V5 Implement crashkernel=auto

V4 -> V5:
- Rename the global functions, as suggested by Andrew.
- Save some macros, as suggested by Andrew.
- Change the high threshold, from 32G to 4G.
- Change the low threshold on ppc, suggested by ppc developers.
- Make the mm part as a seperate function, suggest by Eric.
- Make the IA64 code more readable.
- Reorder the patchset again, since review from mm people is done.

V3 -> V4:
- Reorder the patches.
- Really free the reserved memory, instead of remapping it. (Thanks to KAMEZAWA Hiroyuki!)
- Release the reserved memory resource when the size is 0.
- Use strict_strtoul() instead of simple_strtoul().

V2 -> V3:
- Use more clever way to calculate reserved memory size, especially for IA64.
- Add that patch that implements shrinking reserved memory

V1 -> V2:
- Use include/asm-generic/kexec.h, suggested by Neil.
- Rename a local variable, suggested by Fenghua.
- Fix some style problems found by checkpatch.pl.
- Unify the Kconfig docs.

This series of patch implements automatically reserved memory for crashkernel, by introducing a new boot option "crashkernel=auto". This idea is from Neil.

In case of breaking user-space applications, it modifies this boot option after it decides how much memory should be reserved.

On different arch, the threshold and reserved memory size is different. Please refer patch 7/8 which contains an update for the documentation.

Patch 8/8 implements shrinking reserved memory at run-time, which is useful when more than enough memory is reserved automatically.

This patchset _is_ already tested on x86_64, IA64 and ppc64.

---
 Documentation/kdump/kdump.txt    |   29 +++++++++
 arch/ia64/Kconfig                |   14 ++++
 arch/ia64/include/asm/kexec.h    |   44 +++++++++++++
 arch/powerpc/Kconfig             |   11 +++
 arch/powerpc/include/asm/kexec.h |   39 ++++++++++++
 arch/x86/Kconfig                 |   13 ++++
 arch/x86/include/asm/kexec.h     |    1 
 include/asm-generic/kexec.h      |   16 +++++
 include/linux/kexec.h            |    2 
 kernel/kexec.c                   |  124 +++++++++++++++++++++++++++++++++++++++
 kernel/ksysfs.c                  |   21 ++++++
 11 files changed, 314 insertions(+)

Introduce a new config option KEXEC_AUTO_RESERVE for x86.

---

Index: linux-2.6/arch/x86/Kconfig =================================--- linux-2.6.orig/arch/x86/Kconfig +++ linux-2.6/arch/x86/Kconfig
@@ -1482,6 +1482,19 @@ config KEXEC
support. As of this writing the exact hardware interface is strongly in flux, so no good recommendation can be made.

+config KEXEC_AUTO_RESERVE
+ bool "automatically reserve memory for kexec kernel"
+ depends on KEXEC
+ default y
+ ---help---

+	  Automatically reserve memory for a kexec kernel, so that you don't
+	  need to specify numbers for the "crashkernel=X@Y" boot option,
+	  instead you can use "crashkernel=auto". To make this work, you need
+	  to have more than 4G memory.

+
+ On x86_32, 128M is reserved, on x86_64 1/32 of your memory is
+ reserved, but it will not exceed 4G.
+
config CRASH_DUMP
bool "kernel crash dumps"
depends on X86_64 || (X86_32 && HIGHMEM)

Implement "crashkernel=auto" for x86 first, other arch will be added in the following patches.

The kernel will modify this command line with the actually reserved size, in case of breaking any user-space programs.

---

Index: linux-2.6/kernel/kexec.c =================================--- linux-2.6.orig/kernel/kexec.c +++ linux-2.6/kernel/kexec.c
@@ -37,6 +37,7 @@
#include <asm/io.h>
#include <asm/system.h>
#include <asm/sections.h>
+#include <asm/setup.h>

/* Per cpu memory for storing cpu states in case of system crash. */ note_buf_t* crash_notes;
@@ -1269,6 +1270,37 @@ static int __init parse_crashkernel_simp
return 0;
}

