Reading/Writing to PCI Config Space of PLX bridge
in [Device Drivers]
|
From: alberto on 21 Jan 2010 16:15 I have used CF0/CFC accesses on and off in these last years, and I haven't seen them crash a system yet. The driver for the predecessor chip, which has been in production for a good 10 years now (I didn't write it!), also uses that kind of access to access its bridge at init time. But I'm the first to admit, we're talking about closed systems in dedicated application server environments here - this isn't exactly the kind of system you can buy at your friendly hardware store! Why can't the OS just export to us a mutex or spinlock that guarantees atomic access to bus space ? You know, sometimes we actually do need to play with that bridge! Alberto. On Jan 21, 2:31 pm, "Don Burn" <b...(a)stopspam.windrvr.com> wrote: > I believe you are correct about the platforms, but Alberto specifically > called out using the legacy interface and that will nail a system sometime. > > -- > Don Burn (MVP, Windows DKD) > Windows Filesystem and Driver Consulting > Website:http://www.windrvr.com > Blog:http://msmvps.com/blogs/WinDrvr > Remove StopSpam to reply > > "Pavel A." <pave...(a)12fastmail34.fm> wrote in message > > news:eeCNv9smKHA.1708(a)TK2MSFTNGP05.phx.gbl... > > > > > > > "Don Burn" <b...(a)stopspam.windrvr.com> wrote in message > >news:uUbwW0smKHA.6096(a)TK2MSFTNGP02.phx.gbl... > >> Yes it does crash, you have a 2 step process with no access to the lock, > >> how do you prevent your code from stopping the actions of the kernel in > >> the middle. I.E. they have set the address register but not accessed > >> the data register, then you blow away the address register. > > >> -- > >> Don Burn (MVP, Windows DKD) > >> Windows Filesystem and Driver Consulting > >> Website:http://www.windrvr.com > >> Blog:http://msmvps.com/blogs/WinDrvr > >> Remove StopSpam to reply > > > AFAIK at least on some current Intel platforms, PCI config > > space of bridges (and maybe of all PCI(e) devices) > > is permanently mapped to memory. So pci.sys does not > > use the legacy i/o interface. > > This may explain why Alberto is so lucky ;) > > > Regards, > > -- pa > > >> "alberto" <amore...(a)ieee.org> wrote in message > >>news:7daea67a-c15d-4e01-9b75-5593dd71da77(a)k17g2000yqh.googlegroups.com.... > > >> Do it right, and it doesn't crash. In fact, the closest you are to the > >> hardware, the less things can go wrong, and if I have to decide > >> between trusting the hardware or trusting the OS, I'll trust the > >> hardware 100% of the time. All it takes is to know your hardware > >> well. > > >> I do direct bus access a couple of times in my own driver - what do > >> you know, to fiddle with the onboard PCI-X to PCIe bridge - and yet > >> the driver is now at a stage that it runs in stress conditions 24/7 > >> for weeks uninterrupted. > > >> But what is it so important in a PCI config register that it requires > >> access at interrupt time ? For most cases I can think of, a shadow > >> will do fine. > > >> Alberto. > > >> On Jan 20, 1:45 pm, "Don Burn" <b...(a)stopspam.windrvr.com> wrote: > >>> It is a wonderful way to crash a system, this is one of the worst > >>> suggestions that could be made. > > >>> -- > >>> Don Burn (MVP, Windows DKD) > >>> Windows Filesystem and Driver Consulting > >>> Website:http://www.windrvr.com > >>> Blog:http://msmvps.com/blogs/WinDrvr > >>> Remove StopSpam to reply > > >>> "alberto" <amore...(a)ieee.org> wrote in message > > >>>news:a03b6446-0abd-4dad-a9b8-ac18effa20ae(a)p8g2000yqb.googlegroups.com.... > > >>> Have you tried to use CF8/CFC ? I know this is a big no-no in > >>> Windowsland, but sometimes there isn't any better alternative. > > >>> Alberto. > > >>> On Jan 19, 6:03 pm, Patrick <Patr...