Merge branch 'decompressor-memory-tracking' into features

Vasily Gorbik says:

===================
Combine and generalize all methods for finding unused memory in
decompressor, while decreasing complexity, add memory holes support,
while improving error handling (especially in low-memory conditions)
and debug-ability.
===================

Signed-off-by: Heiko Carstens <hca@linux.ibm.com>

23 files changed, 911 insertions, 1004 deletions

diff --git a/arch/s390/boot/Makefile b/arch/s390/boot/Makefile
index cebd4ca16916..c7c81e5f9218 100644
--- a/arch/s390/boot/Makefile
+++ b/arch/s390/boot/Makefile
@@ -35,7 +35,7 @@ endif
 
 CFLAGS_sclp_early_core.o += -I$(srctree)/drivers/s390/char
 
-obj-y	:= head.o als.o startup.o mem_detect.o ipl_parm.o ipl_report.o vmem.o
+obj-y	:= head.o als.o startup.o physmem_info.o ipl_parm.o ipl_report.o vmem.o
 obj-y	+= string.o ebcdic.o sclp_early_core.o mem.o ipl_vmparm.o cmdline.o
 obj-y	+= version.o pgm_check_info.o ctype.o ipl_data.o machine_kexec_reloc.o
 obj-$(findstring y, $(CONFIG_PROTECTED_VIRTUALIZATION_GUEST) $(CONFIG_PGSTE))	+= uv.o
diff --git a/arch/s390/boot/boot.h b/arch/s390/boot/boot.h
index 58ce701d6110..872963c8a0ab 100644
--- a/arch/s390/boot/boot.h
+++ b/arch/s390/boot/boot.h
@@ -8,6 +8,8 @@
 
 #ifndef __ASSEMBLY__
 
+#include <asm/physmem_info.h>
+
 struct machine_info {
 	unsigned char has_edat1 : 1;
 	unsigned char has_edat2 : 1;
@@ -30,24 +32,44 @@ struct vmlinux_info {
 	unsigned long init_mm_off;
 	unsigned long swapper_pg_dir_off;
 	unsigned long invalid_pg_dir_off;
+#ifdef CONFIG_KASAN
+	unsigned long kasan_early_shadow_page_off;
+	unsigned long kasan_early_shadow_pte_off;
+	unsigned long kasan_early_shadow_pmd_off;
+	unsigned long kasan_early_shadow_pud_off;
+	unsigned long kasan_early_shadow_p4d_off;
+#endif
 };
 
 void startup_kernel(void);
-unsigned long detect_memory(unsigned long *safe_addr);
-void mem_detect_set_usable_limit(unsigned long limit);
+unsigned long detect_max_physmem_end(void);
+void detect_physmem_online_ranges(unsigned long max_physmem_end);
+void physmem_set_usable_limit(unsigned long limit);
+void physmem_reserve(enum reserved_range_type type, unsigned long addr, unsigned long size);
+void physmem_free(enum reserved_range_type type);
+/* for continuous/multiple allocations per type */
+unsigned long physmem_alloc_top_down(enum reserved_range_type type, unsigned long size,
+				     unsigned long align);
+/* for single allocations, 1 per type */
+unsigned long physmem_alloc_range(enum reserved_range_type type, unsigned long size,
+				  unsigned long align, unsigned long min, unsigned long max,
+				  bool die_on_oom);
+bool ipl_report_certs_intersects(unsigned long addr, unsigned long size,
+				 unsigned long *intersection_start);
 bool is_ipl_block_dump(void);
 void store_ipl_parmblock(void);
-unsigned long read_ipl_report(unsigned long safe_addr);
+int read_ipl_report(void);
+void save_ipl_cert_comp_list(void);
 void setup_boot_command_line(void);
 void parse_boot_command_line(void);
 void verify_facilities(void);
 void print_missing_facilities(void);
 void sclp_early_setup_buffer(void);
 void print_pgm_check_info(void);
-unsigned long get_random_base(unsigned long safe_addr);
+unsigned long get_random_base(void);
 void setup_vmem(unsigned long asce_limit);
-unsigned long vmem_estimate_memory_needs(unsigned long online_mem_total);
 void __printf(1, 2) decompressor_printk(const char *fmt, ...);
+void print_stacktrace(unsigned long sp);
 void error(char *m);
 
 extern struct machine_info machine;
@@ -62,7 +84,7 @@ extern char __boot_data_start[], __boot_data_end[];
 extern char __boot_data_preserved_start[], __boot_data_preserved_end[];
 extern char _decompressor_syms_start[], _decompressor_syms_end[];
 extern char _stack_start[], _stack_end[];
-extern char _end[];
+extern char _end[], _decompressor_end[];
 extern unsigned char _compressed_start[];
 extern unsigned char _compressed_end[];
 extern struct vmlinux_info _vmlinux_info;
@@ -70,5 +92,10 @@ extern struct vmlinux_info _vmlinux_info;
 
