Thread (30 messages) 30 messages, 2 authors, 2019-10-15

Re: [PATCH v6 08/17] arm64: hibernate: add trans_pgd public functions

flat view

From: James Morse <james.morse@arm.com>
Date: 2019-10-11 18:18:39
Also in: kexec, linux-mm, lkml

Hi Pavel,

On 04/10/2019 19:52, Pavel Tatashin wrote:
trans_pgd_create_copy() and trans_pgd_map_page() are going to be
the basis for new shared code that handles page tables for cases
which are between kernels: kexec, and hibernate.

Note: Eventually, get_safe_page() will be moved into a function pointer
passed via argument, but for now keep it as is.
quoted hunk ↗ jump to hunk
diff --git a/arch/arm64/kernel/hibernate.c b/arch/arm64/kernel/hibernate.c
index ce60bceed357..ded9034b9d39 100644
--- a/arch/arm64/kernel/hibernate.c
+++ b/arch/arm64/kernel/hibernate.c
quoted hunk ↗ jump to hunk
@@ -242,6 +218,44 @@ static int create_safe_exec_page(void *src_start, size_t length,
quoted hunk ↗ jump to hunk
+/*
+ * Copies length bytes, starting at src_start into an new page,
+ * perform cache maintenance, then maps it at the specified address low
+ * address as executable.
+ *
+ * This is used by hibernate to copy the code it needs to execute when
+ * overwriting the kernel text. This function generates a new set of page
+ * tables, which it loads into ttbr0.
+ *
+ * Length is provided as we probably only want 4K of data, even on a 64K
+ * page system.
+ */
+static int create_safe_exec_page(void *src_start, size_t length,
+				 unsigned long dst_addr,
+				 phys_addr_t *phys_dst_addr)
+{
+	void *page = (void *)get_safe_page(GFP_ATOMIC);
+	pgd_t *trans_pgd;
+	int rc;
+
+	if (!page)
+		return -ENOMEM;
+
+	memcpy(page, src_start, length);
+	__flush_icache_range((unsigned long)page, (unsigned long)page + length);
+
+	trans_pgd = (void *)get_safe_page(GFP_ATOMIC);
+	if (!trans_pgd)
+		return -ENOMEM;
+
+	rc = trans_pgd_map_page(trans_pgd, page, dst_addr,
+				PAGE_KERNEL_EXEC);
+	if (rc)
+		return rc;
+
 	/*
 	 * Load our new page tables. A strict BBM approach requires that we
 	 * ensure that TLBs are free of any entries that may overlap with the
(I suspect you are going to to duplicate this in the kexec code. Kexec has the same
pattern: instructions that have to be copied to do the relocation of the rest of memory)

quoted hunk ↗ jump to hunk
@@ -462,6 +476,24 @@ static int copy_page_tables(pgd_t *dst_pgdp, unsigned long start,
quoted hunk ↗ jump to hunk
+int trans_pgd_create_copy(pgd_t **dst_pgdp, unsigned long start,
+			  unsigned long end)
+{
+	int rc;
+	pgd_t *trans_pgd = (pgd_t *)get_safe_page(GFP_ATOMIC);
+
+	if (!trans_pgd) {
+		pr_err("Failed to allocate memory for temporary page tables.\n");
+		return -ENOMEM;
+	}
+
+	rc = copy_page_tables(trans_pgd, start, end);
+	if (!rc)
+		*dst_pgdp = trans_pgd;
*dst_pgdp was already allocated in swsusp_arch_resume().

quoted hunk ↗ jump to hunk
+
+	return rc;
+}
+
 /*
  * Setup then Resume from the hibernate image using swsusp_arch_suspend_exit().
  *
@@ -488,7 +520,7 @@ int swsusp_arch_resume(void)
 		pr_err("Failed to allocate memory for temporary page tables.\n");
 		return -ENOMEM;
 	}
If the allocation moves into 'trans_pgd_create_copy()', please move the code just above
here (cut off by the diff) that allocates it in swsusp_arch_resume().

Its actually okay to leak memory like this, hibernate's allocator acts as a memory pool.
It either gets freed if we fail to resume, or vanishes when the resumed kernel takes over.

Reviewed-by: James Morse <james.morse@arm.com>

quoted hunk ↗ jump to hunk
-	rc = copy_page_tables(tmp_pg_dir, PAGE_OFFSET, PAGE_END);
+	rc = trans_pgd_create_copy(&tmp_pg_dir, PAGE_OFFSET, PAGE_END);
 	if (rc)
 		return rc;

Thanks,

James

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help