fs/proc/kcore.c: allow translation of physical memory addresses

When /proc/kcore is read an attempt to read the first two pages results in
HW-specific page swap on s390 and another (so called prefix) pages are
accessed instead.  That leads to a wrong read.

Allow architecture-specific translation of memory addresses using
kc_xlate_dev_mem_ptr() and kc_unxlate_dev_mem_ptr() callbacks similarily
to /dev/mem xlate_dev_mem_ptr() and unxlate_dev_mem_ptr() callbacks.  That
way an architecture can deal with specific physical memory ranges.

Re-use the existing /dev/mem callback implementation on s390, which
handles the described prefix pages swapping correctly.

For other architectures the default callback is basically NOP.  It is
expected the condition (vaddr == __va(__pa(vaddr))) always holds true for
KCORE_RAM memory type.

Link: https://lkml.kernel.org/r/20240930122119.1651546-1-agordeev@linux.ibm.com
Signed-off-by: Alexander Gordeev <agordeev@linux.ibm.com>
Suggested-by: Heiko Carstens <hca@linux.ibm.com>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

arch/s390/include/asm/io.h

@@ -16,8 +16,10 @@
 #include <asm/pci_io.h>
 
 #define xlate_dev_mem_ptr xlate_dev_mem_ptr
+#define kc_xlate_dev_mem_ptr xlate_dev_mem_ptr
 void *xlate_dev_mem_ptr(phys_addr_t phys);
 #define unxlate_dev_mem_ptr unxlate_dev_mem_ptr
+#define kc_unxlate_dev_mem_ptr unxlate_dev_mem_ptr
 void unxlate_dev_mem_ptr(phys_addr_t phys, void *addr);
 
 #define IO_SPACE_LIMIT 0

fs/proc/kcore.c

@@ -50,6 +50,20 @@ static struct proc_dir_entry *proc_root_kcore;
 #define	kc_offset_to_vaddr(o) ((o) + PAGE_OFFSET)
 #endif
 
+#ifndef kc_xlate_dev_mem_ptr
+#define kc_xlate_dev_mem_ptr kc_xlate_dev_mem_ptr
+static inline void *kc_xlate_dev_mem_ptr(phys_addr_t phys)
+{
+	return __va(phys);
+}
+#endif
+#ifndef kc_unxlate_dev_mem_ptr
+#define kc_unxlate_dev_mem_ptr kc_unxlate_dev_mem_ptr
+static inline void kc_unxlate_dev_mem_ptr(phys_addr_t phys, void *virt)
+{
+}
+#endif
+
 static LIST_HEAD(kclist_head);
 static DECLARE_RWSEM(kclist_lock);
 static int kcore_need_update = 1;
@@ -471,6 +485,8 @@ static ssize_t read_kcore_iter(struct kiocb *iocb, struct iov_iter *iter)
 	while (buflen) {
 		struct page *page;
 		unsigned long pfn;
+		phys_addr_t phys;
+		void *__start;
 
 		/*
 		 * If this is the first iteration or the address is not within
@@ -537,7 +553,8 @@ static ssize_t read_kcore_iter(struct kiocb *iocb, struct iov_iter *iter)
 			}
 			break;
 		case KCORE_RAM:
-			pfn = __pa(start) >> PAGE_SHIFT;
+			phys = __pa(start);
+			pfn =  phys >> PAGE_SHIFT;
 			page = pfn_to_online_page(pfn);
 
 			/*
@@ -557,13 +574,28 @@ static ssize_t read_kcore_iter(struct kiocb *iocb, struct iov_iter *iter)
 			fallthrough;
 		case KCORE_VMEMMAP:
 		case KCORE_TEXT:
+			if (m->type == KCORE_RAM) {
+				__start = kc_xlate_dev_mem_ptr(phys);
+				if (!__start) {
+					ret = -ENOMEM;
+					if (iov_iter_zero(tsz, iter) != tsz)
+						ret = -EFAULT;
+					goto out;
+				}
+			} else {
+				__start = (void *)start;
+			}
+
 			/*
 			 * Sadly we must use a bounce buffer here to be able to
 			 * make use of copy_from_kernel_nofault(), as these
 			 * memory regions might not always be mapped on all
 			 * architectures.
 			 */
-			if (copy_from_kernel_nofault(buf, (void *)start, tsz)) {
+			ret = copy_from_kernel_nofault(buf, __start, tsz);
+			if (m->type == KCORE_RAM)
+				kc_unxlate_dev_mem_ptr(phys, __start);
+			if (ret) {
 				if (iov_iter_zero(tsz, iter) != tsz) {
 					ret = -EFAULT;
 					goto out;