 #define __abs_lowcore_pa(x)	(((unsigned long)(x) - __abs_lowcore) % sizeof(struct lowcore))
 
+static inline bool intersects(unsigned long addr0, unsigned long size0,
+			      unsigned long addr1, unsigned long size1)
+{
+	return addr0 + size0 > addr1 && addr1 + size1 > addr0;
+}
 #endif /* __ASSEMBLY__ */
 #endif /* BOOT_BOOT_H */
diff --git a/arch/s390/boot/ipl_report.c b/arch/s390/boot/ipl_report.c
index 9b14045065b6..1803035e68d2 100644
--- a/arch/s390/boot/ipl_report.c
+++ b/arch/s390/boot/ipl_report.c
@@ -5,6 +5,7 @@
 #include <asm/sclp.h>
 #include <asm/sections.h>
 #include <asm/boot_data.h>
+#include <asm/physmem_info.h>
 #include <uapi/asm/ipl.h>
 #include "boot.h"
 
@@ -16,20 +17,16 @@ unsigned long __bootdata_preserved(ipl_cert_list_size);
 unsigned long __bootdata(early_ipl_comp_list_addr);
 unsigned long __bootdata(early_ipl_comp_list_size);
 
+static struct ipl_rb_certificates *certs;
+static struct ipl_rb_components *comps;
+static bool ipl_report_needs_saving;
+
 #define for_each_rb_entry(entry, rb) \
 	for (entry = rb->entries; \
 	     (void *) entry + sizeof(*entry) <= (void *) rb + rb->len; \
 	     entry++)
 
-static inline bool intersects(unsigned long addr0, unsigned long size0,
-			      unsigned long addr1, unsigned long size1)
-{
-	return addr0 + size0 > addr1 && addr1 + size1 > addr0;
-}
-
-static unsigned long find_bootdata_space(struct ipl_rb_components *comps,
-					 struct ipl_rb_certificates *certs,
-					 unsigned long safe_addr)
+static unsigned long get_cert_comp_list_size(void)
 {
 	struct ipl_rb_certificate_entry *cert;
 	struct ipl_rb_component_entry *comp;
@@ -44,36 +41,27 @@ static unsigned long find_bootdata_space(struct ipl_rb_components *comps,
 	ipl_cert_list_size = 0;
 	for_each_rb_entry(cert, certs)
 		ipl_cert_list_size += sizeof(unsigned int) + cert->len;
-	size = ipl_cert_list_size + early_ipl_comp_list_size;
+	return ipl_cert_list_size + early_ipl_comp_list_size;
+}
 
-	/*
-	 * Start from safe_addr to find a free memory area large
-	 * enough for the IPL report boot data. This area is used
-	 * for ipl_cert_list_addr/ipl_cert_list_size and
-	 * early_ipl_comp_list_addr/early_ipl_comp_list_size. It must
-	 * not overlap with any component or any certificate.
-	 */
-repeat:
-	if (IS_ENABLED(CONFIG_BLK_DEV_INITRD) && initrd_data.start && initrd_data.size &&
-	    intersects(initrd_data.start, initrd_data.size, safe_addr, size))
-		safe_addr = initrd_data.start + initrd_data.size;
-	for_each_rb_entry(comp, comps)
-		if (intersects(safe_addr, size, comp->addr, comp->len)) {
-			safe_addr = comp->addr + comp->len;
-			goto repeat;
-		}
-	for_each_rb_entry(cert, certs)
-		if (intersects(safe_addr, size, cert->addr, cert->len)) {
-			safe_addr = cert->addr + cert->len;
-			goto repeat;
-		}
-	early_ipl_comp_list_addr = safe_addr;
-	ipl_cert_list_addr = safe_addr + early_ipl_comp_list_size;
+bool ipl_report_certs_intersects(unsigned long addr, unsigned long size,
+				 unsigned long *intersection_start)
+{
+	struct ipl_rb_certificate_entry *cert;
 
-	return safe_addr + size;
+	if (!ipl_report_needs_saving)
+		return false;
+
+	for_each_rb_entry(cert, certs) {
+		if (intersects(addr, size, cert->addr, cert->len)) {
+			*intersection_start = cert->addr;
+			return true;
+		}
+	}
+	return false;
 }
 
