pcidev: On arm64 we are not getting a valid BAR0 address #136
Conversation
There was a problem hiding this comment.
Thanks! There seems to be a typo, but otherwise the patch looks pretty good.
It would be nice if we can add some context. The original workaround was added in 4059598 and cited libpci 2.2.8 as a problematic version. However I took a quick look at it and am not sure where the issue got fixed... pci_generic_fill_info() in libpci seems to be basically unchanged since 2.2.8. Did you have a chance to look into it deeper?
|
|
||
| /* Don't use dev->base_addr[x] (as value for 'bar'), won't work on older libpci. */ | ||
| addr = pci_read_long(dev, bar); | ||
| if (PCI_LIB_VERSION >= 0x030500 && bar_number > -1) { |
There was a problem hiding this comment.
Any particular reason to choose 3.5.0?
There was a problem hiding this comment.
Nope, just that on my platform ( arm64) , that was the lowest version i tried.
|
Another way way that I can see is to only do this on ARM64. I was looking into how lspci did come up with the correct value. |
It seems that there is some code to work around a possible bug in old versions of libpci. This trusts libpci if it is new enough Change-Id: I8bd32c6344b0831a949c3853abeb84905420922a Signed-off-by: Martijn Berger <martijn.berger@gmail.com>
martijnberger
force-pushed
the
arm64_pci_bar
branch
from
a098a93
to
fe92293
Compare
It seems that there is some code to work around a possible bug in
old versions of libpci. This trusts libpci if it is new enough
Change-Id: I8bd32c6344b0831a949c3853abeb84905420922a
Signed-off-by: Martijn Berger martijn.berger@gmail.com