KDL: SMAP violation (intel_extreme)

[waddlesplash]

Yes, the driver in the screenshot is in non-packaged, but it was "just built" before rebooting.

I guess radeon_hd probably will have the same issue?

[korli]

Please try to change at ​https://git.haiku-os.org/haiku/tree/src/add-ons/kernel/drivers/graphics/intel_extreme/bios.cpp#n138

B_READ_AREA => B_KERNEL_READ_AREA

[korli]

for radeon_hd:

​https://git.haiku-os.org/haiku/tree/src/add-ons/kernel/drivers/graphics/radeon_hd/radeon_hd.cpp#n54

B_READ_AREA => B_KERNEL_READ_AREA

​https://git.haiku-os.org/haiku/tree/src/add-ons/kernel/drivers/graphics/radeon_hd/radeon_hd.cpp#n88

B_READ_AREA | B_WRITE_AREA => B_KERNEL_READ_AREA | B_KERNEL_WRITE_AREA

​https://git.haiku-os.org/haiku/tree/src/add-ons/kernel/drivers/graphics/radeon_hd/radeon_hd.cpp#n108

B_READ_AREA => B_KERNEL_READ_AREA

[waddlesplash]

With that change, the intel_extreme driver gets farther, but then there is another KDL upon reaching accelerant init.

[korli]

Please try to change at ​https://git.haiku-os.org/haiku/tree/src/add-ons/kernel/drivers/graphics/intel_extreme/intel_extreme.cpp#n319 0 => B_KERNEL_READ_AREA | B_KERNEL_WRITE_AREA | B_USER_CLONEABLE_AREA

radeon_hd ​https://git.haiku-os.org/haiku/tree/src/add-ons/kernel/drivers/graphics/radeon_hd/radeon_hd.cpp#n543 0 => B_KERNEL_READ_AREA | B_KERNEL_WRITE_AREA | B_USER_CLONEABLE_AREA

[waddlesplash]

Didn't help. Same KDL, same location (+ 0xe0, memset.)

[waddlesplash]

With all of the changes, I can successfully boot to desktop with SMAP disabled, so at least they don't hurt. Applied the first one in hrev52002, since that one seems like an obvious fix.

[korli]

I don't know, the same happens in vesa: ​https://git.haiku-os.org/haiku/tree/src/add-ons/kernel/drivers/graphics/vesa/vesa.cpp#n367

[waddlesplash]

Some more fixes done in hrev52032. The culprit behind the attached KDL, however, is the agp_gart bus. The patch for that is:

--- a/src/add-ons/kernel/busses/agp_gart/intel_gart.cpp
+++ b/src/add-ons/kernel/busses/agp_gart/intel_gart.cpp
@@ -542,14 +542,16 @@ intel_map(intel_info &info)
        info.aperture_area = apertureMapper.Map("intel graphics aperture",
                info.aperture_physical_base, info.aperture_size,
                B_ANY_KERNEL_BLOCK_ADDRESS | B_MTR_WC,
-               B_READ_AREA | B_WRITE_AREA, (void**)&info.aperture_base);
+               B_KERNEL_READ_AREA | B_KERNEL_WRITE_AREA | B_USER_CLONEABLE_AREA,
+               (void**)&info.aperture_base);
        if (apertureMapper.InitCheck() < B_OK) {
                // try again without write combining
                ERROR("enabling write combined mode failed.\n");
 
                info.aperture_area = apertureMapper.Map("intel graphics aperture",
                        info.aperture_physical_base, info.aperture_size,
-                       B_ANY_KERNEL_BLOCK_ADDRESS, B_READ_AREA | B_WRITE_AREA,
+                       B_ANY_KERNEL_BLOCK_ADDRESS,
+                       B_KERNEL_READ_AREA | B_KERNEL_WRITE_AREA | B_USER_CLONEABLE_AREA,
                        (void**)&info.aperture_base);
        }
        if (apertureMapper.InitCheck() < B_OK) {

However with this, the screen goes black after the rocket icon, so it seems the accelerant expects these areas to be writeable, not cloneable. I haven't looked into that yet.

[korli]

Please try again with hrev52052. Works for me (even if the display output is broken).

[korli]

No feedback.