-static void copy_components_bootdata(struct ipl_rb_components *comps)
+static void copy_components_bootdata(void)
 {
 	struct ipl_rb_component_entry *comp, *ptr;
 
@@ -82,7 +70,7 @@ static void copy_components_bootdata(struct ipl_rb_components *comps)
 		memcpy(ptr++, comp, sizeof(*ptr));
 }
 
-static void copy_certificates_bootdata(struct ipl_rb_certificates *certs)
+static void copy_certificates_bootdata(void)
 {
 	struct ipl_rb_certificate_entry *cert;
 	void *ptr;
@@ -96,10 +84,8 @@ static void copy_certificates_bootdata(struct ipl_rb_certificates *certs)
 	}
 }
 
-unsigned long read_ipl_report(unsigned long safe_addr)
+int read_ipl_report(void)
 {
-	struct ipl_rb_certificates *certs;
-	struct ipl_rb_components *comps;
 	struct ipl_pl_hdr *pl_hdr;
 	struct ipl_rl_hdr *rl_hdr;
 	struct ipl_rb_hdr *rb_hdr;
@@ -112,7 +98,7 @@ unsigned long read_ipl_report(unsigned long safe_addr)
 	 */
 	if (!ipl_block_valid ||
 	    !(ipl_block.hdr.flags & IPL_PL_FLAG_IPLSR))
-		return safe_addr;
+		return -1;
 	ipl_secure_flag = !!(ipl_block.hdr.flags & IPL_PL_FLAG_SIPL);
 	/*
 	 * There is an IPL report, to find it load the pointer to the
@@ -150,16 +136,30 @@ unsigned long read_ipl_report(unsigned long safe_addr)
 	 * With either the component list or the certificate list
 	 * missing the kernel will stay ignorant of secure IPL.
 	 */
-	if (!comps || !certs)
-		return safe_addr;
+	if (!comps || !certs) {
+		certs = NULL;
+		return -1;
+	}
 
-	/*
-	 * Copy component and certificate list to a safe area
-	 * where the decompressed kernel can find them.
-	 */
-	safe_addr = find_bootdata_space(comps, certs, safe_addr);
-	copy_components_bootdata(comps);
-	copy_certificates_bootdata(certs);
+	ipl_report_needs_saving = true;
+	physmem_reserve(RR_IPLREPORT, (unsigned long)pl_hdr,
+			(unsigned long)rl_end - (unsigned long)pl_hdr);
+	return 0;
+}
+
+void save_ipl_cert_comp_list(void)
+{
+	unsigned long size;
+
+	if (!ipl_report_needs_saving)
+		return;
+
+	size = get_cert_comp_list_size();
+	early_ipl_comp_list_addr = physmem_alloc_top_down(RR_CERT_COMP_LIST, size, sizeof(int));
+	ipl_cert_list_addr = early_ipl_comp_list_addr + early_ipl_comp_list_size;
 
-	return safe_addr;
+	copy_components_bootdata();
+	copy_certificates_bootdata();
+	physmem_free(RR_IPLREPORT);
+	ipl_report_needs_saving = false;
 }
diff --git a/arch/s390/boot/kaslr.c b/arch/s390/boot/kaslr.c
index 3e3d846400b4..71f75f03f800 100644
--- a/arch/s390/boot/kaslr.c
+++ b/arch/s390/boot/kaslr.c
@@ -3,7 +3,7 @@
  * Copyright IBM Corp. 2019
  */
 #include <linux/pgtable.h>
-#include <asm/mem_detect.h>
+#include <asm/physmem_info.h>
 #include <asm/cpacf.h>
 #include <asm/timex.h>
 #include <asm/sclp.h>
@@ -91,113 +91,16 @@ static int get_random(unsigned long limit, unsigned long *value)
 	return 0;
 }
 