+#ifdef CONFIG_KEXEC_AUTO_RESERVE
+#ifndef arch_default_crash_size
+unsigned long long arch_default_crash_size(unsigned long long total_size) +{
+ if (total_size < KEXEC_AUTO_THRESHOLD)
+ return 0;
+ else {
+#ifdef CONFIG_64BIT
+ if (total_size > (1ULL<<37)) /* 128G */
+ return KEXEC_AUTO_RESERVED_SIZE
+ * ((1ULL<<37) / KEXEC_AUTO_THRESHOLD);
+ return 1ULL<<ilog2(roundup(total_size/32, 1ULL<<21));
+#else
+ return KEXEC_AUTO_RESERVED_SIZE;
+#endif
+ }
+}
+#define arch_default_crash_size arch_default_crash_size +#endif
+
+#ifndef arch_default_crash_base
+unsigned long long arch_default_crash_base(void)
+{
+ /* 0 means find the base address automatically. */
+ return 0;
+}
+#define arch_default_crash_base arch_default_crash_base +#endif
+
+#endif /*CONFIG_KEXEC_AUTO_RESERVE*/
+
/*
* That function is the entry point for command line parsing and should be
* called from the arch-specific code.
@@ -1297,6 +1329,39 @@ int __init parse_crashkernel(char *cm

ck_cmdline += 12; /* strlen("crashkernel=") */

+#ifdef CONFIG_KEXEC_AUTO_RESERVE
+ if (strncmp(ck_cmdline, "auto", 4) = 0) {
+ unsigned long long size;
+ int len;
+ char tmp[32];
+
+ size = arch_default_crash_size(system_ram);
+ if (size != 0) {
+ *crash_size = size;
+ *crash_base = arch_default_crash_base();
+ len = scnprintf(tmp, sizeof(tmp), "%luM@%luM",
+ (unsigned long)(*crash_size)>>20,
+ (unsigned long)(*crash_base)>>20);
+ /* 'len' can't be <= 4. */
+ if (likely((len - 4 + strlen(cmdline))
+ < COMMAND_LINE_SIZE - 1)) {
+ memmove(ck_cmdline + len, ck_cmdline + 4,
+ strlen(cmdline) - (ck_cmdline + 4 - cmdline) + 1);
+ memcpy(ck_cmdline, tmp, len);
+ }
+ return 0;
+ } else {
+ /*
+ * We can't reserve memory auotmatcally,
+ * remove "crashkernel=auto" from cmdline.
+ */
+ ck_cmdline += 4; /* strlen("auto") */
+ memmove(ck_cmdline - 16, ck_cmdline,
+ strlen(cmdline) - (ck_cmdline - cmdline) + 1);
+ return -ENOMEM;
+ }
+ }
+#endif
/*
* if the commandline contains a ':', then that's the extended
* syntax -- if not, it must be the classic syntax
Index: linux-2.6/arch/x86/include/asm/kexec.h =================================--- linux-2.6.orig/arch/x86/include/asm/kexec.h +++ linux-2.6/arch/x86/include/asm/kexec.h
@@ -23,6 +23,7 @@

#include <asm/page.h>
#include <asm/ptrace.h>
+#include <asm-generic/kexec.h>

/*
* KEXEC_SOURCE_MEMORY_LIMIT maximum page get_free_page can return.
Index: linux-2.6/include/asm-generic/kexec.h =================================--- /dev/null
+++ linux-2.6/include/asm-generic/kexec.h
@@ -0,0 +1,16 @@
+#ifndef _ASM_GENERIC_KEXEC_H
+#define _ASM_GENERIC_KEXEC_H
+
+#ifdef CONFIG_KEXEC_AUTO_RESERVE
+
+#ifndef KEXEC_AUTO_RESERVED_SIZE
+#define KEXEC_AUTO_RESERVED_SIZE (1ULL<<27) /* 128M */ +#endif
+#ifndef KEXEC_AUTO_THRESHOLD
+#define KEXEC_AUTO_THRESHOLD (1ULL<<32) /* 4G */
+#endif
+
+
+#endif /* CONFIG_KEXEC_AUTO_RESERVE */
+
+#endif

Introduce a new config option KEXEC_AUTO_RESERVE for ia64.

---

Index: linux-2.6/arch/ia64/Kconfig =================================--- linux-2.6.orig/arch/ia64/Kconfig +++ linux-2.6/arch/ia64/Kconfig
@@ -582,6 +582,20 @@ config KEXEC
support. As of this writing the exact hardware interface is strongly in flux, so no good recommendation can be made.

+config KEXEC_AUTO_RESERVE
+ bool "automatically reserve memory for kexec kernel"
+ depends on KEXEC
+ default y
+ ---help---

+	  Automatically reserve memory for a kexec kernel, so that you don't
+	  need to specify numbers for the "crashkernel=X@Y" boot option,
+	  instead you can use "crashkernel=auto". To make this work, you need
+	  to have more than 4G memory.
+
+	  The reserved memory size is different depends on how much memory
+	  you actually have. Please check Documentation/kdump/kdump.txt.
+	  If you doubt, say N.