(a)discussions.microsoft.com> wrote: > > >>> > Hi all, > > >>> > I've a system which containing several PLX bridges. I need access to a > >>> > particular GPIO register which exists in the extended PCI Config > >>> > Space. > >>> > Previously we were using HalGetBusData/SetBusData to toggle the GPIO, > >>> > but > >>> > these calls are now invalid in Vista. > > >>> > The driver for the PLX bridges are the default pci.sys (from > >>> > Microsoft). > > >>> > Is there a method to do this ? Some of the newer calls requires the > >>> > PDO or DO, and these calls only allow me to read/write the PCI Config > >>> > space > >>> > of my own device. What I need is the ability to read/write the PCI > >>> > Config > >>> > Space of another device. For Vista, how can a driver do (1) scan all > >>> > devices in the pci bridge (2) select a device, and go modify the PCI > >>> > Config > >>> > Space ? > > >>> > At this point, all I can get is the bus/device/function id of the PLX > >>> > bridge that I need to modify the GPIO, but I am not really sure how I > >>> > can > >>> > get > >>> > access to that device's config space to make the change. Any advice > >>> > will > >>> > be > >>> > much appreciated. > > > __________ Information from ESET NOD32 Antivirus, version of virus > > signature database 4794 (20100121) __________ > > > The message was checked by ESET NOD32 Antivirus. > > >http://www.eset.com > > __________ Information from ESET NOD32 Antivirus, version of virus signature database 4794 (20100121) __________ > > The message was checked by ESET NOD32 Antivirus. > > http://www.eset.com- Hide quoted text - > > - Show quoted text -
From: Alexander Grigoriev on 21 Jan 2010 22:27 The OS exports the interface for that, actually 3 interfaces. You should be using it (at least HalGetBusData). But they can't be used at DIRQL (in ISR). "alberto" <amoreira(a)ieee.org> wrote in message news:30d64d6c-66ba-4d66-b5d4-ba1f13bffe83(a)n31g2000vbt.googlegroups.com... I have used CF0/CFC accesses on and off in these last years, and I haven't seen them crash a system yet. The driver for the predecessor chip, which has been in production for a good 10 years now (I didn't write it!), also uses that kind of access to access its bridge at init time. But I'm the first to admit, we're talking about closed systems in dedicated application server environments here - this isn't exactly the kind of system you can buy at your friendly hardware store! Why can't the OS just export to us a mutex or spinlock that guarantees atomic access to bus space ? You know, sometimes we actually do need to play with that bridge! Alberto. On Jan 21, 2:31 pm, "Don Burn" <b...(a)stopspam.windrvr.com> wrote: > I believe you are correct about the platforms, but Alberto specifically > called out using the legacy interface and that will nail a system > sometime. > > -- > Don Burn (MVP, Windows DKD) > Windows Filesystem and Driver Consulting > Website:http://www.windrvr.com > Blog:http://msmvps.com/blogs/WinDrvr > Remove StopSpam to reply > > "Pavel A." <pave...(a)12fastmail34.fm> wrote in message > > news:eeCNv9smKHA.1708(a)TK2MSFTNGP05.phx.gbl... > > > > > > > "Don Burn" <b...(a)stopspam.windrvr.com> wrote in message > >news:uUbwW0smKHA.6096(a)TK2MSFTNGP02.phx.gbl... > >> Yes it does crash, you have a 2 step process with no access to the > >> lock, > >> how do you prevent your code from stopping the actions of the kernel in > >> the middle. I.E. they have set the address register but not accessed > >> the data register, then you blow away the address register. > > >> -- > >> Don Burn (MVP, Windows DKD) > >> Windows Filesystem and Driver Consulting > >> Website:http://www.windrvr.com > >> Blog:http://msmvps.com/blogs/WinDrvr > >> Remove StopSpam to reply > > > AFAIK at least on some current Intel platforms, PCI config > > space of bridges (and maybe of all PCI(e) devices) > > is permanently mapped to memory. So pci.sys does not > > use the legacy i/o interface. > > This may explain why Alberto is so lucky ;) > > > Regards, > > -- pa > > >> "alberto" <amore...