From: Frank van der Linden <fvdl@google.com>
To: Andrew Morton <akpm@linux-foundation.org>, linux-mm@kvack.org
Cc: Frank van der Linden <fvdl@google.com>,
Anshuman Khandual <anshuman.khandual@arm.com>
Subject: [PATCH v3] mm/debug: use valid physical memory for pmd/pud tests
Date: Tue, 10 Jan 2023 18:12:08 +0000 [thread overview]
Message-ID: <20230110181208.1633879-1-fvdl@google.com> (raw)
The page table debug tests need a physical address to validate
low-level page table manipulation with. The memory at this address
is not actually touched, it just encoded in the page table entries
at various levels during the tests only.
Since the memory is not used, the code just picks the physical
address of the start_kernel symbol. This value is then truncated
to get a properly aligned address that is to be used for various
tests. Because of the truncation, the address might not actually
exist, or might not describe a complete huge page. That's not a
problem for most tests, but the arch-specific code may check
for attribute validity and consistency. The x86 version of
{pud,pmd}_set_huge actually validates the MTRRs for the PMD/PUD
range. This may fail with an address derived from start_kernel,
depending on where the kernel was loaded and what the physical
memory layout of the system is. This then leads to false negatives
for the {pud,pmd}_set_huge tests.
Avoid this by finding a properly aligned memory range that exists
and is usable. If such a range is not found, skip the tests that
needed it.
Fixes: 399145f9eb6c ("mm/debug: add tests validating architecture page table helpers")
Cc: Anshuman Khandual <anshuman.khandual@arm.com>
Signed-off-by: Frank van der Linden <fvdl@google.com>
---
mm/debug_vm_pgtable.c | 102 ++++++++++++++++++++++++++++++++++--------
1 file changed, 83 insertions(+), 19 deletions(-)
diff --git a/mm/debug_vm_pgtable.c b/mm/debug_vm_pgtable.c
index c631ade3f1d2..bb3328f46126 100644
--- a/mm/debug_vm_pgtable.c
+++ b/mm/debug_vm_pgtable.c
@@ -15,6 +15,7 @@
#include <linux/hugetlb.h>
#include <linux/kernel.h>
#include <linux/kconfig.h>
+#include <linux/memblock.h>
#include <linux/mm.h>
#include <linux/mman.h>
#include <linux/mm_types.h>
@@ -80,6 +81,7 @@ struct pgtable_debug_args {
unsigned long pmd_pfn;
unsigned long pte_pfn;
+ unsigned long fixed_alignment;
unsigned long fixed_pgd_pfn;
unsigned long fixed_p4d_pfn;
unsigned long fixed_pud_pfn;
@@ -430,7 +432,8 @@ static void __init pmd_huge_tests(struct pgtable_debug_args *args)
{
pmd_t pmd;
- if (!arch_vmap_pmd_supported(args->page_prot))
+ if (!arch_vmap_pmd_supported(args->page_prot) ||
+ args->fixed_alignment < PMD_SIZE)
return;
pr_debug("Validating PMD huge\n");
@@ -449,7 +452,8 @@ static void __init pud_huge_tests(struct pgtable_debug_args *args)
{
pud_t pud;
- if (!arch_vmap_pud_supported(args->page_prot))
+ if (!arch_vmap_pud_supported(args->page_prot) ||
+ args->fixed_alignment < PUD_SIZE)
return;
pr_debug("Validating PUD huge\n");
@@ -1077,10 +1081,85 @@ debug_vm_pgtable_alloc_huge_page(struct pgtable_debug_args *args, int order)
return page;
}
+/*
+ * Check if a physical memory range described by <pstart, pend> contains
+ * an area that is of size psize, and aligned to psize.
+ *
+ * Don't use address 0, an all-zeroes physical address might mask bugs, and
+ * it's not used on x86.
