64-128 local processors with tens or even hundreds of thousands of processor cores, Windows just won't cut it and being completely closed off, no one can add extensions to make it work. I'm not a Linux fan-boy. I'm no longer a Windows (or Windows Embedded Compact) fan boy. Windows is still ahead of Linux when it comes to the average non-technical user who just wants to open their laptop, do a presentation, read a web page and then watch a movie. Mac pretty much owns anything multi-media related. Each has its place. The trend I find extremely disturbing is not the OS itself but the amount of lock-in MS is forcing on its users. Windows 10 doesn't even ask permission any more before installing whatever Microsoft wants on your computer. There's not even a way to stop it! You can slow it down by marking every network connection as "metered" but that's not going to last forever. Microsoft is forcing everyone using their products like Skype into a Single Sign On model and are pushing it hard even on Windows 10. I didn't catch it before MS reconfigured my Windows 10 tablet to go through their SSO "service" and all of their password rules I did not want. It took some doing (and lost data) to get it back to local control. Greg --- xxxxx@lists.osr.com wrote: From: "xxxxx@hotmail.com" <xxxxx@lists.osr.com> To: "Windows System Software Devs Interest List" <xxxxx@lists.osr.com> Subject: RE:[ntdev] Driver Developer Conference Date: Sat, 7 Oct 2017 19:10:15 -0400 (EDT) Don, Another point to consider is that the distinction between "SMP support" and "the OS designed for multi-processing" becomes obvious only when the sufficiently large number of logical CPUs are around. Otherwise, the whole thing may be pretty pointless in a sense that it gives you a marginal performance improvement on average while GREATLY increasing code complexity. For example, consider the OS with the following kernel features: 1. Adaptive kernel mutexes with support for priority inheritance 2. A mutex owner thread is allowed to block infinitely or with a timeout on semaphore or event only as long as its current (i.e. adapter) priority does not exceed its basic one - otherwise a blocking call has to return E_SIGNAL 3. Every kernel object in existence (i.e. a thread, a semaphore, a mutex, an event,a runqueue,etc) has its own spinlock - there is no such thing as a single dispatcher spinlock The question is "Can such a system be implemented?" The answer is positive, but in order to implement a system like that one needs quite complex locking hierarchies and protocols. It does not really makes sense to bother yourself with something like that if we are speaking about the system with,say, 2- 4 logical CPUs - the whole thing is simply not worth an effort. However, a system with 64-128 logical CPUs may, indeed, get a very significant and noticeable improvement from this model Anton Bassov --- NTDEV is sponsored by OSR Visit the list online at: <http://www.osronline.com/showlists.cfm?list=ntdev> MONTHLY seminars on crash dump analysis, WDF, Windows internals and software drivers! Details at <http://www.osr.com/seminars> To unsubscribe, visit the List Server section of OSR Online at <http://www.osronline.com/page.cfm?name=ListServer>
  Message 42 of 71  
09 Oct 17 21:26
anton bassov
xxxxxx@hotmail.com
Join Date: 16 Jul 2006
Posts To This List: 4384
Driver Developer Conference

<quote> This is why Linux OWNS the entire HPC (High Performance Computing) domain. When you're dealing with local scheduling of 64-128 local processors with tens or even hundreds of thousands of processor cores, Windows just won't cut it and being completely closed off, no one can add extensions to make it work. </quote> I would not be too sure about the original causes and resulting consequences in this respect.... I would rather suspect the original reasons for using Linux in HPC to be more of economic nature. A typical supercomputer is composed of tens ( or, probably, even hundreds) thousands of nodes. If you think of every node as of a separate OS installation that requires a separate license it becomes obvious that using Windows for HPC may be prohibitively expensive. Therefore, they started using Linux for this purpose. Once Linux happens to be an open-source kernel they started introducing the features that are particularly suitable for HPC (elimination of global locks, lockless algorithms like RCU, distributed filesystems like Lustre,etc). Once Linux evolves all the time these features eventually made their way into the main kernel tree, effectively making Linux more suitable for HPC from the technical standpoint. <quote> The trend I find extremely disturbing is not the OS itself but the amount of lock-in MS is forcing on its users. Windows 10 doesn't even ask permission any more before installing whatever Microsoft wants on your computer. There's not even a way to stop it! You can slow it down by marking every network connection as "metered" but that's not going to last forever. Microsoft is forcing everyone using their products like Skype into a Single Sign On model and are pushing it hard even on Windows 10. I didn't catch it before MS reconfigured my Windows 10 tablet to go through their SSO "service" and all of their password rules I did not want. It took some doing (and lost data) to get it back to local control. </quote> I think you should go to some university and take a course on the word "defenestration". At this point you will realise that this is the very first step that should be taken by any reasonable person upon computer purchase - either Windows has to go out of the computer, or computer has to go out of the window.... Anton Bassov
  Message 43 of 71  
10 Oct 17 13:24
Tim Roberts
xxxxxx@probo.com
Join Date: 28 Jan 2005
Posts To This List: 11622
Driver Developer Conference

xxxxx@pdq.net wrote: > The trend I find extremely disturbing is not the OS itself but the amount of lock-in MS is forcing on its users. Windows 10 doesn't even ask permission any more before installing whatever Microsoft wants on your computer. There's not even a way to stop it! As a former Windows "fan boy", I have to say this has also been a serious consideration for me.  At home, I switched to MacOS about three years ago (initially as an experiment, because the MacBook Air is so darned sexy), and I was surprised at how quickly and completely I made that switch.  I haven't even turned on my Windows laptop in 18 months. At work, one of our long-term real-time projects moved its signal processing from Windows Server 2008 to a Linux server last year, exactly because of these privacy and forced update issues.  I've had an education in cross-platform programming. Now, it's quite possible that Microsoft doesn't need to care about individuals and niche markets.  They still own the Fortune 500 and much of the non-technical consumer mindshare, and that's probably enough to sustain them for a long time to come. -- Tim Roberts, xxxxx@probo.com Providenza & Boekelheide, Inc.
  Message 44 of 71  
11 Oct 17 10:58
anton bassov
xxxxxx@hotmail.com
Join Date: 16 Jul 2006
Posts To This List: 4384
Driver Developer Conference

<quote> As a former Windows "fan boy", I have to say this has also been a serious c= onsideration for me.=C2 At home, I switched to MacOS about three years ago= (initially as an experiment, because the MacBook Air is so darned sexy), a= nd I was surprised at how quickly and completely I made that switch.=C2 I = haven't even turned on my Windows laptop in 18 months.=20 </quote> Oh dear....... I have always thought more along the lines of "Windows fanboy is not just a= fashion - it is the way of life"; "There is no such thing as a former Wind= ows fanboy"; etc etc, etc.As I can see now, this is not always the case - a= s we can see, Windows fanboys may get disappointed in their adoration objec= t and abandon their religion as well It also makes me wonder what happened to Max. Could it somehow happen that = he just got disappointed in Windows? After all, in some of his latest posts= he even forgets to mention his MVP status, i.e. something that he seemed t= o be so proud of...... Anton Bassov
  Message 45 of 71  
16 Oct 17 09:13
Slava Imameev
xxxxxx@hotmail.com
Join Date: 13 Sep 2013
Posts To This List: 204
Driver Developer Conference

