On Tuesday, October 21, 2014 9:29 AM, Ian Abbott wrote: > On 20/10/14 18:57, H Hartley Sweeten wrote: >> This driver currently passes the timer "channel" in a manner that violates >> the comedi API. Fix the timer subdevice so that the timers "channels" are >> correctly used from the insn->chanspec. >> >> Fix the I/O access for the boards registers. Currently the iobase used to >> access the counters and the main board registers are incorrect. According >> to the ADDI-DATA datasheet the PCI BARs are: >> >> PCI BAR0 - amcc chip registers >> PCI BAR1 - main board registers >> PCI BAR2 - counter registers >> >> Move the register map defines from the included hwdrv_apci1564.c file to >> the main driver source file. >> >> H Hartley Sweeten (9): >> staging: comedi: addi_apci_1564: remove APCI1564_COUNTER[1234] defines >> staging: comedi: addi_apci_1564: remove private data 'mode_select_register' >> staging: comedi: addi_apci_1564: fix counter register access >> staging: comedi: addi_apci_1564: board has 4 timers >> staging: comedi: addi_apci_1564: fix board register access >> staging: comedi: addi_apci_1564: tidy up private data 'amcc_iobase' >> staging: comedi: addi_apci_1564: move register map defines to driver >> staging: comedi: addi_apci_1564: move APCI1564_DI_IRQ_REG bit defines >> staging: comedi: addi_apci_1564: move APCI1564_DO_INT_CTRL/STATUS_REG bit defines >> >> .../comedi/drivers/addi-data/hwdrv_apci1564.c | 155 ++++++------------ >> drivers/staging/comedi/drivers/addi_apci_1564.c | 179 +++++++++++++-------- >> 2 files changed, 159 insertions(+), 175 deletions(-) >> > > The changes to PCI BARs worry me. Perhaps the old PCI BARs worked with > older models of the card. Presumably they stopped using the AMCC chip > some years ago as it's no longer available. They might have moved the > registers around and changed the PCI IDs at the same time, or possibly > moved the registers around and *not* changed the PCI IDs! (That nearly > happened with the MeasurementComputing/ComputerBoards PCI-DIO48H which > is now in the comedi 8255_pci driver, except it did turn out that they > changed the PCI subdevice IDs after switching from the AMCC S5933 (I > think) to a PLX PCI9050/PCI9052.) > > Well, ADDI-DATA have their own Linux driver for this board which only > seems to use BAR0 and BAR1, so something fishy is going on. > > http://www.addi-data.com/treiber/linux/kernel_ioctl_api/apci1564_r1283.tar.bz2 Ian, The I/O Mapping document I have from ADDI-DATA is dated 27.02.2004. It was modified on 29.07.08 with the counter information. The ADD-DATA driver only using PCI BARS 0 and 1 makes sense if that driver does not have any counter support. As messed up as the ADDI-DATA drivers were, my guess is the PCI BAR issue was initially a screw up when the drivers were initially merged. I will ask ADDI-DATA about the PCI BARs to see if they have any additional information. Regards, Hartley _______________________________________________ devel mailing list devel@xxxxxxxxxxxxxxxxxxxxxx http://driverdev.linuxdriverproject.org/mailman/listinfo/driverdev-devel
