riscv: Properly export reserved regions in /proc/iomem

The /proc/iomem represents the kernel's memory map. Regions marked
with "Reserved" tells the user that the range should not be tampered
with. Kexec-tools, when using the older kexec_load syscall relies on
the "Reserved" regions to build the memory segments, that will be the
target of the new kexec'd kernel.

The RISC-V port tries to expose all reserved regions to userland, but
some regions were not properly exposed: Regions that resided in both
the "regular" and reserved memory block, e.g. the EFI Memory Map. A
missing entry could result in reserved memory being overwritten.

It turns out, that arm64, and loongarch had a similar issue a while
back:

  commit d91680e687f4 ("arm64: Fix /proc/iomem for reserved but not memory regions")
  commit 50d7ba36b916 ("arm64: export memblock_reserve()d regions via /proc/iomem")

Similar to the other ports, resolve the issue by splitting the regions
in an arch initcall, since we need a working allocator.

Fixes: ffe0e5261268 ("RISC-V: Improve init_resources()")
Signed-off-by: Björn Töpel <bjorn@rivosinc.com>
Reviewed-by: Alexandre Ghiti <alexghiti@rivosinc.com>
Link: https://lore.kernel.org/r/20250409182129.634415-1-bjorn@kernel.org
Signed-off-by: Alexandre Ghiti <alexghiti@rivosinc.com>

diff --git a/arch/riscv/kernel/setup.c b/arch/riscv/kernel/setup.c
index c174544eefc8eeb4abd01737925d192b9b099664..f7c9a1caa83e62a7999d82cfa24a9345c3046351 100644 (file)
--- a/arch/riscv/kernel/setup.c
+++ b/arch/riscv/kernel/setup.c
@@ -66,6 +66,9 @@ static struct resource bss_res = { .name = "Kernel bss", };
 static struct resource elfcorehdr_res = { .name = "ELF Core hdr", };
 #endif
 
+static int num_standard_resources;
+static struct resource *standard_resources;
+
 static int __init add_resource(struct resource *parent,
                                struct resource *res)
 {
@@ -139,7 +142,7 @@ static void __init init_resources(void)
        struct resource *res = NULL;
        struct resource *mem_res = NULL;
        size_t mem_res_sz = 0;
-       int num_resources = 0, res_idx = 0;
+       int num_resources = 0, res_idx = 0, non_resv_res = 0;
        int ret = 0;
 
        /* + 1 as memblock_alloc() might increase memblock.reserved.cnt */
@@ -193,6 +196,7 @@ static void __init init_resources(void)
        /* Add /memory regions to the resource tree */
        for_each_mem_region(region) {
                res = &mem_res[res_idx--];
+               non_resv_res++;
 
                if (unlikely(memblock_is_nomap(region))) {
                        res->name = "Reserved";
@@ -210,6 +214,9 @@ static void __init init_resources(void)
                        goto error;
        }
 
+       num_standard_resources = non_resv_res;
+       standard_resources = &mem_res[res_idx + 1];
+
        /* Clean-up any unused pre-allocated resources */
        if (res_idx >= 0)
                memblock_free(mem_res, (res_idx + 1) * sizeof(*mem_res));
@@ -221,6 +228,33 @@ static void __init init_resources(void)
        memblock_free(mem_res, mem_res_sz);
 }
 
+static int __init reserve_memblock_reserved_regions(void)
+{
+       u64 i, j;
+
+       for (i = 0; i < num_standard_resources; i++) {
+               struct resource *mem = &standard_resources[i];
+               phys_addr_t r_start, r_end, mem_size = resource_size(mem);
+
+               if (!memblock_is_region_reserved(mem->start, mem_size))
+                       continue;
+
+               for_each_reserved_mem_range(j, &r_start, &r_end) {
+                       resource_size_t start, end;
+
+                       start = max(PFN_PHYS(PFN_DOWN(r_start)), mem->start);
+                       end = min(PFN_PHYS(PFN_UP(r_end)) - 1, mem->end);
+
+                       if (start > mem->end || end < mem->start)
+                               continue;
+
+                       reserve_region_with_split(mem, start, end, "Reserved");
+               }
+       }
+
+       return 0;
+}
+arch_initcall(reserve_memblock_reserved_regions);
 
 static void __init parse_dtb(void)
 {