Windows development isn't "cool" anymore. Windows brand isn't appealing for university students. Can anybody point to any research based on Windows kernel done outside Microsoft? Microsoft had some spectacular failures with Windows Mobile and Windows IoT, though the latter is not yet officially dead. You can't rely on Windows IoT for your project as Microsoft might cut support as happened with some devboards. There is no way to modify Windows kernel or user subsystem except begging Microsoft and your requests will fall on deaf ear if you are not listed in Fortune 100. The lack of source level debugging doesn't make the situation better. I don't see any merit in Microsoft hosting Driver Developer Conference. There is nothing to talk about except good old x86 consumer desktop PC platform where Microsoft is practically a monopoly.
  Message 46 of 71  
16 Oct 17 09:48
Don Burn
xxxxxx@windrvr.com
Join Date: 23 Feb 2011
Posts To This List: 1349
Driver Developer Conference

Slavaim, Congratulations you have become the open source fanboy of the week. Let's look at most of your points: >> Windows development isn't "cool" anymore. Windows brand isn't appealing for university students. I just got asked by a major engineering university in my area to apply to teach Windows kernel work. They came to me since their students wanted more data on Windows, and I was working with one of the researchers. >> Can anybody point to any research based on Windows kernel done outside Microsoft? I finally dropped the ACM SigOPs a few years ago, but there were a significant number of papers on Windows in the Operating System Group journal. The sad thing was so many of them were wrong, in part because it was a researcher who did not know Windows well, so made assumptions based on open source. >> There is no way to modify Windows kernel or user subsystem It is true you cannot modify the Windows kernel, but that does not mean you cannot do innovative work. Over the years I've done some major projects that Microsoft stated could not be done, and two them got highlighted at WinHEC. One was a video filesystem that delivered data faster than PCI bus speeds at the time, and the other was the first headless, diskless, server for Windows. You have to think outside of the box, but you can do interesting things with Windows. Note: given the dominance of RedHat Linux in some spaces, you have the same challenge there since they are worse than Microsoft on accepting drivers and changes. >> The lack of source level debugging doesn't make the situation better. I assume you mean source level debugging of Microsoft's kernel code. It is available but it does take some work to get. At the same time, Microsoft is putting out the source of more pieces such as the WDF source, than ever before. >> I don't see any merit in Microsoft hosting Driver Developer Conference. Every one of the things you complain about is exactly why a conference (which it is obvious you have never been to any before) is of great help to the community. It is a chance to interact with Microsoft developers and other Windows kernel dev's and share ideas and needs. When they did have conferences, I saw many more papers on Windows, and we did point out papers that were wrong and many times a response came from the community. Overall your post is one of the stupider and most biased I have seen in a long time. I used to get annoyed at Anton, but your bias means I have to give him an apology for his somewhat off approach. Don Burn Windows Driver Consulting Website: http://www.windrvr.com
  Message 47 of 71  
16 Oct 17 12:13
Slava Imameev
xxxxxx@hotmail.com
Join Date: 13 Sep 2013
Posts To This List: 204
Driver Developer Conference

Hi Don, There is no need to be so defensive. If we had lived in your world dominate= d by research papers on Windows and students queuing to enrol in Windows ke= rnel design courses we would have had Windows Mobile with at least 50% of t= he market, two billion devices with Windows IoT, every second car with Wind= ows entertainment system, Windows Servers driving Facebook and Google. You demonstrated a classical pattern of K=FCbler-Ross model with denial as = the first reaction. Windows is not a working horse for research and university courses on OS de= sign. This is the fact. You don't accept it but it doesn't change it. Some students might enrol in a course on Windows OS design but 99.5% studen= ts in CS attend Operating System design courses based on Linux, MINIX, BSD,= L4. The fact that a local university approached a person outside academic = world to conduct a Windows course proves that Windows is not a mainstream i= n OS classes. Windows kernel source level debugging is not available as long as it is not= working out of the box for everyone. Requirements to sign NDA and being ap= proved by MS mean the code is not available for majority of developers outs= ide "qualified customers, enterprises, governments, and partners". I praise your tenacity in solving problems that doesn't exist in open sourc= e world. You can blame others for being stupid and lazy but this is how it = works. Companies take the easy road by not inventing new problems with a cl= osed source system.
  Message 48 of 71  
16 Oct 17 12:21
Prokash Sinha
xxxxxx@gmail.com
Join Date: 28 Jun 2012
Posts To This List: 99
Driver Developer Conference

Good Luck Don !!! I've been living in Bay Area. It has some of the Best Universities in the world, and way too many companies. The story is completely different. Yeah, I'm now an OPEN SOAR fanboy :) -pro On Mon, Oct 16, 2017 at 6:48 AM, xxxxx@windrvr.com <xxxxx@lists.osr.com> wrote: > Slavaim, > > Congratulations you have become the open source fanboy of the week. Let's > look at most of your points: > >>> Windows development isn't "cool" anymore. Windows brand isn't appealing > for university students. > > I just got asked by a major engineering university in my area to apply > to teach Windows kernel work. They came to me since their students wanted <...excess quoted lines suppressed...>
  Message 49 of 71  
16 Oct 17 14:41
anton bassov
xxxxxx@hotmail.com
Join Date: 16 Jul 2006
Posts To This List: 4384
Driver Developer Conference