-/*
- * To randomize kernel base address we have to consider several facts:
- * 1. physical online memory might not be continuous and have holes. mem_detect
- *    info contains list of online memory ranges we should consider.
- * 2. we have several memory regions which are occupied and we should not
- *    overlap and destroy them. Currently safe_addr tells us the border below
- *    which all those occupied regions are. We are safe to use anything above
- *    safe_addr.
- * 3. the upper limit might apply as well, even if memory above that limit is
- *    online. Currently those limitations are:
- *    3.1. Limit set by "mem=" kernel command line option
- *    3.2. memory reserved at the end for kasan initialization.
- * 4. kernel base address must be aligned to THREAD_SIZE (kernel stack size).
- *    Which is required for CONFIG_CHECK_STACK. Currently THREAD_SIZE is 4 pages
- *    (16 pages when the kernel is built with kasan enabled)
- * Assumptions:
- * 1. kernel size (including .bss size) and upper memory limit are page aligned.
- * 2. mem_detect memory region start is THREAD_SIZE aligned / end is PAGE_SIZE
- *    aligned (in practice memory configurations granularity on z/VM and LPAR
- *    is 1mb).
- *
- * To guarantee uniform distribution of kernel base address among all suitable
- * addresses we generate random value just once. For that we need to build a
- * continuous range in which every value would be suitable. We can build this
- * range by simply counting all suitable addresses (let's call them positions)
- * which would be valid as kernel base address. To count positions we iterate
- * over online memory ranges. For each range which is big enough for the
- * kernel image we count all suitable addresses we can put the kernel image at
- * that is
- * (end - start - kernel_size) / THREAD_SIZE + 1
- * Two functions count_valid_kernel_positions and position_to_address help
- * to count positions in memory range given and then convert position back
- * to address.
- */
-static unsigned long count_valid_kernel_positions(unsigned long kernel_size,
-						  unsigned long _min,
-						  unsigned long _max)
-{
-	unsigned long start, end, pos = 0;
-	int i;
-
-	for_each_mem_detect_usable_block(i, &start, &end) {
-		if (_min >= end)
-			continue;
-		if (start >= _max)
-			break;
-		start = max(_min, start);
-		end = min(_max, end);
-		if (end - start < kernel_size)
-			continue;
-		pos += (end - start - kernel_size) / THREAD_SIZE + 1;
-	}
-
-	return pos;
-}
-
-static unsigned long position_to_address(unsigned long pos, unsigned long kernel_size,
-				 unsigned long _min, unsigned long _max)
-{
-	unsigned long start, end;
-	int i;
-
-	for_each_mem_detect_usable_block(i, &start, &end) {
-		if (_min >= end)
-			continue;
-		if (start >= _max)
-			break;
-		start = max(_min, start);
-		end = min(_max, end);
-		if (end - start < kernel_size)
-			continue;
-		if ((end - start - kernel_size) / THREAD_SIZE + 1 >= pos)
-			return start + (pos - 1) * THREAD_SIZE;
-		pos -= (end - start - kernel_size) / THREAD_SIZE + 1;
-	}
-
-	return 0;
-}
-
-unsigned long get_random_base(unsigned long safe_addr)
+unsigned long get_random_base(void)
 {
-	unsigned long usable_total = get_mem_detect_usable_total();
-	unsigned long memory_limit = get_mem_detect_end();
-	unsigned long base_pos, max_pos, kernel_size;
-	int i;
+	unsigned long vmlinux_size = vmlinux.image_size + vmlinux.bss_size;
+	unsigned long minimal_pos = vmlinux.default_lma + vmlinux_size;
+	unsigned long random;
 
-	/*
-	 * Avoid putting kernel in the end of physical memory
-	 * which vmem and kasan code will use for shadow memory and
-	 * pgtable mapping allocations.
-	 */
-	memory_limit -= kasan_estimate_memory_needs(usable_total);
-	memory_limit -= vmem_estimate_memory_needs(usable_total);
-
-	safe_addr = ALIGN(safe_addr, THREAD_SIZE);
-	kernel_size = vmlinux.image_size + vmlinux.bss_size;
-	if (safe_addr + kernel_size > memory_limit)
+	/* [vmlinux.default_lma + vmlinux.image_size + vmlinux.bss_size : physmem_info.usable] */
+	if (get_random(physmem_info.usable - minimal_pos, &random))
 		return 0;
 