+
config CRASH_DUMP
bool "kernel crash dumps"
depends on IA64_MCA_RECOVERY && !IA64_HP_SIM && (!SMP || HOTPLUG_CPU)

Since in patch 2/8 we already implement the generic part, this will add the rest part for ia64.

---

Index: linux-2.6/arch/ia64/include/asm/kexec.h =================================--- linux-2.6.orig/arch/ia64/include/asm/kexec.h +++ linux-2.6/arch/ia64/include/asm/kexec.h
@@ -1,6 +1,7 @@
#ifndef _ASM_IA64_KEXEC_H
#define _ASM_IA64_KEXEC_H

+#include <asm-generic/kexec.h>

/* Maximum physical address we can use pages from */ #define KEXEC_SOURCE_MEMORY_LIMIT (-1UL)
@@ -19,6 +20,49 @@

                 flush_icache_range(page_addr, page_addr + PAGE_SIZE); \
         } while(0)

+#ifdef CONFIG_KEXEC_AUTO_RESERVE
+#define MBYTES(n) ((n)*1024*1024ULL)
+#define GBYTES(n) ((n)*1024*1024*1024ULL)
+/*

+       Memory size     Reserved memory
+       ======     =======+       [4G, 12G)       256M
+       [12G, 128G)     512M
+       [128G, 256G)    768M
+       [256G, 378G)    1024M
+       [378G, 512G)    1536M
+       [512G, 768G)    2048M
+       [768G, )        3072M

+ */
+static inline
+unsigned long long arch_default_crash_size(unsigned long long total_size) +{
+ unsigned long long ret;
+
+ if (total_size >= GBYTES(4) && total_size < GBYTES(12))
+ ret = MBYTES(256);
+ else if (total_size >= GBYTES(12) && total_size < GBYTES(128))
+ ret = MBYTES(512);
+ else if (total_size >= GBYTES(128) && total_size < GBYTES(256))
+ ret = MBYTES(768);
+ else if (total_size >= GBYTES(256) && total_size < GBYTES(378))
+ ret = MBYTES(1024);
+ else if (total_size >= GBYTES(318) && total_size < GBYTES(512))
+ ret = MBYTES(1536);
+ else if (total_size >= GBYTES(512) && total_size < GBYTES(768))
+ ret = MBYTES(2048);
+ else
+ ret = MBYTES(3072);
+ if (!ia64_platform_is("sn2") && !ia64_platform_is("uv"))
+ if (ret > GBYTES(4))
+ ret = ret / GBYTES(4) * MBYTES(256);
+ return ret;
+}
+#undef GBYTES
+#undef MBYTES
+#define arch_default_crash_size arch_default_crash_size +#endif
+
extern struct kimage *ia64_kimage;
extern const unsigned int relocate_new_kernel_size;
extern void relocate_new_kernel(unsigned long, unsigned long,

Yeah, seems I missed Andrew's comment... :-/ I will update it.

This statement will be ALWAYS false:
+ if (ret > GBYTES(4))
since ret's max value is 3GB after the above total_size checks.

So the code doesn't make sense. Did I miss something?

Thanks.

-Fenghua

Oops, in fact it was supposed to be:

if (ret + base > GBYTES(4))
//blah blah

Hmm, but is it possible for ia64 to choose a base address above 1G? If no, that code can be removed.

Thanks!

I think you don't need to consider the base here. This arch_default_crash_size() only returns size. Later on, ia64 setup_crashkernel() will find a base value in reserved region and check if the base value is under 4GB etc.

Thanks.

-Fenghua

Introduce a new config option KEXEC_AUTO_RESERVE for powerpc.

---

Index: linux-2.6/arch/powerpc/Kconfig =================================--- linux-2.6.orig/arch/powerpc/Kconfig +++ linux-2.6/arch/powerpc/Kconfig
@@ -346,6 +346,17 @@ config KEXEC
support. As of this writing the exact hardware interface is strongly in flux, so no good recommendation can be made.

+config KEXEC_AUTO_RESERVE
+ bool "automatically reserve memory for kexec kernel"
+ depends on KEXEC
+ default y
+ ---help---

+	  Automatically reserve memory for a kexec kernel, so that you don't
+	  need to specify numbers for the "crashkernel=X@Y" boot option,
+	  instead you can use "crashkernel=auto". To make this work, you need
+	  to have more than 2G memory. On PPC, 128M or 256M is reserved, on
+	  PPC64 1/32 of your physical memory, but it will not exceed 4G.

+
config CRASH_DUMP
bool "Build a kdump crash kernel"
depends on PPC64 || 6xx

Since in patch 2/8 we already implement the generic part, this will add the rest part for powerpc.