+ */
+static void __init phys_align_check(phys_addr_t pstart,
+ phys_addr_t pend, unsigned long psize,
+ phys_addr_t *physp, unsigned long *alignp)
+{
+ phys_addr_t aligned_start, aligned_end;
+
+ if (pstart == 0)
+ pstart = PAGE_SIZE;
+
+ aligned_start = ALIGN(pstart, psize);
+ aligned_end = aligned_start + psize;
+
+ if (aligned_end > aligned_start && aligned_end <= pend) {
+ *alignp = psize;
+ *physp = aligned_start;
+ }
+}
+
+static void __init init_fixed_pfns(struct pgtable_debug_args *args)
+{
+ u64 idx;
+ phys_addr_t phys, pstart, pend;
+
+ /*
+ * Initialize the fixed pfns. To do this, try to find a
+ * valid physical range, preferably aligned to PUD_SIZE,
+ * but settling for aligned to PMD_SIZE as a fallback. If
+ * neither of those is found, use the physical address of
+ * the start_kernel symbol.
+ *
+ * The memory doesn't need to be allocated, it just needs to exist
+ * as usable memory. It won't be touched.
+ *
+ * The alignment is recorded, and can be checked to see if we
+ * can run the tests that require an actual valid physical
+ * address range on some architectures ({pmd,pud}_huge_test
+ * on x86).
+ */
+
+ phys = __pa_symbol(&start_kernel);
+ args->fixed_alignment = PAGE_SIZE;
+
+ for_each_mem_range(idx, &pstart, &pend) {
+ /* First check for a PUD-aligned area */
+ phys_align_check(pstart, pend, PUD_SIZE, &phys,
+ &args->fixed_alignment);
+
+ /* If a PUD-aligned area is found, we're done */
+ if (args->fixed_alignment == PUD_SIZE)
+ break;
+
+ /*
+ * If no PMD-aligned area found yet, check for one,
+ * but continue the loop to look for a PUD-aligned area.
+ */
+ if (args->fixed_alignment < PMD_SIZE)
+ phys_align_check(pstart, pend, PMD_SIZE, &phys,
+ &args->fixed_alignment);
+ }
+
+ args->fixed_pgd_pfn = __phys_to_pfn(phys & PGDIR_MASK);
+ args->fixed_p4d_pfn = __phys_to_pfn(phys & P4D_MASK);
+ args->fixed_pud_pfn = __phys_to_pfn(phys & PUD_MASK);
+ args->fixed_pmd_pfn = __phys_to_pfn(phys & PMD_MASK);
+ args->fixed_pte_pfn = __phys_to_pfn(phys & PAGE_MASK);
+ WARN_ON(!pfn_valid(args->fixed_pte_pfn));
+}
+
+
static int __init init_args(struct pgtable_debug_args *args)
{
struct page *page = NULL;
- phys_addr_t phys;
int ret = 0;
/*
@@ -1160,22 +1239,7 @@ static int __init init_args(struct pgtable_debug_args *args)
args->start_ptep = pmd_pgtable(READ_ONCE(*args->pmdp));
WARN_ON(!args->start_ptep);
- /*
- * PFN for mapping at PTE level is determined from a standard kernel
- * text symbol. But pfns for higher page table levels are derived by
- * masking lower bits of this real pfn. These derived pfns might not
- * exist on the platform but that does not really matter as pfn_pxx()
- * helpers will still create appropriate entries for the test. This
- * helps avoid large memory block allocations to be used for mapping
- * at higher page table levels in some of the tests.
- */
- phys = __pa_symbol(&start_kernel);
- args->fixed_pgd_pfn = __phys_to_pfn(phys & PGDIR_MASK);
- args->fixed_p4d_pfn = __phys_to_pfn(phys & P4D_MASK);
- args->fixed_pud_pfn = __phys_to_pfn(phys & PUD_MASK);
- args->fixed_pmd_pfn = __phys_to_pfn(phys & PMD_MASK);
- args->fixed_pte_pfn = __phys_to_pfn(phys & PAGE_MASK);
- WARN_ON(!pfn_valid(args->fixed_pte_pfn));
+ init_fixed_pfns(args);
/*
* Allocate (huge) pages because some of the tests need to access
--
2.39.0.314.g84b9a713c41-goog
next reply other threads:[~2023-01-10 18:12 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-01-10 18:12 Frank van der Linden [this message]
2023-01-11 4:38 ` Anshuman Khandual
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20230110181208.1633879-1-fvdl@google.com \
--to=fvdl@google.com \
--cc=akpm@linux-foundation.org \
--cc=anshuman.khandual@arm.com \
--cc=linux-mm@kvack.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox