* [RFC PATCH 0/4] makedumpfile/arm64: support flipped VA and 52-bit kernel VA
@ 2021-01-14 8:25 kazuhito.hagio
2021-01-14 8:25 ` [RFC PATCH 1/4] Use ELF vmcoreinfo note for --mem-usage option kazuhito.hagio
` (3 more replies)
0 siblings, 4 replies; 10+ messages in thread
From: kazuhito.hagio @ 2021-01-14 8:25 UTC (permalink / raw)
To: bhsharma, bhupesh.linux, alexander.kamensky42, piliu, kexec
Cc: aadiga, ioanna-maria.alifieraki
Hi Bhupesh, Alexander, Pingfan and arm64 guys,
I've rewritten Bhupesh's v5 patch [1] in order to avoid some limitations and
now this patchset
- can be built on some old arm64 environments hopefully,
- can be built on x86_64 machines with TARGET=aarch64,
- can process vmcores on different version of kernel from captured version.
Could you test this? The patchset is also on GitHub [2] and the commit
messages are to be written later.
However the following conditions are probably not supported:
- 5.11-rc is not taken into account
- 5.4 <= kernels <= 5.8 (*1)
* if PA_BITS=52 && VA_BITS!=52
* with -x option if vabits_actual=52
- kernels < 5.4 with CONFIG_ARM64_USER_VA_BITS_52=y (e.g. RHEL8 kernels)
(*1) supported with kernel commit bbdbc11804ff and 1d50e5d0c5052.
I think especially MAX_PHYSMEM_BITS cannot be guessed correctly only with
the entries in vmcoreinfo (even with Bhupesh's original patch), so basically
I would like distributions that uses 5.4 to 5.8 to backport Bhupesh's kernel
patches above if the patchset does not work well.
[1] http://lists.infradead.org/pipermail/kexec/2020-September/021336.html
[2] https://github.com/k-hagio/makedumpfile/commits/arm64.kh.test2
Kazuhito Hagio (4):
Use ELF vmcoreinfo note for --mem-usage option
arm64: use NUMBER(VA_BITS) in vmcoreinfo
Get kernel version from OSRELEASE in vmcoreinfo
arm64: support flipped VA and 52-bit kernel VA
arch/arm64.c | 155 ++++++++++++++++++++++++++++++++++++++++++++-------------
makedumpfile.c | 32 ++++++++++--
makedumpfile.h | 1 +
3 files changed, 149 insertions(+), 39 deletions(-)
--
2.9.3
_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec
^ permalink raw reply [flat|nested] 10+ messages in thread
* [RFC PATCH 1/4] Use ELF vmcoreinfo note for --mem-usage option
2021-01-14 8:25 [RFC PATCH 0/4] makedumpfile/arm64: support flipped VA and 52-bit kernel VA kazuhito.hagio
@ 2021-01-14 8:25 ` kazuhito.hagio
2021-01-14 8:25 ` [RFC PATCH 2/4] arm64: use NUMBER(VA_BITS) in vmcoreinfo kazuhito.hagio
` (2 subsequent siblings)
3 siblings, 0 replies; 10+ messages in thread
From: kazuhito.hagio @ 2021-01-14 8:25 UTC (permalink / raw)
To: bhsharma, bhupesh.linux, alexander.kamensky42, piliu, kexec
Cc: aadiga, ioanna-maria.alifieraki
From: Kazuhito Hagio <k-hagio-ab@nec.com>
Signed-off-by: Kazuhito Hagio <k-hagio-ab@nec.com>
---
makedumpfile.c | 26 +++++++++++++++++++++-----
1 file changed, 21 insertions(+), 5 deletions(-)
diff --git a/makedumpfile.c b/makedumpfile.c
index ecd63fa..199748b 100644
--- a/makedumpfile.c
+++ b/makedumpfile.c
@@ -11426,6 +11426,7 @@ int show_mem_usage(void)
{
uint64_t vmcoreinfo_addr, vmcoreinfo_len;
struct cycle cycle = {0};
+ int vmcoreinfo = FALSE;
if (!is_crashkernel_mem_reserved()) {
ERRMSG("No memory is reserved for crashkernel!\n");
@@ -11437,9 +11438,22 @@ int show_mem_usage(void)
if (!open_files_for_creating_dumpfile())
return FALSE;
- if (!get_elf_loads(info->fd_memory, info->name_memory))
+ if (!get_elf_info(info->fd_memory, info->name_memory))
return FALSE;
+ /*
+ * /proc/kcore on Linux 4.19 and later kernels have vmcoreinfo note in
+ * NOTE segment. See commit 23c85094fe18.
+ */
+ if (has_vmcoreinfo()) {
+ off_t offset;
+ unsigned long size;
+
+ get_vmcoreinfo(&offset, &size);
+ vmcoreinfo = read_vmcoreinfo_from_vmcore(offset, size, FALSE);
+ DEBUG_MSG("Read vmcoreinfo from NOTE segment: %d\n", vmcoreinfo);
+ }
+
if (!get_page_offset())
return FALSE;
@@ -11447,11 +11461,13 @@ int show_mem_usage(void)
if (!get_phys_base())
return FALSE;
- if (!get_sys_kernel_vmcoreinfo(&vmcoreinfo_addr, &vmcoreinfo_len))
- return FALSE;
+ if (!vmcoreinfo) {
+ if (!get_sys_kernel_vmcoreinfo(&vmcoreinfo_addr, &vmcoreinfo_len))
+ return FALSE;
- if (!set_kcore_vmcoreinfo(vmcoreinfo_addr, vmcoreinfo_len))
- return FALSE;
+ if (!set_kcore_vmcoreinfo(vmcoreinfo_addr, vmcoreinfo_len))
+ return FALSE;
+ }
if (!initial())
return FALSE;
--
2.9.3
_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec
^ permalink raw reply related [flat|nested] 10+ messages in thread
* [RFC PATCH 2/4] arm64: use NUMBER(VA_BITS) in vmcoreinfo
2021-01-14 8:25 [RFC PATCH 0/4] makedumpfile/arm64: support flipped VA and 52-bit kernel VA kazuhito.hagio
2021-01-14 8:25 ` [RFC PATCH 1/4] Use ELF vmcoreinfo note for --mem-usage option kazuhito.hagio
@ 2021-01-14 8:25 ` kazuhito.hagio
2021-01-14 8:25 ` [RFC PATCH 3/4] Get kernel version from OSRELEASE " kazuhito.hagio
2021-01-14 8:25 ` [RFC PATCH 4/4] arm64: support flipped VA and 52-bit kernel VA kazuhito.hagio
3 siblings, 0 replies; 10+ messages in thread
From: kazuhito.hagio @ 2021-01-14 8:25 UTC (permalink / raw)
To: bhsharma, bhupesh.linux, alexander.kamensky42, piliu, kexec
Cc: aadiga, ioanna-maria.alifieraki
From: Kazuhito Hagio <k-hagio-ab@nec.com>
Signed-off-by: Bhupesh Sharma <bhsharma@redhat.com>
Signed-off-by: Kazuhito Hagio <k-hagio-ab@nec.com>
---
arch/arm64.c | 64 ++++++++++++++++++++++++++++++++++++++++--------------------
1 file changed, 43 insertions(+), 21 deletions(-)
diff --git a/arch/arm64.c b/arch/arm64.c
index 3d7b416..61ec89a 100644
--- a/arch/arm64.c
+++ b/arch/arm64.c
@@ -345,6 +345,44 @@ get_stext_symbol(void)
return(found ? kallsym : FALSE);
}
+static int
+get_va_bits_from_stext_arm64(void)
+{
+ ulong _stext;
+
+ _stext = get_stext_symbol();
+ if (!_stext) {
+ ERRMSG("Can't get the symbol of _stext.\n");
+ return FALSE;
+ }
+
+ /*
+ * Derive va_bits as per arch/arm64/Kconfig. Note that this is a
+ * best case approximation at the moment, as there can be
+ * inconsistencies in this calculation (for e.g., for
+ * 52-bit kernel VA case, the 48th bit is set in
+ * the _stext symbol).