>Overall your post is one of the stupider and most biased I have seen in a long time. Oh dear......... Did you forget to take your medication today or what? Look - your overall reaction to Mr.Imameyev's post is really hysterical and, in terms of adequacy, comparable to your not-so-unusual requests to publish the company name and/or shouting in capitals all over NTDEV. Concerning Mr.Imameyev's post itself, the only thing that I don't agree with is an indirect suggestion that Windows had ever been "cool" in anyone eye's, let alone that it had ever been appealing to the university students. Please note that looking "cool" does not necessarily imply being superior in technical respects and/or overall system architecture. If you need some proof, just look at the system architecture of "cool" Linux, and "compare it to the ones of "not-so-cool" FreeBSD and "totally uncool" Solaris, so that you will have a chance to see it with your own eyes. Concerning the rest, Mr.Imameyev is absolutely correct. If anymore or less objective person looks at overall Windows architecture he is immediately going to realise that this particular OS is far from being optimal for the research purposes even for those who have a full access to its source code and is free to modify it whatever they wish. If you want to introduce a new component or subsystem you have to present it as a device that has to be dealt with via IO Manager - this is the only possible way how you can expose it to the rest of the kernel,as well as a userland. As a result, you are left to deal with quite a few things that bear no relationship to your experimental component/subsystem whatsoever. In general, one would normally prefer some other mechanism that is more flexible than IO Manager. Under a UNIX-like system this can be achieved by exposing your target subsystem as a set of function pointers to the kernel-mode components and as a file system to the userland. The latter can be optionally combined with extra system calls - for example, this is how the socket interface works under all major systems that combine sockfs with socket() and friends system calls. File system approach is particularly useful if you want to allow a shared buffer between an app and your component/subsystem. The entire System V shared memory API is based upon this approach. However, Windows system architecture would not allow something like that due to the lack of VFS - in order to do something like that you have to use IOCTLs plus share events between a driver and an app, and, as result, deal with quite a few unnecessary complications that VFS would spare you from. This is how things work for those who have a full access to Windows source code. For the rest of the world inability to see the kernel sources, let alone to modify them, takes complications to a basically new level, and makes Windows almost completely unsuitable for the research purposes. Concerning the claim of Windows gradually becoming irrelevant, Mr.Imameyev seems to be correct as well. Apart from x86-based desktop/server market, the only serious topic of concern for MSFT's seems to be a cloud-space stuff like Azure. The rest seems to be, from MSFT's perspective, just some kind of a niche market that is not really worth any more or less serious investment of time and/or money into. > Note: given the dominance of RedHat Linux in some spaces, you have the same challenge > there since they are worse than Microsoft on accepting drivers and changes. Yes, but why on Earth would any more or less reasonable person use a production-grade enterprise system like RHEL for the purposes of the research??? > Over the years I've done some major projects that Microsoft stated could not be done, > and two them got highlighted at WinHEC. One was a video filesystem that delivered > data faster than PCI bus speeds at the time, and the other was the first headless, > diskless, server for Windows. Why haven't you tried to bring ZFS to the Windows world yet??? Please don't tell me your usual mantras about the patents and lawsuits - once it is released under the terms of CDDL, the only type of code that cannot be used with it is GPLed one. Both proprietary code and permissively-licensed open-source one can be used with it without a slightest problem. The only issues that you may have are purely of a technical nature and are inherent to the very architecture of Windows kernel, but, once you claim that you can find a way around them, you should be able to do it as well. Let's see how easy it is to build ZFS around the interactions between IO Manager/Cache Manager/Memory Manager trio and to mount it on top of Ftdisk stack.... > I used to get annoyed at Anton, but your bias means I have to give him an apology > for his somewhat off approach. Now I feel really insulted. Look - Mr.Imameyev's post is absolutely harmless by my standards. When it comes to attacking MSFT and Windows, I go SIGNIFICANTLY further that than in 90+% of my posts. I really hope you are not going to tell me that, when it comes to trolling you, Mr.imameyev is better than me as well, do you. OTOH, judging from your reaction to his post, I am already not so sure.... Anton Bassov
  Message 50 of 71  
16 Oct 17 20:04
Prokash Sinha
xxxxxx@gmail.com
Join Date: 28 Jun 2012
Posts To This List: 99
Driver Developer Conference

Anton, What you just said is not entirely true ... Before Unix src was given to universities ( around late 70s to early 80s), the research was mostly within the company like: IBM, AT&T, DEC, HoneyWell, GE etc. Once Unix gave the source to educational institutions, being simple enough to do research, lot of them adopted. So that is where the start.... Then MS gave source licenses and perhaps funds to quite a bit of universities to do research. But due to no cost OS, lot of UNIX like OS ( except Half-open OS X ) like Linux came to embedded space, and many other areas. -pro On Mon, Oct 16, 2017 at 11:40 AM, xxxxx@hotmail.com <xxxxx@lists.osr.com> wrote: > >Overall your post is one of the stupider and most biased I have seen in= a long time. > > > Oh dear......... > > > Did you forget to take your medication today or what? Look - your overall= reaction to Mr.Imameyev's post is really hysterical and, in terms of adequ= acy, comparable to your > not-so-unusual requests to publish the company name and/or shouting in ca= pitals all over NTDEV. > > > > Concerning Mr.Imameyev's post itself, the only thing that I don't agree w= ith is an indirect suggestion that Windows had ever been "cool" in anyone e= ye's, let alone that it had ever been appealing to the university students.= Please note that looking "cool" does not necessarily imply being superior = in technical respects and/or overall system architecture. If you need some = proof, just look at the system architecture of "cool" Linux, and "compare i= t to the ones of "not-so-cool" FreeBSD and "totally uncool" Solaris, so tha= t you will have a chance to see it with your own eyes. > > > > Concerning the rest, Mr.Imameyev is absolutely correct. If anymore or les= s objective person looks at overall Windows architecture he is immediately = going to realise that this particular OS is far from being optimal for the = research purposes even for those who have a full access to its source code = and is free to modify it whatever they wish. If you want to introduce a new= component or subsystem you have to present it as a device that has to be d= ealt with via IO Manager - this is the only possible way how you can expose= it to the rest of the kernel,as well as a userland. As a result, you are l= eft to deal with quite a few things that bear no relationship to your exper= imental component/subsystem whatsoever. In general, one would normally pref= er some other mechanism > that is more flexible than IO Manager. > > > Under a UNIX-like system this can be achieved by exposing your target sub= system as a set of function pointers to the kernel-mode components and as a= file system to the userland. The latter can be optionally combined with ex= tra system calls - for example, this is how the socket interface works unde= r all major systems that combine sockfs with socket() and friends system ca= lls. > File system approach is particularly useful if you want to allow a shared= buffer between an app and your component/subsystem. The entire System V sh= ared memory API is based upon this approach. However, Windows system archit= ecture would not allow something like that due to the lack of VFS - in orde= r to do something like that you have to use IOCTLs plus share events betwee= n a driver and an app, and, as result, deal with quite a few unnecessary co= mplications that VFS would spare you from. > > > > This is how things work for those who have a full access to Windows sourc= e code. For the rest of the world inability to see the kernel sources, let = alone to modify them, takes complications to a basically new level, and ma= kes Windows almost completely unsuitable for the research purposes. > > > > Concerning the claim of Windows gradually becoming irrelevant, Mr.Imameye= v seems to be correct as well. Apart from x86-based desktop/server market, = the only serious topic of concern for MSFT's seems to be a cloud-space stu= ff like Azure. The rest seems to be, from MSFT's perspective, just some kin= d of a niche market that is not really worth any more or less serious inves= tment of time and/or money into. > > >> Note: given the dominance of RedHat Linux in some spaces, you have the s= ame challenge >> there since they are worse than Microsoft on accepting drivers and chang= es. > > Yes, but why on Earth would any more or less reasonable person use a prod= uction-grade > enterprise system like RHEL for the purposes of the research??? > > >> Over the years I've done some major projects that Microsoft stated could= not be done, >> and two them got highlighted at WinHEC. One was a video filesystem that = delivered >> data faster than PCI bus speeds at the time, and the other was the first= headless, >> diskless, server for Windows. > > > Why haven't you tried to bring ZFS to the Windows world yet??? Please don= 't tell me your usual > mantras about the patents and lawsuits - once it is released under the te= rms of CDDL, the only type of code that cannot be used with it is GPLed one= . Both proprietary code and permissively-licensed open-source one can be us= ed with it without a slightest problem. The only issues that you may have a= re purely of a technical nature and are inherent to the very architecture o= f Windows kernel, but, once you claim that you can find a way around them, = you should be able to do it as well. Let's see how easy it is to build ZFS = around the interactions between > IO Manager/Cache Manager/Memory Manager trio and to mount it on top of Ft= disk stack.... > > > >> I used to get annoyed at Anton, but your bias means I have to give him a= n apology >> for his somewhat off approach. > > > Now I feel really insulted. Look - Mr.Imameyev's post is absolutely harml= ess by my standards. > When it comes to attacking MSFT and Windows, I go SIGNIFICANTLY further t= hat than in 90+% of my posts. I really hope you are not going to tell me th= at, when it comes to trolling you, Mr.imameyev is better than me as well,= do you. OTOH, judging from your reaction to his post, > I am already not so sure.... > > > Anton Bassov > > > --- > NTDEV is sponsored by OSR > > Visit the list online at: <http://www.osronline.com/showlists.cfm?list=3D= ntdev> > > MONTHLY seminars on crash dump analysis, WDF, Windows internals and softw= are drivers! > Details at <http://www.osr.com/seminars> > > To unsubscribe, visit the List Server section of OSR Online at <http://ww= w.osronline.com/page.cfm?name=3DListServer>
  Message 51 of 71  
16 Oct 17 22:06
Shaarang Tyagi
xxxxxx@gmail.com
Join Date: 08 Mar 2017
Posts To This List: 6
Driver Developer Conference

