d/freebsd-dev
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

[PowerPC] Fix translation-related crashes during startup

Browse Source 
After spending a lot of time trying to track down what was going on, I have
isolated the "black screen" failures when using boot1 to boot a G4 machine.

It turns out we were replacing the traps before installing the temporary
BAT entry for the bottom of physical memory. That meant that until the MMU
was bootstrapped, the cached translations were the only thing keeping us
from losing.

Throwing boot1 into the mix was affecting execution flow enough to cause us
to hit an uncached page and crash.

Fix this by properly setting up the initial BAT entry at the same time we
are replacing the OpenFirmware traps, so we can continue executing in
segment 0 until the rest of the DMAP has been set up.

A second thing discovered while researching this is that we were entering a
BAT region for segment 16. It turns out this range was a) considered part
of KVA, and b) has firmware mappings with varying attributes.

If we ever accessed an unmapped page in segment 16, it would cause a BAT
entry to be installed for the whole segment, which would bypass the
existing mappings until it was flushed out again.

Instead, translate the OFW memory attributes into VM memory attributes and
install the ranges into the kernel address space properly.

Reviewed by:	adalava
MFC after:	3 weeks
Sponsored by:	Tag1 Consulting, Inc.
Differential Revision:	https://reviews.freebsd.org/D25547
This commit is contained in:
Brandon Bergren 2020-08-21 03:31:01 +00:00
parent b94b2fcd61
commit f10baa4049
Notes: svn2git 2020-12-20 02:59:44 +00:00 
svn path=/head/; revision=364447
2 changed files with 56 additions and 29 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

19
sys/powerpc/aim/aim_machdep.c
View File

@ -390,6 +390,25 @@ aim_cpu_init(vm_offset_t toc)
bcopy(&restorebridge, (void *)EXC_MCHK, trap_offset);
bcopy(&restorebridge, (void *)EXC_TRC, trap_offset);
bcopy(&restorebridge, (void *)EXC_BPT, trap_offset);
} else {
/*
* Use an IBAT and a DBAT to map the bottom 256M segment.
*
* It is very important to do it *now* to avoid taking a
* fault in .text / .data before the MMU is bootstrapped,
* because until then, the translation data has not been
* copied over from OpenFirmware, so our DSI/ISI will fail
* to find a match.
*/
battable[0x0].batl = BATL(0x00000000, BAT_M, BAT_PP_RW);
battable[0x0].batu = BATU(0x00000000, BAT_BL_256M, BAT_Vs);
__asm (".balign 32; \n"
"mtibatu 0,%0; mtibatl 0,%1; isync; \n"
"mtdbatu 0,%0; mtdbatl 0,%1; isync"
:: "r"(battable[0].batu), "r"(battable[0].batl));
}
#else
trapsize = (size_t)&hypertrapcodeend - (size_t)&hypertrapcode;

66
sys/powerpc/aim/mmu_oea.c
View File

@ -420,6 +420,32 @@ moea_calc_wimg(vm_paddr_t pa, vm_memattr_t ma)
return pte_lo;
}
/*
* Translate OFW translations into VM attributes.
*/
static __inline vm_memattr_t
moea_bootstrap_convert_wimg(uint32_t mode)
{
switch (mode) {
case (PTE_I | PTE_G):
/* PCI device memory */
return VM_MEMATTR_UNCACHEABLE;
case (PTE_M):
/* Explicitly coherent */
return VM_MEMATTR_CACHEABLE;
case 0: /* Default claim */
case 2: /* Alternate PP bits set by OF for the original payload */
/* "Normal" memory. */
return VM_MEMATTR_DEFAULT;
default:
/* Err on the side of caution for unknowns */
/* XXX should we panic instead? */
return VM_MEMATTR_UNCACHEABLE;
}
}
static void
tlbie(vm_offset_t va)
{
@ -670,13 +696,6 @@ moea_bootstrap(vm_offset_t kernelstart, vm_offset_t kernelend)
vm_size_t size, physsz, hwphyssz;
vm_offset_t pa, va, off;
void *dpcpu;
register_t msr;
/*
* Set up BAT0 to map the lowest 256 MB area
*/
battable[0x0].batl = BATL(0x00000000, BAT_M, BAT_PP_RW);
battable[0x0].batu = BATU(0x00000000, BAT_BL_256M, BAT_Vs);
/*
* Map PCI memory space.
@ -693,24 +712,7 @@ moea_bootstrap(vm_offset_t kernelstart, vm_offset_t kernelend)
battable[0xb].batl = BATL(0xb0000000, BAT_I|BAT_G, BAT_PP_RW);
battable[0xb].batu = BATU(0xb0000000, BAT_BL_256M, BAT_Vs);
/*
* Map obio devices.
*/
battable[0xf].batl = BATL(0xf0000000, BAT_I|BAT_G, BAT_PP_RW);
battable[0xf].batu = BATU(0xf0000000, BAT_BL_256M, BAT_Vs);
/*
* Use an IBAT and a DBAT to map the bottom segment of memory
* where we are. Turn off instruction relocation temporarily
* to prevent faults while reprogramming the IBAT.
*/
msr = mfmsr();
mtmsr(msr & ~PSL_IR);
__asm (".balign 32; \n"
"mtibatu 0,%0; mtibatl 0,%1; isync; \n"
"mtdbatu 0,%0; mtdbatl 0,%1; isync"
:: "r"(battable[0].batu), "r"(battable[0].batl));
mtmsr(msr);
powerpc_sync();
/* map pci space */
__asm __volatile("mtdbatu 1,%0" :: "r"(battable[8].batu));
@ -910,15 +912,21 @@ moea_bootstrap(vm_offset_t kernelstart, vm_offset_t kernelend)
/*
* If the mapping is 1:1, let the RAM and device
* on-demand BAT tables take care of the translation.
*
* However, always enter mappings for segment 16,
* which is mixed-protection and therefore not
* compatible with a BAT entry.
*/
if (translations[i].om_va == translations[i].om_pa)
continue;
if ((translations[i].om_va >> ADDR_SR_SHFT) != 0xf &&
translations[i].om_va == translations[i].om_pa)
continue;
/* Enter the pages */
for (off = 0; off < translations[i].om_len;
off += PAGE_SIZE)
moea_kenter(translations[i].om_va + off,
translations[i].om_pa + off);
moea_kenter_attr(translations[i].om_va + off,
translations[i].om_pa + off,
moea_bootstrap_convert_wimg(translations[i].om_mode));
}
}