mm/cma: make detection of highmem_start more robust

Pratyush Yadav reports the following crash:

    ------------[ cut here ]------------
    kernel BUG at arch/x86/mm/physaddr.c:23!
    ception 0x06 IP 10:ffffffff812ebbf8 error 0 cr2 0xffff88903ffff000
    CPU: 0 UID: 0 PID: 0 Comm: swapper Not tainted 6.15.0-rc6+ #231 PREEMPT(undef)
    Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Arch Linux 1.16.3-1-1 04/01/2014
    RIP: 0010:__phys_addr+0x58/0x60
    Code: 01 48 89 c2 48 d3 ea 48 85 d2 75 05 e9 91 52 cf 00 0f 0b 48 3d ff ff ff 1f 77 0f 48 8b 05 20 54 55 01 48 01 d0 e9 78 52 cf 00 <0f> 0b 90 0f 1f 44 00 00 90 90 90 90 90 90 90 90 90 90 90 90 90 90
    RSP: 0000:ffffffff82803dd8 EFLAGS: 00010006 ORIG_RAX: 0000000000000000
    RAX: 000000007fffffff RBX: 00000000ffffffff RCX: 0000000000000000
    RDX: 000000007fffffff RSI: 0000000280000000 RDI: ffffffffffffffff
    RBP: ffffffff82803e68 R08: 0000000000000000 R09: 0000000000000000
    R10: ffffffff83153180 R11: ffffffff82803e48 R12: ffffffff83c9aed0
    R13: 0000000000000000 R14: 0000001040000000 R15: 0000000000000000
    FS:  0000000000000000(0000) GS:0000000000000000(0000) knlGS:0000000000000000
    CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
    CR2: ffff88903ffff000 CR3: 0000000002838000 CR4: 00000000000000b0
    Call Trace:
     <TASK>
     ? __cma_declare_contiguous_nid+0x6e/0x340
     ? cma_declare_contiguous_nid+0x33/0x70
     ? dma_contiguous_reserve_area+0x2f/0x70
     ? setup_arch+0x6f1/0x870
     ? start_kernel+0x52/0x4b0
     ? x86_64_start_reservations+0x29/0x30
     ? x86_64_start_kernel+0x7c/0x80
     ? common_startup_64+0x13e/0x141

  The reason is that __cma_declare_contiguous_nid() does:

          highmem_start = __pa(high_memory - 1) + 1;

  If dma_contiguous_reserve_area() (or any other CMA declaration) is
  called before free_area_init(), high_memory is uninitialized. Without
  CONFIG_DEBUG_VIRTUAL, it will likely work but use the wrong value for
  highmem_start.

The issue occurs because commit e120d1b ("arch, mm: set high_memory
in free_area_init()") moved initialization of high_memory after the call
to dma_contiguous_reserve() -> __cma_declare_contiguous_nid() on several
architectures.

In the case CONFIG_HIGHMEM is enabled, some architectures that actually
support HIGHMEM (arm, powerpc and x86) have initialization of high_memory
before a possible call to __cma_declare_contiguous_nid() and some
initialized high_memory late anyway (arc, csky, microblase, mips, sparc,
xtensa) even before the commit e120d1b so they are fine with using
uninitialized value of high_memory.

And in the case CONFIG_HIGHMEM is disabled high_memory essentially becomes
the first address after memory end, so instead of relying on high_memory
to calculate highmem_start use memblock_end_of_DRAM() and eliminate the
dependency of CMA area creation on high_memory in majority of
configurations.

Link: https://lkml.kernel.org/r/20250519171805.1288393-1-rppt@kernel.org
Fixes: e120d1b ("arch, mm: set high_memory in free_area_init()")
Signed-off-by: Mike Rapoport (Microsoft) <rppt@kernel.org>
Reported-by: Pratyush Yadav <ptyadav@amazon.de>
Tested-by: Pratyush Yadav <ptyadav@amazon.de>
Tested-by: Alexandre Ghiti <alexghiti@rivosinc.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/cma.c

@@ -608,7 +608,10 @@ static int __init __cma_declare_contiguous_nid(phys_addr_t *basep,
 	 * complain. Find the boundary by adding one to the last valid
 	 * address.
 	 */
-	highmem_start = __pa(high_memory - 1) + 1;
+	if (IS_ENABLED(CONFIG_HIGHMEM))
+		highmem_start = __pa(high_memory - 1) + 1;
+	else
+		highmem_start = memblock_end_of_DRAM();
 	pr_debug("%s(size %pa, base %pa, limit %pa alignment %pa)\n",
 		__func__, &size, &base, &limit, &alignment);