Disability of the universities to do research on a particular thing does not make it irrelevant, universities unlike industries are always free to choose an easier path for thier research. On Oct 17, 2017 5:34 AM, "xxxxx@gmail.com" <xxxxx@lists.osr.com> wrote: > Anton, > > What you just said is not entirely true ... > > Before Unix src was given to universities ( around late 70s to early > 80s), the research was mostly within the company like: IBM, AT&T, DEC, > HoneyWell, GE etc. Once Unix gave the source to educational > institutions, being simple enough to do research, lot of them adopted. > So that is where the start.... > <...excess quoted lines suppressed...> --
  Message 52 of 71  
16 Oct 17 23:37
Slava Imameev
xxxxxx@hotmail.com
Join Date: 13 Sep 2013
Posts To This List: 204
Driver Developer Conference

Yes, it is bizarre that my genuine concern that Windows becomes more and more irrelevant caused such reaction from Mr. Burn. If Mr. Burn has such intimate connections with guys in Redmond to be considered by MS as being one of the "qualified customers, enterprises, governments, and partners" he would better start lobbying MS to make Windows more appealing for students, researches and developers by open sourcing desktop and server versions for research and open sourcing Windows IoT for commercial applications by adopting Android licensing model.
  Message 53 of 71  
17 Oct 17 01:18
Prokash Sinha
xxxxxx@gmail.com
Join Date: 28 Jun 2012
Posts To This List: 99
Driver Developer Conference

Well yes and no! Once it is a favorite OS of most university researchers, the companies get trained employees right from there. In most cases students do the work, professors guide and put their name first on research, and it is fact. On top of it, lot of study materials that are actually good comes with it. A good example is if you ever want to look at the network stack of any BSD product, you can look at Vol1 and Vol2 of TCP/IP book. Sometime, there is line by line match with the current source too. That gives enormous lift to someone really want to understand. On the no side, it is up to the company to pursue that it does not become irrelevant. At the time Windows came to the market, it was the best, reasonably priced platform. It built lot of eco-system, unlike anything before. But it did not catchup with the wave of other areas. Perhaps it was intentional. On the other hand, BSD if not for anything, the licensing mechanism helped lot small to moderate size cos. Once they got something going, and want to have their known parts as IP, they are on course, and sometime several major version behind the current trunk. But that's fine, and its their choice. GNU license is not that good from that point of view. By giving source to Universities, UNIX make it relevant. ATT made money on commercial OS off of it. IBM moved to support/embrace UNIX first, now Linux. -pro On Mon, Oct 16, 2017 at 7:05 PM, xxxxx@gmail.com <xxxxx@lists.osr.com> wrote: > Disability of the universities to do research on a particular thing does not > make it irrelevant, universities unlike industries are always free to choose > an easier path for thier research. > > On Oct 17, 2017 5:34 AM, "xxxxx@gmail.com" <xxxxx@lists.osr.com> wrote: >> >> Anton, >> >> What you just said is not entirely true ... >> <...excess quoted lines suppressed...>
  Message 54 of 71  
17 Oct 17 04:09
maik peterson
xxxxxx@googlemail.com
Join Date: 17 Nov 2016
Posts To This List: 3
Driver Developer Conference

just my two cents. we are a pretty big company and internally there are strong forces to replace windows overall (today it really just survived in some office domains). that seems to be a trend... i bet as soon as mobiles enter the offices windows is dead within 10 years. 2017-10-17 7:17 GMT+02:00 xxxxx@gmail.com <xxxxx@lists.osr.com>: > Well yes and no! Once it is a favorite OS of most university > researchers, the companies get trained employees right from there. In > most cases students do the work, professors guide and put their name > first on research, and it is fact. On top of it, lot of study > materials that are actually good comes with it. A good example is if > you ever want to look at the network stack of any BSD product, you can > look at Vol1 and Vol2 of TCP/IP book. Sometime, there is line by line > match with the current source too. That gives enormous lift to someone > really want to understand. > <...excess quoted lines suppressed...>
  Message 55 of 71  
17 Oct 17 08:22
anton bassov
xxxxxx@hotmail.com
Join Date: 16 Jul 2006
Posts To This List: 4384
Driver Developer Conference