+ */
+ if ((_stext & PAGE_OFFSET_48) == PAGE_OFFSET_48) {
+ va_bits = 48;
+ } else if ((_stext & PAGE_OFFSET_47) == PAGE_OFFSET_47) {
+ va_bits = 47;
+ } else if ((_stext & PAGE_OFFSET_42) == PAGE_OFFSET_42) {
+ va_bits = 42;
+ } else if ((_stext & PAGE_OFFSET_39) == PAGE_OFFSET_39) {
+ va_bits = 39;
+ } else if ((_stext & PAGE_OFFSET_36) == PAGE_OFFSET_36) {
+ va_bits = 36;
+ } else {
+ ERRMSG("Cannot find a proper _stext for calculating VA_BITS\n");
+ return FALSE;
+ }
+
+ DEBUG_MSG("va_bits : %d (approximation via _stext)\n", va_bits);
+
+ return TRUE;
+}
+
int
get_machdep_info_arm64(void)
{
@@ -398,27 +436,11 @@ get_xen_info_arm64(void)
int
get_versiondep_info_arm64(void)
{
- ulong _stext;
-
- _stext = get_stext_symbol();
- if (!_stext) {
- ERRMSG("Can't get the symbol of _stext.\n");
- return FALSE;
- }
-
- /* Derive va_bits as per arch/arm64/Kconfig */
- if ((_stext & PAGE_OFFSET_36) == PAGE_OFFSET_36) {
- va_bits = 36;
- } else if ((_stext & PAGE_OFFSET_39) == PAGE_OFFSET_39) {
- va_bits = 39;
- } else if ((_stext & PAGE_OFFSET_42) == PAGE_OFFSET_42) {
- va_bits = 42;
- } else if ((_stext & PAGE_OFFSET_47) == PAGE_OFFSET_47) {
- va_bits = 47;
- } else if ((_stext & PAGE_OFFSET_48) == PAGE_OFFSET_48) {
- va_bits = 48;
- } else {
- ERRMSG("Cannot find a proper _stext for calculating VA_BITS\n");
+ if (NUMBER(VA_BITS) != NOT_FOUND_NUMBER) {
+ va_bits = NUMBER(VA_BITS);
+ DEBUG_MSG("va_bits : %d (vmcoreinfo)\n", va_bits);
+ } else if (get_va_bits_from_stext_arm64() == FALSE) {
+ ERRMSG("Can't determine va_bits.\n");
return FALSE;
}
--
2.9.3
_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec
^ permalink raw reply related [flat|nested] 10+ messages in thread
* [RFC PATCH 3/4] Get kernel version from OSRELEASE in vmcoreinfo
2021-01-14 8:25 [RFC PATCH 0/4] makedumpfile/arm64: support flipped VA and 52-bit kernel VA kazuhito.hagio
2021-01-14 8:25 ` [RFC PATCH 1/4] Use ELF vmcoreinfo note for --mem-usage option kazuhito.hagio
2021-01-14 8:25 ` [RFC PATCH 2/4] arm64: use NUMBER(VA_BITS) in vmcoreinfo kazuhito.hagio
@ 2021-01-14 8:25 ` kazuhito.hagio
2021-01-14 8:25 ` [RFC PATCH 4/4] arm64: support flipped VA and 52-bit kernel VA kazuhito.hagio
3 siblings, 0 replies; 10+ messages in thread
From: kazuhito.hagio @ 2021-01-14 8:25 UTC (permalink / raw)
To: bhsharma, bhupesh.linux, alexander.kamensky42, piliu, kexec
Cc: aadiga, ioanna-maria.alifieraki
From: Kazuhito Hagio <k-hagio-ab@nec.com>
Signed-off-by: Kazuhito Hagio <k-hagio-ab@nec.com>
---
makedumpfile.c | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/makedumpfile.c b/makedumpfile.c
index 199748b..fdfe437 100644
--- a/makedumpfile.c
+++ b/makedumpfile.c
@@ -1182,6 +1182,7 @@ check_release(void)
}
}
+ info->kernel_version = FALSE;
info->kernel_version = get_kernel_version(info->system_utsname.release);
if (info->kernel_version == FALSE) {
ERRMSG("Can't get the kernel version.\n");
@@ -2480,6 +2481,9 @@ read_vmcoreinfo_basic_info(void)
if (strlen(info->release))
continue;
strcpy(info->release, buf + strlen(STR_OSRELEASE));
+
+ if (!info->kernel_version)
+ info->kernel_version = get_kernel_version(info->release);
}
if (strncmp(buf, STR_PAGESIZE, strlen(STR_PAGESIZE)) == 0) {
page_size = strtol(buf+strlen(STR_PAGESIZE),&endp,10);
--
2.9.3
_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec
^ permalink raw reply related [flat|nested] 10+ messages in thread
* [RFC PATCH 4/4] arm64: support flipped VA and 52-bit kernel VA
2021-01-14 8:25 [RFC PATCH 0/4] makedumpfile/arm64: support flipped VA and 52-bit kernel VA kazuhito.hagio
` (2 preceding siblings ...)