-	max_pos = count_valid_kernel_positions(kernel_size, safe_addr, memory_limit);
-	if (!max_pos) {
-		sclp_early_printk("KASLR disabled: not enough memory\n");
-		return 0;
-	}
-
-	/* we need a value in the range [1, base_pos] inclusive */
-	if (get_random(max_pos, &base_pos))
-		return 0;
-	return position_to_address(base_pos + 1, kernel_size, safe_addr, memory_limit);
+	return physmem_alloc_range(RR_VMLINUX, vmlinux_size, THREAD_SIZE,
+				   vmlinux.default_lma, minimal_pos + random, false);
 }
diff --git a/arch/s390/boot/mem_detect.c b/arch/s390/boot/mem_detect.c
deleted file mode 100644
index 35f4ba11f7fd..000000000000
--- a/arch/s390/boot/mem_detect.c
+++ /dev/null
@@ -1,191 +0,0 @@
-// SPDX-License-Identifier: GPL-2.0
-#include <linux/errno.h>
-#include <linux/init.h>
-#include <asm/setup.h>
-#include <asm/processor.h>
-#include <asm/sclp.h>
-#include <asm/sections.h>
-#include <asm/mem_detect.h>
-#include <asm/sparsemem.h>
-#include "decompressor.h"
-#include "boot.h"
-
-struct mem_detect_info __bootdata(mem_detect);
-
-/* up to 256 storage elements, 1020 subincrements each */
-#define ENTRIES_EXTENDED_MAX						       \
-	(256 * (1020 / 2) * sizeof(struct mem_detect_block))
-
-static struct mem_detect_block *__get_mem_detect_block_ptr(u32 n)
-{
-	if (n < MEM_INLINED_ENTRIES)
-		return &mem_detect.entries[n];
-	return &mem_detect.entries_extended[n - MEM_INLINED_ENTRIES];
-}
-
-/*
- * sequential calls to add_mem_detect_block with adjacent memory areas
- * are merged together into single memory block.
- */
-void add_mem_detect_block(u64 start, u64 end)
-{
-	struct mem_detect_block *block;
-
-	if (mem_detect.count) {
-		block = __get_mem_detect_block_ptr(mem_detect.count - 1);
-		if (block->end == start) {
-			block->end = end;
-			return;
-		}
-	}
-
-	block = __get_mem_detect_block_ptr(mem_detect.count);
-	block->start = start;
-	block->end = end;
-	mem_detect.count++;
-}
-
-static int __diag260(unsigned long rx1, unsigned long rx2)
-{
-	unsigned long reg1, reg2, ry;
-	union register_pair rx;
-	psw_t old;
-	int rc;
-
-	rx.even = rx1;
-	rx.odd	= rx2;
-	ry = 0x10; /* storage configuration */
-	rc = -1;   /* fail */
-	asm volatile(
-		"	mvc	0(16,%[psw_old]),0(%[psw_pgm])\n"
-		"	epsw	%[reg1],%[reg2]\n"
-		"	st	%[reg1],0(%[psw_pgm])\n"
-		"	st	%[reg2],4(%[psw_pgm])\n"
-		"	larl	%[reg1],1f\n"
-		"	stg	%[reg1],8(%[psw_pgm])\n"
-		"	diag	%[rx],%[ry],0x260\n"
-		"	ipm	%[rc]\n"
-		"	srl	%[rc],28\n"
-		"1:	mvc	0(16,%[psw_pgm]),0(%[psw_old])\n"
-		: [reg1] "=&d" (reg1),
-		  [reg2] "=&a" (reg2),
-		  [rc] "+&d" (rc),
-		  [ry] "+&d" (ry),
-		  "+Q" (S390_lowcore.program_new_psw),
-		  "=Q" (old)
-		: [rx] "d" (rx.pair),
-		  [psw_old] "a" (&old),
-		  [psw_pgm] "a" (&S390_lowcore.program_new_psw)
-		: "cc", "memory");
-	return rc == 0 ? ry : -1;
-}
-
-static int diag260(void)
-{
-	int rc, i;
-
-	struct {
-		unsigned long start;
-		unsigned long end;
-	} storage_extents[8] __aligned(16); /* VM supports up to 8 extends */