---

Index: linux-2.6/arch/powerpc/include/asm/kexec.h =================================--- linux-2.6.orig/arch/powerpc/include/asm/kexec.h +++ linux-2.6/arch/powerpc/include/asm/kexec.h
@@ -39,6 +39,45 @@ typedef void (*crash_shutdown_t)(void);

#ifdef CONFIG_KEXEC

+#ifdef CONFIG_KEXEC_AUTO_RESERVE
+
+#ifndef KEXEC_AUTO_THRESHOLD
+#define KEXEC_AUTO_THRESHOLD (1ULL<<31) /* 2G */
+#endif
+
+#ifndef arch_default_crash_base
+static inline
+unsigned long long arch_default_crash_base(void)
+{
+ return KDUMP_KERNELBASE;
+}
+#define arch_default_crash_base arch_default_crash_base +#endif
+
+#ifndef arch_default_crash_size
+static inline
+unsigned long long arch_default_crash_size(unsigned long long total_size) +{
+ if (total_size < KEXEC_AUTO_THRESHOLD)
+ return 0;
+ if (total_size < (1ULL<<32))
+ return 1ULL<<27;
+ else {
+#ifdef CONFIG_64BIT
+ if (total_size > (1ULL<<37)) /* 128G */
+ return 1ULL<<32; /* 4G */
+ return 1ULL<<ilog2(roundup(total_size/32, 1ULL<<21));
+#else
+ return 1ULL<<28;
+#endif
+ }
+}
+#define arch_default_crash_size arch_default_crash_size +#endif
+
+#endif
+#include <asm-generic/kexec.h>
+
/*
* This function is responsible for capturing register states if coming
* via panic or invoking dump using sysrq-trigger.

Update the document for kdump.

---

Index: linux-2.6/Documentation/kdump/kdump.txt =================================--- linux-2.6.orig/Documentation/kdump/kdump.txt +++ linux-2.6/Documentation/kdump/kdump.txt
@@ -147,6 +147,15 @@ System kernel config options
analysis tools require a vmlinux with debug symbols in order to read and analyze a dump file.

+4) Enable "automatically reserve memory for kexec kernel" in
+ "Processor type and features."
+
+ CONFIG_KEXEC_AUTO_RESERVE=y
+

+   This will let you to use "crashkernel=auto", instead of specifying
+   numbers for "crashkernel=". Note, you need to have enough memory.
+   The threshold and reserved memory size are arch-dependent.

+
Dump-capture kernel config options (Arch Independent)
-----------------------------------------------------

@@ -266,6 +275,26 @@ This would mean:
2) if the RAM size is between 512M and 2G (exclusive), then reserve 64M
3) if the RAM size is larger than 2G, then reserve 128M

+Or you can use:
+
+ crashkernel=auto
+
+if you have enough memory. The threshold is 4G on ia64 and x86, 2G on powerpc, +below which this won't work.
+
+The automatically reserved memory size would be 128M on x86_32, 128M or 256M +on ppc, 1/32 of your physical memory size on x86_64 and ppc64 (but it will not +grow if you have more than 128G memory). IA64 has its own policy, shown below:
+
+ Memory size Reserved memory
+ ====== =======+ [4G, 12G) 256M
+ [12G, 128G) 512M
+ [128G, 256G) 768M
+ [256G, 378G) 1024M
+ [378G, 512G) 1536M
+ [512G, 768G) 2048M
+ [768G, ) 3072M

Boot into System Kernel

This patch implements shrinking the reserved memory for crash kernel, if it is more than enough.

For example, if you have already reserved 128M, now you just want 100M, you can do:

# echo $((100*1024*1024)) > /sys/kernel/kexec_crash_size

Note, you can only do this before loading the crash kernel.

---

Index: linux-2.6/include/linux/kexec.h =================================--- linux-2.6.orig/include/linux/kexec.h +++ linux-2.6/include/linux/kexec.h
@@ -206,6 +206,8 @@ extern size_t vmcoreinfo_max_size;

int __init parse_crashkernel(char *cmdline, unsigned long long system_ram,
unsigned long long *crash_size, unsigned long long *crash_base);
+int crash_shrink_memory(unsigned long new_size);
+size_t crash_get_memory_size(void);

#else /* !CONFIG_KEXEC */
struct pt_regs;
Index: linux-2.6/kernel/kexec.c =================================--- linux-2.6.orig/kernel/kexec.c +++ linux-2.6/kernel/kexec.c
@@ -31,6 +31,7 @@
#include <linux/cpu.h>
#include <linux/console.h>
#include <linux/vmalloc.h>
+#include <linux/swap.h>