> Then MS gave source licenses and perhaps funds to quite a bit of universities to do research. Are you aware of any technology,filesystem, network protocol, API, etc that had resulted from this research? If we were speaking about BSD I would be able to name quite a few things, with TCP/IP and socket API being the most obvious examples. However, I am not aware of any technology that either came from MSFT research or got developed on Windows. As far as I am concerned, all MSFT "technology" comes either from acquisitions, or from "borrowing" ideas from the UNIX world (COM-CORBA, .NET-Java relationships are the very first examples that come to my mind in this respect), or by any combination thereof. Being more or less reasonable person I DO realize that if I happen to be unaware of something it does not necessarily imply that the thing in question does not exist. Therefore, there is a good chance that MSFT research had, in actuality, produced quite a few things that I am just unaware of. Can you somehow enlighten me in this respect? <quote> On the other hand, BSD if not for anything, the licensing mechanism helped lot small to moderate size cos. Once they got something going, and want to have their known parts as IP, they are on course, and sometime several major version behind the current trunk. But that's fine, and its their choice. GNU license is not that good from that point of view. </quote> Actually, there is a not-so-uncommon opinion that GPL license happens to be, probably, the most attractive part of Linux kernel that makes it so popular in the embedded world. Let's say you have a brand-new architecture or at least a brand-new ARM or MIPS board. Which well-known open-source OS do you think it would be easier for you to port to your hardware, especially if we want to make use of some of its existing drivers (once we are not speaking about a server-grade system high-performance capabilities and optimisations of the target OS are of little concern to us)? NetBSD seems to be the most obvious option, don't you think? However, people tend to chose Linux for some reason. Linux may be a primary example in this respect, but far from being the only one . For example, look at relatively popular open-source RTOSes like RTEMS and eCos - they are also using GPL variants, rather than BSD or any other permissive license. What makes GPL so attractive for the embedded development? I heard that, despite BSD et al license's seeming permissiveness there are quite a few shady areas that may potentially leave you vulnerable to certain legal issues if you use BSD-licensed code in your product. Let's consider a practical example. Let's say a company X releases the code of, say, a filesystem under some license with the terms that are comparable to, say, the ones that Mr.Kirby released his Filedisk under. One of the recipients of this code decides to port it to some OS that is licensed under the terms of BSD, and does it while staying compliant with the terms of company X's original license, i.e. uses it as a reference and maintains a copyright notice. A company ABC likes it so much that it decides to use the enhanced version of this filesystem in its product that gets based upon the modified version of the OS in question. Some years later the company X gets hostilely taken over by the company Y that decides to close the sources of the filesystem in question, and sends "cease-and-desist" notices to all the recipients of its code. The question is "Where does company ABC stand in legal terms under this scenario"? It is hard to say anything without asking the professional lawyer's advice. Please note that if your lawyer says everything is OK the company Y's lawyers may have a different opinion on the subject, and take you to the court. No matter how it all goes, the whole thing will inevitably result in unnecessary complications, aggravations and legal costs. Why would you want all this hassle if you can simply use Linux kernel instead ???? Anton Bassov
  Message 56 of 71  
17 Oct 17 08:51
anton bassov
xxxxxx@hotmail.com
Join Date: 16 Jul 2006
Posts To This List: 4384
Driver Developer Conference

> i bet as soon as mobiles enter the offices windows is dead within 10 years. Are you requesting Mr.Burn to show all his "exceptional talents" on this thread or what? Anton Bassov
  Message 57 of 71  
17 Oct 17 16:12
Prokash Sinha
xxxxxx@gmail.com
Join Date: 28 Jun 2012
Posts To This List: 99
Driver Developer Conference

First on the licensing ( I'm not an expert). But I know a thing or two because I saw a thing or two ( famous Farmers' Insurance ads) from different cos, no one likes GPL. You can argue your case if you happens to see one of those from the legal side of any cos :) For the case of MS research, we should not be fool enough to discount that. Also you can google for it if you can toss out the soft Micro phobia. The research encompasses way larger area than OS/desktop etc. On the implementation side, you have been in this list for quite sometime, so you can tell what was designed that was first in the industry. For example, the Developer studio, and generating boilerplate source code -- I've seen both SUN and HP always praised that. Early adoption of threading (which Rick Rashid did in CMU) and integrating file map ( i.e. mmap) with the paging system. There are plenty. And lot of it comes from previous research etc. So there are bad example of CORBA-SOM, and there are usable implementation of COM/DCOM, I've seen you how bad were the former one. Linux is no invention, what you happen to praise all the time. Even Unix is a byproduct of some joint research. So there is a past trail already there, and that was either as a learning platform to build something better on top of it. Like IBM VM, APL language etc. are mostly invention I would say, but it is not just it happened all off a sudden. Some combination of things happened then something else followed thru. Anyway, I'm not a fanboy of anything, but not biased either to a point make really really stupid statements. -pro On Tue, Oct 17, 2017 at 5:21 AM, xxxxx@hotmail.com <xxxxx@lists.osr.com> wrote: >> Then MS gave source licenses and perhaps funds to quite a bit of univers= ities to do research. > > > Are you aware of any technology,filesystem, network protocol, API, etc th= at had resulted from this research? If we were speaking about BSD I would b= e able to name quite a few things, with TCP/IP and socket API being the mos= t obvious examples. However, I am not aware of any technology that either c= ame from MSFT research or got developed on Windows. As far as I am concerne= d, all MSFT "technology" comes either from acquisitions, or from "borrowing= " ideas from the UNIX world (COM-CORBA, .NET-Java relationships are the ver= y first examples that come to my mind in this respect), or by any combinati= on thereof. > > > Being more or less reasonable person I DO realize that if I happen to be = unaware of something it does not necessarily imply that the thing in questi= on does not exist. Therefore, there is a good chance that MSFT research had= , in actuality, produced quite a few things that I am just unaware of. Can= you somehow enlighten me in this respect? > > > <quote> > > On the other hand, BSD if not for anything, the licensing mechanism help= ed lot small to moderate size cos. Once they got something going, and want = to have their known parts as IP, they are on course, and sometime several m= ajor version behind the current trunk. But that's fine, and its their choic= e. GNU license is not that good from that point of view. > > </quote> > > > Actually, there is a not-so-uncommon opinion that GPL license happens to= be, probably, the most > attractive part of Linux kernel that makes it so popular in the embedded = world. Let's say you have a brand-new architecture or at least a brand-new = ARM or MIPS board. Which well-known open-source OS do you think it would be= easier for you to port to your hardware, especially if we want to make use= of some of its existing drivers (once we are not speaking about a server-g= rade system high-performance capabilities and optimisations of the target O= S are of little concern to us)? NetBSD seems to be the most obvious option,= don't you think? However, people tend to chose Linux for some reason. > > > Linux may be a primary example in this respect, but far from being the on= ly one . For example, look at relatively popular open-source RTOSes like RT= EMS and eCos - they are also using GPL variants, rather than BSD or any oth= er permissive license. What makes GPL so attractive for the embedded develo= pment? > > > I heard that, despite BSD et al license's seeming permissiveness there ar= e quite a few shady areas > that may potentially leave you vulnerable to certain legal issues if you = use BSD-licensed code in your product. Let's consider a practical example. > > > Let's say a company X releases the code of, say, a filesystem under some = license with the terms that are comparable to, say, the ones that Mr.Kirby = released his Filedisk under. One of the recipients of this code decides to = port it to some OS that is licensed under the terms of BSD, and does it whi= le staying compliant with the terms of company X's original license, i.e. u= ses it as a reference and maintains a copyright notice. A company ABC likes= it so much that it decides to use the enhanced version of this filesystem = in its product that gets based upon the modified version of the OS in quest= ion. Some years later the company X gets hostilely taken over by the compan= y Y that decides to close the sources of the filesystem in question, and s= ends "cease-and-desist" notices to all the recipients of its code. > > > The question is "Where does company ABC stand in legal terms under this s= cenario"? It is hard to say anything without asking the professional lawyer= 's advice. Please note that if your lawyer says everything is OK the compan= y Y's lawyers may have a different opinion on the subject, and take you to = the court. No matter how it all goes, the whole thing will inevitably resul= t in unnecessary complications, aggravations and legal costs. Why would you= want all this hassle if you can simply use Linux kernel instead ???? > > > > Anton Bassov > > > > --- > NTDEV is sponsored by OSR > <...excess quoted lines suppressed...> ntdev> > > MONTHLY seminars on crash dump analysis, WDF, Windows internals and softw= are drivers! > Details at <http://www.osr.com/seminars> > > To unsubscribe, visit the List Server section of OSR Online at <http://ww= w.osronline.com/page.cfm?name=3DListServer>
  Message 58 of 71  
17 Oct 17 17:11
Mark Roddy
xxxxxx@gmail.com
Join Date: 25 Feb 2000
Posts To This List: 4012
Driver Developer Conference