(a)ieee.org> wrote in message > >>news:7daea67a-c15d-4e01-9b75-5593dd71da77(a)k17g2000yqh.googlegroups.com... > > >> Do it right, and it doesn't crash. In fact, the closest you are to the > >> hardware, the less things can go wrong, and if I have to decide > >> between trusting the hardware or trusting the OS, I'll trust the > >> hardware 100% of the time. All it takes is to know your hardware > >> well. > > >> I do direct bus access a couple of times in my own driver - what do > >> you know, to fiddle with the onboard PCI-X to PCIe bridge - and yet > >> the driver is now at a stage that it runs in stress conditions 24/7 > >> for weeks uninterrupted. > > >> But what is it so important in a PCI config register that it requires > >> access at interrupt time ? For most cases I can think of, a shadow > >> will do fine. > > >> Alberto. > > >> On Jan 20, 1:45 pm, "Don Burn" <b...(a)stopspam.windrvr.com> wrote: > >>> It is a wonderful way to crash a system, this is one of the worst > >>> suggestions that could be made. > > >>> -- > >>> Don Burn (MVP, Windows DKD) > >>> Windows Filesystem and Driver Consulting > >>> Website:http://www.windrvr.com > >>> Blog:http://msmvps.com/blogs/WinDrvr > >>> Remove StopSpam to reply > > >>> "alberto" <amore...(a)ieee.org> wrote in message > > >>>news:a03b6446-0abd-4dad-a9b8-ac18effa20ae(a)p8g2000yqb.googlegroups.com... > > >>> Have you tried to use CF8/CFC ? I know this is a big no-no in > >>> Windowsland, but sometimes there isn't any better alternative. > > >>> Alberto. > > >>> On Jan 19, 6:03 pm, Patrick <Patr...(a)discussions.microsoft.com> wrote: > > >>> > Hi all, > > >>> > I've a system which containing several PLX bridges. I need access to > >>> > a > >>> > particular GPIO register which exists in the extended PCI Config > >>> > Space. > >>> > Previously we were using HalGetBusData/SetBusData to toggle the > >>> > GPIO, > >>> > but > >>> > these calls are now invalid in Vista. > > >>> > The driver for the PLX bridges are the default pci.sys (from > >>> > Microsoft). > > >>> > Is there a method to do this ? Some of the newer calls requires the > >>> > PDO or DO, and these calls only allow me to read/write the PCI > >>> > Config > >>> > space > >>> > of my own device. What I need is the ability to read/write the PCI > >>> > Config > >>> > Space of another device. For Vista, how can a driver do (1) scan all > >>> > devices in the pci bridge (2) select a device, and go modify the PCI > >>> > Config > >>> > Space ? > > >>> > At this point, all I can get is the bus/device/function id of the > >>> > PLX > >>> > bridge that I need to modify the GPIO, but I am not really sure how > >>> > I > >>> > can > >>> > get > >>> > access to that device's config space to make the change. Any advice > >>> > will > >>> > be > >>> > much appreciated. > > > __________ Information from ESET NOD32 Antivirus, version of virus > > signature database 4794 (20100121) __________ > > > The message was checked by ESET NOD32 Antivirus. > > >http://www.eset.com > > __________ Information from ESET NOD32 Antivirus, version of virus > signature database 4794 (20100121) __________ > > The message was checked by ESET NOD32 Antivirus. > > http://www.eset.com- Hide quoted text - > > - Show quoted text -
First
|
Prev
|
Pages: 1 2 3 4 Prev: Display Blanking in Windows 7 Next: Replacing signed drivers during development |