#include <asm/page.h>
#include <asm/uaccess.h>
@@ -1083,6 +1084,64 @@ void crash_kexec(struct pt_regs *regs)
}
}

+size_t crash_get_memory_size(void)
+{
+ size_t size;
+ mutex_lock(&kexec_mutex);
+ size = crashk_res.end - crashk_res.start + 1;
+ mutex_unlock(&kexec_mutex);
+ return size;
+}
+
+static void free_reserved_phys_range(unsigned long begin, unsigned long end) +{
+ unsigned long addr;
+
+ for (addr = begin; addr < end; addr += PAGE_SIZE) {
+ ClearPageReserved(pfn_to_page(addr >> PAGE_SHIFT));
+ init_page_count(pfn_to_page(addr >> PAGE_SHIFT));
+ free_page((unsigned long)__va(addr));
+ totalram_pages++;
+ }
+}
+
+int crash_shrink_memory(unsigned long new_size)
+{
+ int ret = 0;
+ unsigned long start, end;
+
+ mutex_lock(&kexec_mutex);
+
+ if (kexec_crash_image) {
+ ret = -ENOENT;
+ goto unlock;
+ }
+ start = crashk_res.start;
+ end = crashk_res.end;
+
+ if (new_size >= end - start + 1) {
+ ret = -EINVAL;
+ if (new_size = end - start + 1)
+ ret = 0;
+ goto unlock;
+ }
+
+ start = roundup(start, PAGE_SIZE);
+ end = roundup(start + new_size, PAGE_SIZE);
+
+ free_reserved_phys_range(end, crashk_res.end);
+
+ if (start = end) {
+ crashk_res.end = end;
+ release_resource(&crashk_res);
+ } else
+ crashk_res.end = end - 1;
+
+unlock:
+ mutex_unlock(&kexec_mutex);
+ return ret;
+}
+
static u32 *append_elf_note(u32 *buf, char *name, unsigned type, void *data,
size_t data_len)
{
Index: linux-2.6/kernel/ksysfs.c =================================--- linux-2.6.orig/kernel/ksysfs.c +++ linux-2.6/kernel/ksysfs.c
@@ -100,6 +100,26 @@ static ssize_t kexec_crash_loaded_show(s
}
KERNEL_ATTR_RO(kexec_crash_loaded);

+static ssize_t kexec_crash_size_show(struct kobject *kobj,
+ struct kobj_attribute *attr, char *buf)
+{
+ return sprintf(buf, "%lu\n", crash_get_memory_size());
+}
+static ssize_t kexec_crash_size_store(struct kobject *kobj,
+ struct kobj_attribute *attr,
+ const char *buf, size_t count)
+{
+ unsigned long cnt;
+ int ret;
+
+ if (strict_strtoul(buf, 0, &cnt))
+ return -EINVAL;
+
+ ret = crash_shrink_memory(cnt);
+ return ret < 0 ? ret : count;
+}
+KERNEL_ATTR_RW(kexec_crash_size);
+
static ssize_t vmcoreinfo_show(struct kobject *kobj,
struct kobj_attribute *attr, char *buf)
{
@@ -147,6 +167,7 @@ static struct attribute * kernel_attrs[]
#ifdef CONFIG_KEXEC
&kexec_loaded_attr.attr, &kexec_crash_loaded_attr.attr,
+ &kexec_crash_size_attr.attr,
&vmcoreinfo_attr.attr,
#endif
NULL

I don't want to sound like a micro-kernel zealot, I'm not, but I'm still unconvinced as to why the auto logic needs to go in the kernel. What is the compelling reason that the kernel needs to do this calculation vs some userspace tool? We already have the syntax that allows defining a different crash size depending on the size of RAM.

The shrinking of reserved memory is cool.

cheers

Michael I am in agreement with you.
The shrinking is good.

If we can come up with some simple and generic logic that we can use to reserve memory then I am in favor. However all this patchset is doing is moving user space specific arbitrary hacks into the kernel we do that perfectly well on the command line. Having the amount to reserve be arch specific is complete non-sense and a major maintenance pain.

Eric

Ok, since you guys think we should not do sth in kernel space if we can do it in user space, why not removing the extended crash kernel syntax??

E.g.
crashkernelQ2M-2G:64M,2G-:128M

I am *quite* sure this can be done in user space too. In theory, only you need is just:

crashkernel=X@Y

all the rest things can be done in user space. Enjoy!

Take a look at the x86 part, it can not be done _directly_ in kernel command line with the syntax.