This thread is in disaster territory. The answer to the question should we have another Driver Dev Con, based on this thread, is "no", or actually "no fucking way", although it might be amusing to watch this nonsense live. Mark Roddy On Tue, Oct 17, 2017 at 4:11 PM, xxxxx@gmail.com <xxxxx@lists.osr.com> wrote: > First on the licensing ( I'm not an expert). But I know a thing or two > because I saw a thing or two ( famous Farmers' Insurance ads) from > different cos, no one likes GPL. You can argue your case if you > happens to see one of those from the legal side of any cos :) > > For the case of MS research, we should not be fool enough to discount > that. Also you can google for it if you can toss out the soft Micro > phobia. The research encompasses way larger area than OS/desktop etc. > > On the implementation side, you have been in this list for quite <...excess quoted lines suppressed...> --
  Message 59 of 71  
17 Oct 17 17:33
anton bassov
xxxxxx@hotmail.com
Join Date: 16 Jul 2006
Posts To This List: 4384
Driver Developer Conference

<quote> For the case of MS research, we should not be fool enough to discount that. Also you can google for it if you can toss out the soft Micro phobia. The research encompasses way larger area than OS/desktop etc. </quote> But this is not what I am asking you about, is it? The only thing I am asking you to do is to name some technology used by MSFT that had been designed/developed by MSFT research. For example, if you asked me about the SUN Microsystems and their research, on the practical side I would immediately name NFS, ZFS and Java as products/technologies. On the theoretical one, I would name VFS and SunOS memory model as architectural concepts; zones as an innovative light-weight approach to resource management; tasks and projects as an approach to workload management; interrupt threads,task queues and slab memory allocation as implementation ideas/approaches. All above mentioned things originated at SUN and subsequently made their way to SunOS/Solaris and, in this or that form, to the other OSes. Can you name something like that for MSFT? <quote> On the implementation side, you have been in this list for quite sometime, so you can tell what was designed that was first in the industry. For example, the Developer studio, and generating boilerplate source code -- I've seen both SUN and HP always praised that. </quote> Fair enough - when it comes to user friendliness, MSFT, indeed, has always been miles ahead of all UNIX-like systems combined, which allowed them to become monopolists in the desktop market. > Early adoption of threading (which Rick Rashid did in CMU) and integrating file map ( i.e. mmap) > with the paging system. There are plenty. Memory mapping and multithreading first appeared in 32-bit Windows, i.e. in mid-90s, and were,indeed, truly innovative concepts in the Windows world. The "only" problem is that both concepts had been already known in the UNIX one for almost a decade at the time. Too much for "innovation", don't you think.... > Linux is no invention Agreed - Linux is, indeed, mainly implements the concepts and ideas that come from the UNIX world (mainly Solaris ones, although with a certain influence coming from BSD derivatives). At the same time, Linux sometimes tries some interesting ideas of its own. For example,it was the first one to adjust the concept of fair scheduling (i.e. the concept that had been used for packet scheduling for quite a while) to scheduling tasks, effectively eliminating the very concept of thread quantum from its scheduler. Anton Bassov
  Message 60 of 71  
18 Oct 17 14:00
anton bassov
xxxxxx@hotmail.com
Join Date: 16 Jul 2006
Posts To This List: 4384
Driver Developer Conference

<quote> The answer to the question should we have another Driver Dev Con, based on this thread, is "no", or actually "no fucking way", although it might be amusing to watch this nonsense live. </quote> <ironical mode> Is the above statement somehow based upon the public order and safety concerns? Judging from this thread there is, indeed, a good chance that the whole thing may end up in a mass brawl more or less the following way: https://www.youtube.com/watch?v=Im_Icto6IIg I imagine Mr.Burn walking towards Mr.Imameyev with the bottle in his hands, shouting "Ay! Are you (C-word)"; our hosts trying to calm the thing down like "We are Windows driver developers, we don't fight each other, do we. Just fucking leave it out, Windows driver developers stick together", and many other"funny" moments reminiscent of the above video.... </ironical mode> Anton Bassov
  Message 61 of 71  
18 Oct 17 16:43
Pavel A
xxxxxx@fastmail.fm
Join Date: 21 Jul 2008
Posts To This List: 2401
Driver Developer Conference

> The answer to the question should we have another Driver Dev Con, based on this thread, is "no" <snip> <snip> Then, perhaps, few plugfests for problematic technologies? Based on feedback in MSDN forums, Bluetooth (BLE) is especially troubling. IMHO most of Windows kernel jobs these days revolve around "security" and "cyber" topics, which are very far from developing normal, sincere device drivers... The new ARM64 platforms may be intriguing, but still not available for outsiders. IoT... not serious, sorry. Regards, -- pa
  Message 62 of 71  
18 Oct 17 16:54
ashish kohli
xxxxxx@hotmail.com
Join Date: 29 Nov 2014
Posts To This List: 42
Driver Developer Conference

https://blogs.windows.com/devices/2017/10/17/introducing-surface-book-2-the-most- powerful-surface-book-ever/ may be this is microsoft take on mac.
  Message 63 of 71  
18 Oct 17 21:56
Prokash Sinha
xxxxxx@gmail.com
Join Date: 28 Jun 2012
Posts To This List: 99
Driver Developer Conference