-
-	memset(storage_extents, 0, sizeof(storage_extents));
-	rc = __diag260((unsigned long)storage_extents, sizeof(storage_extents));
-	if (rc == -1)
-		return -1;
-
-	for (i = 0; i < min_t(int, rc, ARRAY_SIZE(storage_extents)); i++)
-		add_mem_detect_block(storage_extents[i].start, storage_extents[i].end + 1);
-	return 0;
-}
-
-static int tprot(unsigned long addr)
-{
-	unsigned long reg1, reg2;
-	int rc = -EFAULT;
-	psw_t old;
-
-	asm volatile(
-		"	mvc	0(16,%[psw_old]),0(%[psw_pgm])\n"
-		"	epsw	%[reg1],%[reg2]\n"
-		"	st	%[reg1],0(%[psw_pgm])\n"
-		"	st	%[reg2],4(%[psw_pgm])\n"
-		"	larl	%[reg1],1f\n"
-		"	stg	%[reg1],8(%[psw_pgm])\n"
-		"	tprot	0(%[addr]),0\n"
-		"	ipm	%[rc]\n"
-		"	srl	%[rc],28\n"
-		"1:	mvc	0(16,%[psw_pgm]),0(%[psw_old])\n"
-		: [reg1] "=&d" (reg1),
-		  [reg2] "=&a" (reg2),
-		  [rc] "+&d" (rc),
-		  "=Q" (S390_lowcore.program_new_psw.addr),
-		  "=Q" (old)
-		: [psw_old] "a" (&old),
-		  [psw_pgm] "a" (&S390_lowcore.program_new_psw),
-		  [addr] "a" (addr)
-		: "cc", "memory");
-	return rc;
-}
-
-static unsigned long search_mem_end(void)
-{
-	unsigned long range = 1 << (MAX_PHYSMEM_BITS - 20); /* in 1MB blocks */
-	unsigned long offset = 0;
-	unsigned long pivot;
-
-	while (range > 1) {
-		range >>= 1;
-		pivot = offset + range;
-		if (!tprot(pivot << 20))
-			offset = pivot;
-	}
-	return (offset + 1) << 20;
-}
-
-unsigned long detect_memory(unsigned long *safe_addr)
-{
-	unsigned long max_physmem_end = 0;
-
-	sclp_early_get_memsize(&max_physmem_end);
-	mem_detect.entries_extended = (struct mem_detect_block *)ALIGN(*safe_addr, sizeof(u64));
-
-	if (!sclp_early_read_storage_info()) {
-		mem_detect.info_source = MEM_DETECT_SCLP_STOR_INFO;
-	} else if (!diag260()) {
-		mem_detect.info_source = MEM_DETECT_DIAG260;
-		max_physmem_end = max_physmem_end ?: get_mem_detect_end();
-	} else if (max_physmem_end) {
-		add_mem_detect_block(0, max_physmem_end);
-		mem_detect.info_source = MEM_DETECT_SCLP_READ_INFO;
-	} else {
-		max_physmem_end = search_mem_end();
-		add_mem_detect_block(0, max_physmem_end);
-		mem_detect.info_source = MEM_DETECT_BIN_SEARCH;
-	}
-
-	if (mem_detect.count > MEM_INLINED_ENTRIES) {
-		*safe_addr += (mem_detect.count - MEM_INLINED_ENTRIES) *
-			     sizeof(struct mem_detect_block);
-	}
-
-	return max_physmem_end;
-}
-
-void mem_detect_set_usable_limit(unsigned long limit)
-{
-	struct mem_detect_block *block;
-	int i;
-
-	/* make sure mem_detect.usable ends up within online memory block */
-	for (i = 0; i < mem_detect.count; i++) {
-		block = __get_mem_detect_block_ptr(i);
-		if (block->start >= limit)
-			break;
-		if (block->end >= limit) {
-			mem_detect.usable = limit;
-			break;
-		}
-		mem_detect.usable = block->end;
-	}
-}
diff --git a/arch/s390/boot/pgm_check_info.c b/arch/s390/boot/pgm_check_info.c
index c2a1defc79da..0861e3c403f8 100644
--- a/arch/s390/boot/pgm_check_info.c
+++ b/arch/s390/boot/pgm_check_info.c
@@ -123,11 +123,10 @@ out:
 	sclp_early_printk(buf);
 }
 
