Skip to contentSkip to conversation

PCIe interface memory memory mapping issue

DISCUSSION Sanka Piyaratna 2 participants 5 messages Jun 12, 2009 – Jun 14, 2009

lore
Share & tools
lore
#0 lore raw

Hi,

I have developed a PCI express interface using Xilinx ML555 hardware module. I have implemented Linux kernel mode device drivers and everything works correctly as long as I am using the device within a computer with dual channel DDR arrangement. However, as soon as I pug this device into a core i7 or an older single channel DDR machine, the interface memory mapping does not work any more. As if the register map with in the device does no longer exsists. However, "lspci" utility provides correct information. I am not sure if this has anything to do with the number of DDR memory channels the motherboard has or why that would be a problem for PCI express device. However, this seem to be the common link between the machines that demonstrate this issue.

Could some one please let me know if you have ever experience this kind of problem before and/or any ideas to fix this problem?

Thank you very much,

Regards,

Sanka

Need a Holiday? Win a $10,000 Holiday of your choice. Enter now.http://us.lrd.yahoo.com/_ylc=X3oDMTJxN2x2ZmNpBF9zAzIwMjM2MTY2MTMEdG1fZG1lY2gDVGV4dCBMaW5rBHRtX2xuawNVMTEwMzk3NwR0bV9uZXQDWWFob28hBHRtX3BvcwN0YWdsaW5lBHRtX3BwdHkDYXVueg--/SIG=14600t3ni/**http%3A//au.rd.yahoo.com/mail/tagline/creativeholidays/*http%3A//au.docs.yahoo.com/homepageset/%3Fp1=other%26p2=au%26p3=mailtagline

#1 lore raw
Quoted 3 lines · Sanka Piyaratna

Hi,

I have developed a PCI express interface using Xilinx ML555 hardware module. I have implemented Linux kernel mode device drivers and everything works correctly as long as I am using the device within a computer with dual channel DDR arrangement. However, as soon as I pug this device into a core i7 or an older single channel DDR machine, the interface memory mapping does not work any more. As if the register map with in the device does no longer exsists. However, "lspci" utility provides correct information. I am not sure if this has anything to do with the number of DDR memory channels the motherboard has or why that would be a problem for PCI express device. However, this seem to be the common link between the machines that demonstrate this issue.

You'll have to give more details on what you mean by "the interface memory mapping does not work any more. As if the register map with in the device does no longer exsists".

#2 lore raw

Hi Robert

Thanks for your reply.

Quoted 3 lines
Quoted 1 line

I have developed a PCI express interface using Xilinx ML555 hardware module. I have implemented Linux kernel >>mode device drivers and everything works correctly as long as I am using the device within a computer with dual >>channel DDR arrangement. However, as soon as I pug this device into a core i7 or an older single channel DDR >>machine, the interface memory mapping does not work any more. As if the register map with in the device does no >>longer exsists. However, "lspci" utility provides correct information. I am not sure if this has anything to do with the >>number of DDR memory channels the motherboard has or why that would be a problem for PCI express device. >>However, this seem to be the common link between the machines that demonstrate this issue.

You'll have to give more details on what you mean by "the interface memory mapping does not work any more. As if >the register map with in the device does no longer exsists".

What I mean by this is that, when load the driver, the memory mapping of the hardware memory onto the computer memory does not work. I have a setup where I have all the control registers in BAR5 (512 byte) and I also have 64kB chuck of the FPGA memory mapped using BAR0. I am not able to see the register space when the BAR5 area in the computer memory map. However, I am able to write something to BAR0 memory map and read it back.

Thanks and regards,

Sanka

Need a Holiday? Win a $10,000 Holiday of your choice. Enter now.http://us.lrd.yahoo.com/_ylc=X3oDMTJxN2x2ZmNpBF9zAzIwMjM2MTY2MTMEdG1fZG1lY2gDVGV4dCBMaW5rBHRtX2xuawNVMTEwMzk3NwR0bV9uZXQDWWFob28hBHRtX3BvcwN0YWdsaW5lBHRtX3BwdHkDYXVueg--/SIG=14600t3ni/**http%3A//au.rd.yahoo.com/mail/tagline/creativeholidays/*http%3A//au.docs.yahoo.com/homepageset/%3Fp1=other%26p2=au%26p3=mailtagline

#3 lore raw
Quoted 11 lines · Sanka Piyaratna

Hi Robert

Thanks for your reply.

Quoted 3 lines
Quoted 1 line

I have developed a PCI express interface using Xilinx ML555 hardware module. I have implemented Linux kernel >>mode device drivers and everything works correctly as long as I am using the device within a computer with dual >>channel DDR arrangement. However, as soon as I pug this device into a core i7 or an older single channel DDR >>machine, the interface memory mapping does not work any more. As if the register map with in the device does no >>longer exsists. However, "lspci" utility provides correct information. I am not sure if this has anything to do with the >>number of DDR memory channels the motherboard has or why that would be a problem for PCI express device. >>However, this seem to be the common link between the machines that demonstrate this issue.

You'll have to give more details on what you mean by "the interface memory mapping does not work any more. As if >the register map with in the device does no longer exsists".

What I mean by this is that, when load the driver, the memory mapping of the hardware memory onto the computer memory does not work. I have a setup where I have all the control registers in BAR5 (512 byte) and I also have 64kB chuck of the FPGA memory mapped using BAR0. I am not able to see the register space when the BAR5 area in the computer memory map. However, I am able to write something to BAR0 memory map and read it back.

How are you getting the register space into the memory map.. ioremap() ?

#4 lore raw
Quoted 10 lines
Quoted 8 lines
Quoted 6 lines
Quoted 4 lines

I have developed a PCI express interface using Xilinx ML555 hardware module. I have implemented Linux kernel >>mode device drivers and everything works correctly as long as I am using the device within a computer with dual >>channel DDR arrangement. However, as soon as I pug this device into a core i7 or an older single channel DDR >>machine, the interface memory mapping does not work any more. As if the register map with in the device does no >>longer exsists. However, "lspci" utility provides correct information. I am not sure if this has anything to do with the >>number of DDR memory channels the motherboard has or why that would be a problem for PCI express device. >>However, this seem to be the common link between the machines that demonstrate this issue.

You'll have to give more details on what you mean by "the interface memory mapping does not work any more. As if >the register map with in the device does no longer exsists".

What I mean by this is that, when load the driver, the memory mapping of the hardware memory onto the computer memory does not work. I have a setup where I have all the control registers in BAR5 (512 byte) and I also have 64kB chuck of the FPGA memory mapped using BAR0. I am not able to see the register space when the BAR5 area in the computer memory map. However, I am able to write something to BAR0 memory map and read it back.

How are you getting the register space into the memory map.. ioremap() ?

that's right! I am using ioremap(address, size).

Need a Holiday? Win a $10,000 Holiday of your choice. Enter now.http://us.lrd.yahoo.com/_ylc=X3oDMTJxN2x2ZmNpBF9zAzIwMjM2MTY2MTMEdG1fZG1lY2gDVGV4dCBMaW5rBHRtX2xuawNVMTEwMzk3NwR0bV9uZXQDWWFob28hBHRtX3BvcwN0YWdsaW5lBHRtX3BwdHkDYXVueg--/SIG=14600t3ni/**http%3A//au.rd.yahoo.com/mail/tagline/creativeholidays/*http%3A//au.docs.yahoo.com/homepageset/%3Fp1=other%26p2=au%26p3=mailtagline