My apologies to SPAM this list... My only reason to respond is what I knew many moons ago, seems to be not agreed by Anton ! AFAIK, never before NT, the mmap(FileMap) was integrated with the paging systems. Not by any UNIX in 1993/4. IRP mechanism was built to have totally Asynchronous I/O. Also pageable kernel. In fact, most any UNIX had to retrofit even SMP support. Another is the first commercial Log Structured File system, cache integration with File System, not like Buffer Cache. Back in 1998, I was at SUN, at that time they were building the pthread, so I'm not sure if any commercial product before NT had kernel threading. At any rate, Having a Dev conf. for driver / kernel or not begs the very question of how many are doing the developments in US. If far east took all the work, then there is an impact here, and it is visible ... IMO, that does not make Windows no good anymore. For really user friendly toys, Apple took the center stage, at least here. Cost is the only blocking factor to capture in some countries, and Android is filling the gap, that I've seen. Surface may be an alternative, but how would it seamlessly work with Apple and Android toys is the question. BTW, on the apple side, not many cos build apple compatible devices, that require kernel modules from OEM. Their conferences are all about TV, streaming, Watch etc. Good they publish xnu source code, and KDKs, so that anyone can navigate thru the kernel under debugger. Again it is not a true sense kernel debugger, just like kgdb. Here again I see Windbg is far better. Sorry once again ! -pro On Tue, Oct 17, 2017 at 2:32 PM, xxxxx@hotmail.com <xxxxx@lists.osr.com> wrote: > <quote> > > For the case of MS research, we should not be fool enough to discount tha= t. Also you can google for it if you can toss out the soft Micro phobia. Th= e research encompasses way larger area than OS/desktop etc. > > </quote> > > But this is not what I am asking you about, is it? The only thing I am as= king you to do is to name some technology used by MSFT that had been design= ed/developed by MSFT research. For example, if you asked me about the SUN M= icrosystems and their research, on the practical side I would immediately n= ame NFS, ZFS and Java as products/technologies. On the theoretical one, I w= ould name VFS and SunOS memory model as architectural concepts; zones as an= innovative light-weight approach to resource management; tasks and project= s as an approach to workload management; interrupt threads,task queues and = slab memory allocation as implementation ideas/approaches. All above mentio= ned things originated at SUN and subsequently made their way to SunOS/Solar= is and, in this or that form, to the other OSes. Can you name something lik= e that for MSFT? > > > <quote> > > On the implementation side, you have been in this list for quite sometime= , so you can tell what was designed that was first in the industry. For exa= mple, the Developer studio, and generating boilerplate source code -- I've = seen both SUN and HP always praised that. > > </quote> > > Fair enough - when it comes to user friendliness, MSFT, indeed, has alway= s been miles ahead of all UNIX-like systems combined, which allowed them to= become monopolists in the desktop market. > > >> Early adoption of threading (which Rick Rashid did in CMU) and integrat= ing file map ( i.e. mmap) > with the paging system. There are plenty. > > > Memory mapping and multithreading first appeared in 32-bit Windows, i.e. = in mid-90s, and were,indeed, truly innovative concepts in the Windows world= . The "only" problem is that both concepts had been already known in the UN= IX one for almost a decade at the time. Too much for "innovation", don't yo= u think.... > > >> Linux is no invention > > > Agreed - Linux is, indeed, mainly implements the concepts and ideas that= come from the UNIX world (mainly Solaris ones, although with a certain inf= luence coming from BSD derivatives). At the same time, Linux sometimes tri= es some interesting ideas of its own. For example,it was the first one to a= djust the concept of fair scheduling (i.e. the concept that had been used = for packet scheduling for quite a while) to scheduling tasks, effectively e= liminating the very concept of thread quantum from its scheduler. > > > > Anton Bassov > > > --- > NTDEV is sponsored by OSR > > Visit the list online at: <http://www.osronline.com/showlists.cfm?list=3D= ntdev> > > MONTHLY seminars on crash dump analysis, WDF, Windows internals and softw= are drivers! > Details at <http://www.osr.com/seminars> > > To unsubscribe, visit the List Server section of OSR Online at <http://ww= w.osronline.com/page.cfm?name=3DListServer>
  Message 64 of 71  
18 Oct 17 23:30
anton bassov
xxxxxx@hotmail.com
Join Date: 16 Jul 2006
Posts To This List: 4384
Driver Developer Conference

> My apologies to SPAM this list... My only reason to respond is what I knew many > moons ago, seems to be not agreed by Anton ! Well, I guess the best way to stop spamming this list is just to do a bit of research before posting. If you do it this way, the worst thing that can be possibly said of your post is that it goes a bit off-topic while still providing a perfectly valid technical information. However, when you start posting without double-checking things your posts may,indeed, start looking like a spam. We are going to see it shortly.... >AFAIK, never before NT, the mmap(FileMap) was integrated with the paging systems. > Not by any UNIX in 1993/4. Actually, this part was done under the SunOS 4.0 back in 1988, i.e. while it was still BSD-based and well before Windows NT's very conception. Earlier SunOS versions relied upon BSD memory model. SunOS 4.0 memory model integrated memory subsystem with VFS and introduced segment drivers, with the concept of the address spaces as of segments of mapped vnodes and of mapped hardware devices as of segments of hardware-mapped pages. This was about the implementation. Concerning the very idea of file mapping....well, this idea in itself is older than UNIX - IIRC, it originated under MULTICS while AT&T was still working on it. Although the API was totally different from anything that we would think of today, the concept of attaching a segment to the address space is just inseparable from the one of the virtual memory (which,BTW, also originated under MULTICS)... > IRP mechanism was built to have totally Asynchronous I/O. Also pageable kernel. > In fact, most any UNIX had to retrofit even SMP support. Another is the first commercial > Log Structured File system, cache integration with File System, not like Buffer Cache. Concerning "the first commercial Log Structured File system", I have to remind you that JFS had been introduced into AIX by IBM back in 1990, i.e 3 years before NT release. Concerning the pageable kernel and Cache Manager that works on per-file basis, rather than relying upon the buffer cache, as well as IRP-based IO.... well, these are, indeed, the examples of profound idiocy of Wndows-specific kernel concepts - they are specific to Windows and haven't got any analogues in the UNIX world. The only valid point that you made is related to SMP - Windows NT, indeed, introduced the fine-grained spinlocks at the time when most UNIX variants still relied upon the so-called "giant kernel lock". > Back in 1998, I was at SUN, at that time they were building the pthread, so I'm not > sure if any commercial product before NT had kernel threading. Fair enough - the concept of pthreads was originally, indeed, 100% userland thing, so that kernel had no idea about it. Subsequently the whole thing was moved into the kernel by the most of UNIX variants. However, saying that it was Windows who had invented the very concept of multithreading is still a bit of exaggeration..... Anton Bassov
  Message 65 of 71  
19 Oct 17 06:51
Slava Imameev
xxxxxx@hotmail.com
Join Date: 13 Sep 2013
Posts To This List: 204
Driver Developer Conference