-static noinline void print_stacktrace(void)
+void print_stacktrace(unsigned long sp)
 {
 	struct stack_info boot_stack = { STACK_TYPE_TASK, (unsigned long)_stack_start,
 					 (unsigned long)_stack_end };
-	unsigned long sp = S390_lowcore.gpregs_save_area[15];
 	bool first = true;
 
 	decompressor_printk("Call Trace:\n");
@@ -173,7 +172,7 @@ void print_pgm_check_info(void)
 			    gpregs[8], gpregs[9], gpregs[10], gpregs[11]);
 	decompressor_printk("      %016lx %016lx %016lx %016lx\n",
 			    gpregs[12], gpregs[13], gpregs[14], gpregs[15]);
-	print_stacktrace();
+	print_stacktrace(S390_lowcore.gpregs_save_area[15]);
 	decompressor_printk("Last Breaking-Event-Address:\n");
 	decompressor_printk(" [<%016lx>] %pS\n", (unsigned long)S390_lowcore.pgm_last_break,
 			    (void *)S390_lowcore.pgm_last_break);
diff --git a/arch/s390/boot/physmem_info.c b/arch/s390/boot/physmem_info.c
new file mode 100644
index 000000000000..4ee9b7381142
--- /dev/null
+++ b/arch/s390/boot/physmem_info.c
@@ -0,0 +1,323 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <linux/processor.h>
+#include <linux/errno.h>
+#include <linux/init.h>
+#include <asm/physmem_info.h>
+#include <asm/stacktrace.h>
+#include <asm/boot_data.h>
+#include <asm/sparsemem.h>
+#include <asm/sections.h>
+#include <asm/setup.h>
+#include <asm/sclp.h>
+#include <asm/uv.h>
+#include "decompressor.h"
+#include "boot.h"
+
+struct physmem_info __bootdata(physmem_info);
+static unsigned int physmem_alloc_ranges;
+static unsigned long physmem_alloc_pos;
+
+/* up to 256 storage elements, 1020 subincrements each */
+#define ENTRIES_EXTENDED_MAX						       \
+	(256 * (1020 / 2) * sizeof(struct physmem_range))
+
+static struct physmem_range *__get_physmem_range_ptr(u32 n)
+{
+	if (n < MEM_INLINED_ENTRIES)
+		return &physmem_info.online[n];
+	if (unlikely(!physmem_info.online_extended)) {
+		physmem_info.online_extended = (struct physmem_range *)physmem_alloc_range(
+			RR_MEM_DETECT_EXTENDED, ENTRIES_EXTENDED_MAX, sizeof(long), 0,
+			physmem_alloc_pos, true);
+	}
+	return &physmem_info.online_extended[n - MEM_INLINED_ENTRIES];
+}
+
+/*
+ * sequential calls to add_physmem_online_range with adjacent memory ranges
+ * are merged together into single memory range.
+ */
+void add_physmem_online_range(u64 start, u64 end)
+{
+	struct physmem_range *range;
+
+	if (physmem_info.range_count) {
+		range = __get_physmem_range_ptr(physmem_info.range_count - 1);
+		if (range->end == start) {
+			range->end = end;
+			return;
+		}
+	}
+
+	range = __get_physmem_range_ptr(physmem_info.range_count);
+	range->start = start;
+	range->end = end;
+	physmem_info.range_count++;
+}
+
+static int __diag260(unsigned long rx1, unsigned long rx2)
+{
+	unsigned long reg1, reg2, ry;
+	union register_pair rx;
+	psw_t old;
+	int rc;
+
+	rx.even = rx1;
+	rx.odd	= rx2;
+	ry = 0x10; /* storage configuration */
+	rc = -1;   /* fail */
+	asm volatile(
+		"	mvc	0(16,%[psw_old]),0(%[psw_pgm])\n"
+		"	epsw	%[reg1],%[reg2]\n"
+		"	st	%[reg1],0(%[psw_pgm])\n"
+		"	st	%[reg2],4(%[psw_pgm])\n"
+		"	larl	%[reg1],1f\n"
+		"	stg	%[reg1],8(%[psw_pgm])\n"
+		"	diag	%[rx],%[ry],0x260\n"
+		"	ipm	%[rc]\n"
+		"	srl	%[rc],28\n"
+		"1:	mvc	0(16,%[psw_pgm]),0(%[psw_old])\n"
+		: [reg1] "=&d" (reg1),
+		  [reg2] "=&a" (reg2),
+		  [rc] "+&d" (rc),
+		  [ry] "+&d" (ry),
+		  "+Q" (S390_lowcore.program_new_psw),
+		  "=Q" (old)
+		: [rx] "d" (rx.pair),
+		  [psw_old] "a" (&old),
+		  [psw_pgm] "a" (&S390_lowcore.program_new_psw)
+		: "cc", "memory");
+	return rc == 0 ? ry : -1;
+}
+
+static int diag260(void)
+{
+	int rc, i;
+
+	struct {
+		unsigned long start;
+		unsigned long end;
+	} storage_extents[8] __aligned(16); /* VM supports up to 8 extends */
+
+	memset(storage_extents, 0, sizeof(storage_extents));
+	rc = __diag260((unsigned long)storage_extents, sizeof(storage_extents));
+	if (rc == -1)
+		return -1;
+
+	for (i = 0; i < min_t(int, rc, ARRAY_SIZE(storage_extents)); i++)
+		add_physmem_online_range(storage_extents[i].start, storage_extents[i].end + 1);
