At least with CONFIG_PHYSICAL_START=0x100000, if there is < 4 MiB of contiguous
free memory available at this point, the kernel will crash and burn because
memblock_phys_alloc_range returns 0 on failure, which leads memblock_phys_free
to throw the first 4 MiB of physical memory to the wolves.  At a minimum it
should fail gracefully with a meaningful diagnostic, but in fact everything
seems to work fine without the weird reserve allocation.

---
  arch/x86/mm/init.c | 9 +++++++--
  1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/arch/x86/mm/init.c b/arch/x86/mm/init.c
index eb503f5..3696770 100644
--- a/arch/x86/mm/init.c
+++ b/arch/x86/mm/init.c
@@ -640,8 +640,13 @@ static void __init memory_map_top_down(unsigned long 
map_start,
  	 */
  	addr = memblock_phys_alloc_range(PMD_SIZE, PMD_SIZE, map_start,
  					 map_end);
-	memblock_phys_free(addr, PMD_SIZE);
-	real_end = addr + PMD_SIZE;
+	if (unlikely(addr < map_start)) {
+		pr_warn("Failed to release memory for alloc_low_pages()");
+		real_end = ALIGN_DOWN(map_end, PMD_SIZE);
+	} else {
+		memblock_phys_free(addr, PMD_SIZE);
+		real_end = addr + PMD_SIZE;
+	}

  	/* step_size need to be small so pgt_buf from BRK could cover it 
*/
  	step_size = PMD_SIZE;
-- 
2.34.0

* gldrk <me@rarity.fan> wrote:

> At least with CONFIG_PHYSICAL_START=0x100000, if there is < 4 MiB of contiguous
> free memory available at this point, the kernel will crash and burn because
> memblock_phys_alloc_range returns 0 on failure, which leads memblock_phys_free
> to throw the first 4 MiB of physical memory to the wolves.  At a minimum it
> should fail gracefully with a meaningful diagnostic, but in fact everything
> seems to work fine without the weird reserve allocation.
> 
> ---
>  arch/x86/mm/init.c | 9 +++++++--
>  1 file changed, 7 insertions(+), 2 deletions(-)
> 
> diff --git a/arch/x86/mm/init.c b/arch/x86/mm/init.c
> index eb503f5..3696770 100644
> --- a/arch/x86/mm/init.c
> +++ b/arch/x86/mm/init.c
> @@ -640,8 +640,13 @@ static void __init memory_map_top_down(unsigned long
> map_start,
>  	 */
>  	addr = memblock_phys_alloc_range(PMD_SIZE, PMD_SIZE, map_start,
>  					 map_end);
> -	memblock_phys_free(addr, PMD_SIZE);
> -	real_end = addr + PMD_SIZE;
> +	if (unlikely(addr < map_start)) {
> +		pr_warn("Failed to release memory for alloc_low_pages()");
> +		real_end = ALIGN_DOWN(map_end, PMD_SIZE);
> +	} else {
> +		memblock_phys_free(addr, PMD_SIZE);
> +		real_end = addr + PMD_SIZE;
> +	}

Makes sense to fix this bug I suppose, but the usual error check 
pattern for memblock_phys_alloc_range() failure would not be 'addr < 
map_start' but !addr.

( If memblock_phys_alloc_range() succeeds but returns an address below 
  'map_start', that would be a different failure I guess. )

Also, no need to add the 'unlikely()' I suspect - this is early boot 
code, micro-performance of branching is immaterial.

Just curious: what type of system has < 4 MiB of contiguous free memory 
available in early boot? Or was it something intentionally constrained 
via qemu?

Thanks,

	Ingo

> Just curious: what type of system has < 4 MiB of contiguous free memory
> available in early boot? Or was it something intentionally constrained
> via qemu?

Yes, I was able to boot a basic Alpine system off virtiofs with a few MiB
total and a stripped-down config.  I happen to have a memory-starved 486
machine that is technically "supported", but it's not running Linux just
yet.

Here's an updated patch.

Signed-off-by: Philip Redkin <me@rarity.fan>
---
  arch/x86/mm/init.c | 9 +++++++--
  1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/arch/x86/mm/init.c b/arch/x86/mm/init.c
index eb503f5..6c4ec4f 100644
--- a/arch/x86/mm/init.c
+++ b/arch/x86/mm/init.c
@@ -640,8 +640,13 @@ static void __init memory_map_top_down(unsigned long map_start,
  	 */
  	addr = memblock_phys_alloc_range(PMD_SIZE, PMD_SIZE, map_start,
  					 map_end);
-	memblock_phys_free(addr, PMD_SIZE);
-	real_end = addr + PMD_SIZE;
+	if (!addr) {
+		pr_warn("Failed to release memory for alloc_low_pages()");
+		real_end = max(map_start, ALIGN_DOWN(map_end, PMD_SIZE));
+	} else {
+		memblock_phys_free(addr, PMD_SIZE);
+		real_end = addr + PMD_SIZE;
+	}

  	/* step_size need to be small so pgt_buf from BRK could cover it */
  	step_size = PMD_SIZE;
-- 
2.34.0