<QUOTE> Cache Manager that works on per-file basis, rather than relying upon the buffer cache, as well as IRP-based IO.... well, these are, indeed, the examples of profound idiocy of Wndows-specific kernel concepts - they are specific to Windows and haven't got any analogues in the UNIX world. </QUOTE> It is not uncommon to hear about "buffer" cache in *NIX vs Windows NT per file caching. Actually all modern *NIX systems adopted the same concept as Windows NT. For example ramfs file system relies on per file memory caching and is not backed by any block device. In Linux pages are connected to a file via inode->i_mapping->page_tree this constitutes the page cache. The buffer cache shares this pages by connecting a list of buffer_head structures to page->private field, each buffer_head contains a block device offset and size for each block on the page. To map a page from the page cache into the buffer cache the file system's get_page callback is provided. The buffer cache per se is used only for raw data not represented via file system, but even in this case the buffer cache is unified with the page cache by block device's address_space object. macOS uses exactly the same concept connecting vnode and pages by unified buffer cache structure vnode->v_un.vu_ubcinfo.
  Message 66 of 71  
19 Oct 17 06:54
Slava Imameev
xxxxxx@hotmail.com
Join Date: 13 Sep 2013
Posts To This List: 204
Driver Developer Conference

> the file system's get_page should have been get_block
  Message 67 of 71  
19 Oct 17 08:40
Don Matthews
xxxxxx@nexitech.com
Join Date: 24 Jul 2001
Posts To This List: 30
Driver Developer Conference

>> the file system's get_page > > should have been get_block > should have been lock_thread
  Message 68 of 71  
19 Oct 17 09:02
Xiaofan Chen
xxxxxx@gmail.com
Join Date: 09 Sep 2010
Posts To This List: 183
Driver Developer Conference

On Sat, Oct 7, 2017 at 2:00 AM, xxxxx@osr.com <xxxxx@lists.osr.com> wrote: > > Windows CE was a good operating system. There were some > very impressive people designing and maintaining it. But it was entirely > separate and apart from Windows. And there aren't that many companies > that want to pay for developing and maintaining TWO operating systems. > > ... > As I wrote earlier: Industrial automation, medical instrumentation... > these are Windows systems and not Windows CE systems. Just a correction, I work in the Industrial Automation field, there are both Windows and Windows CE based automation product. We still have quite a few active catalogs (industrial display panel, industrial I/Os will Ethernet communication, etc) running on Windows CE. But now we have to consider migrating them away from Windows CE platform. Previously we have done migration for some high volume catalogs to other RTOS platform. -- Xiaofan
  Message 69 of 71  
19 Oct 17 09:20
anton bassov
xxxxxx@hotmail.com
Join Date: 16 Jul 2006
Posts To This List: 4384
Driver Developer Conference

> Actually all modern *NIX systems adopted the same concept as Windows NT. I suggest you should read Rajeev Nagar's book - it describes Cache Manager in a great detail (IIRC, it devotes three chapters to it - one chapter per interface). At this point you will realize that it does not have any analogues in the UNIX world. You can think of it just of a special KM component that maps files to the kernel address space on behalf of its KM filesystem clients exactly the same way userland apps map files into their address spaces. > For example ramfs file system relies on per file memory caching and is not backed > by any block device. To begin with, ramfs and tmpfs are two very special filesystems. RAMFS, as its very name suggests, stores all its data in RAM - the pages that store ramfs data are not reclaimable. Unlike ramfs, tmpfs is backed up by a swap partition. Although you can use it for storing temporary files( i.e.the one that cannot survive a reboot) under /tmp, its main use is to back up System V IPC shared memory segments. These two are very special filesystems that have ABSOLUTELY nothing to do with the ones that are used for the persistent storage. <quote> In Linux pages are connected to a file via inode->i_mapping->page_tree this constitutes the page cache. The buffer cache shares this pages by connecting a list of buffer_head structures to page->private field, each buffer_head contains a block device offset and size for each block on the page. To map a page from the page cache into the buffer cache the file system's get_page callback is provided. The buffer cache per se is used only for raw data not represented via file system, but even in this case the buffer cache is unified with the page cache by block device's address_space object. </quote> Well, the very term "buffer cache" is, indeed, a bit of a misnomer - these days it would be more appropriate to use the one "unified page cache". The very idea behind the unified page cache is that the page reclaiming code should be able to quickly locate all pages that can be freed, no matter if they back up a mapped file, or belong to so-called "anonymous memory" (i.e swappable pages and shared memory segments backed up by tmpfs). When it comes to anonymous pages they can be maintained in a linked list - after all, they are either private to their address spaces or represent shared memory segments. However, when it comes to mapped files, things are not that simple - there may be hoards of them, so that a list of mapped pages may be rather exhaustive. A rather convoluted approach that you have described allows page reclaiming code to quickly locate all pages that back up a given mapped file, i.e. so-called "reverse mapping", and to do so without maintaining the list of page->file relationship descriptors for each page frame. It has absolutely nothing to do with caching per se (by "caching" I mean only buffered reads and writes in this context, i.e. the way Windows Cache Manager works). However, Windows is unable to locate all the pages that back up a given file. For example, a mapping by the CM is totally unrelated to the one that is done by the userland that calls MapViewOfFile(), so that modifications by the cached writes are not immediately reflected in the file view that are mapped by the user apps. Anton Bassov
  Message 70 of 71  
19 Oct 17 09:33
anton bassov
xxxxxx@hotmail.com
Join Date: 16 Jul 2006
Posts To This List: 4384
Driver Developer Conference

> should have been lock_thread Just don't read it if you don't like it, for fuck's sake.... It is just amazing that so many people seem to have a trouble getting it. Is it really a concept that happens to be THAT complex to grasp - after all, it does not seem to be some kind of rocket science, does it.... Anton Bassov
  Message 71 of 71  
19 Oct 17 09:34
Peter Viscarola (OSR)
xxxxxx@osr.com
Join Date:
Posts To This List: 5949
List Moderator
Driver Developer Conference

We've had Godwin's law for ages. Perhaps it's time for a new such thesis: "As a discussion of Windows on NTDEV becomes longer, the probability of Anton making outrageously inflammatory statements about approaches 1." >should have been lock_thread THIS, right here, is where we need time travel debugging. Peter OSR @OSRDrivers
Posting Rules  
You may not post new threads
You may not post replies
You may not post attachments
You must login to OSR Online AND be a member of the ntdev list to be able to post.

All times are GMT -5. The time now is 02:21.


Copyright ©2015, OSR Open Systems Resources, Inc.
Based on vBulletin Copyright ©2000 - 2005, Jelsoft Enterprises Ltd.
Modified under license