+	return 0;
+}
+
+static int tprot(unsigned long addr)
+{
+	unsigned long reg1, reg2;
+	int rc = -EFAULT;
+	psw_t old;
+
+	asm volatile(
+		"	mvc	0(16,%[psw_old]),0(%[psw_pgm])\n"
+		"	epsw	%[reg1],%[reg2]\n"
+		"	st	%[reg1],0(%[psw_pgm])\n"
+		"	st	%[reg2],4(%[psw_pgm])\n"
+		"	larl	%[reg1],1f\n"
+		"	stg	%[reg1],8(%[psw_pgm])\n"
+		"	tprot	0(%[addr]),0\n"
+		"	ipm	%[rc]\n"
+		"	srl	%[rc],28\n"
+		"1:	mvc	0(16,%[psw_pgm]),0(%[psw_old])\n"
+		: [reg1] "=&d" (reg1),
+		  [reg2] "=&a" (reg2),
+		  [rc] "+&d" (rc),
+		  "=Q" (S390_lowcore.program_new_psw.addr),
+		  "=Q" (old)
+		: [psw_old] "a" (&old),
+		  [psw_pgm] "a" (&S390_lowcore.program_new_psw),
+		  [addr] "a" (addr)
+		: "cc", "memory");
+	return rc;
+}
+
+static unsigned long search_mem_end(void)
+{
+	unsigned long range = 1 << (MAX_PHYSMEM_BITS - 20); /* in 1MB blocks */
+	unsigned long offset = 0;
+	unsigned long pivot;
+
+	while (range > 1) {
+		range >>= 1;
+		pivot = offset + range;
+		if (!tprot(pivot << 20))
+			offset = pivot;
+	}
+	return (offset + 1) << 20;
+}
+
+unsigned long detect_max_physmem_end(void)
+{
+	unsigned long max_physmem_end = 0;
+
+	if (!sclp_early_get_memsize(&max_physmem_end)) {
+		physmem_info.info_source = MEM_DETECT_SCLP_READ_INFO;
+	} else {
+		max_physmem_end = search_mem_end();
+		physmem_info.info_source = MEM_DETECT_BIN_SEARCH;
+	}
+	return max_physmem_end;
+}
+
+void detect_physmem_online_ranges(unsigned long max_physmem_end)
+{
+	if (!sclp_early_read_storage_info()) {
+		physmem_info.info_source = MEM_DETECT_SCLP_STOR_INFO;
+	} else if (!diag260()) {
+		physmem_info.info_source = MEM_DETECT_DIAG260;
+	} else if (max_physmem_end) {
+		add_physmem_online_range(0, max_physmem_end);
+	}
+}
+
+void physmem_set_usable_limit(unsigned long limit)
+{
+	physmem_info.usable = limit;
+	physmem_alloc_pos = limit;
+}
+
+static void die_oom(unsigned long size, unsigned long align, unsigned long min, unsigned long max)
+{
+	unsigned long start, end, total_mem = 0, total_reserved_mem = 0;
+	struct reserved_range *range;
+	enum reserved_range_type t;
+	int i;
+
+	decompressor_printk("Linux version %s\n", kernel_version);
+	if (!is_prot_virt_guest() && early_command_line[0])
+		decompressor_printk("Kernel command line: %s\n", early_command_line);
+	decompressor_printk("Out of memory allocating %lx bytes %lx aligned in range %lx:%lx\n",
+			    size, align, min, max);
+	decompressor_printk("Reserved memory ranges:\n");
+	for_each_physmem_reserved_range(t, range, &start, &end) {
+		decompressor_printk("%016lx %016lx %s\n", start, end, get_rr_type_name(t));
+		total_reserved_mem += end - start;
+	}
+	decompressor_printk("Usable online memory ranges (info source: %s [%x]):\n",
+			    get_physmem_info_source(), physmem_info.info_source);
+	for_each_physmem_usable_range(i, &start, &end) {
+		decompressor_printk("%016lx %016lx\n", start, end);
+		total_mem += end - start;
+	}
+	decompressor_printk("Usable online memory total: %lx Reserved: %lx Free: %lx\n",
+			    total_mem, total_reserved_mem,
+			    total_mem > total_reserved_mem ? total_mem - total_reserved_mem : 0);
+	print_stacktrace(current_frame_address());
+	sclp_early_printk("\n\n -- System halted\n");
+	disabled_wait();
+}
+
+void physmem_reserve(enum reserved_range_type type, unsigned long addr, unsigned long size)
+{
+	physmem_info.reserved[type].start = addr;
+	physmem_info.reserved[type].end = addr + size;
+}
+
+void physmem_free(enum reserved_range_type type)
+{
+	physmem_info.reserved[type].start = 0;
+	physmem_info.reserved[type].end = 0;
+}
+
+static bool __physmem_alloc_intersects(unsigned long addr, unsigned long size,
+				       unsigned long *intersection_start)
+{
+	unsigned long res_addr, res_size;
+	int t;
+
+	for (t = 0; t < RR_MAX; t++) {
+		if (!get_physmem_reserved(t, &res_addr, &res_size))
+			continue;
+		if (intersects(addr, size, res_addr, res_size)) {
+			*intersection_start = res_addr;
+			return true;
+		}
+	}
+	return ipl_report_certs_intersects(addr, size, intersection_st