2021-01-14 8:25 ` [RFC PATCH 3/4] Get kernel version from OSRELEASE " kazuhito.hagio
@ 2021-01-14 8:25 ` kazuhito.hagio
2021-01-14 10:03 ` piliu
3 siblings, 1 reply; 10+ messages in thread
From: kazuhito.hagio @ 2021-01-14 8:25 UTC (permalink / raw)
To: bhsharma, bhupesh.linux, alexander.kamensky42, piliu, kexec
Cc: aadiga, ioanna-maria.alifieraki
From: Kazuhito Hagio <k-hagio-ab@nec.com>
Based on Bhupesh's patch and contains Pingfan's idea.
Signed-off-by: Bhupesh Sharma <bhsharma@redhat.com>
Signed-off-by: Kazuhito Hagio <k-hagio-ab@nec.com>
---
arch/arm64.c | 95 ++++++++++++++++++++++++++++++++++++++++++++++++----------
makedumpfile.c | 2 ++
makedumpfile.h | 1 +
3 files changed, 83 insertions(+), 15 deletions(-)
diff --git a/arch/arm64.c b/arch/arm64.c
index 61ec89a..4ece19d 100644
--- a/arch/arm64.c
+++ b/arch/arm64.c
@@ -47,6 +47,8 @@ typedef struct {
static int lpa_52_bit_support_available;
static int pgtable_level;
static int va_bits;
+static int vabits_actual;
+static int flipped_va;
static unsigned long kimage_voffset;
#define SZ_4K 4096
@@ -58,7 +60,6 @@ static unsigned long kimage_voffset;
#define PAGE_OFFSET_42 ((0xffffffffffffffffUL) << 42)
#define PAGE_OFFSET_47 ((0xffffffffffffffffUL) << 47)
#define PAGE_OFFSET_48 ((0xffffffffffffffffUL) << 48)
-#define PAGE_OFFSET_52 ((0xffffffffffffffffUL) << 52)
#define pgd_val(x) ((x).pgd)
#define pud_val(x) (pgd_val((x).pgd))
@@ -218,12 +219,20 @@ pmd_page_paddr(pmd_t pmd)
#define pte_index(vaddr) (((vaddr) >> PAGESHIFT()) & (PTRS_PER_PTE - 1))
#define pte_offset(dir, vaddr) (pmd_page_paddr((*dir)) + pte_index(vaddr) * sizeof(pte_t))
+/*
+ * The linear kernel range starts at the bottom of the virtual address
+ * space. Testing the top bit for the start of the region is a
+ * sufficient check and avoids having to worry about the tag.
+ */
+#define is_linear_addr(addr) (flipped_va ? \
+ (!((unsigned long)(addr) & (1UL << (vabits_actual - 1)))) : \
+ (!!((unsigned long)(addr) & (1UL << (vabits_actual - 1)))))
+
static unsigned long long
__pa(unsigned long vaddr)
{
- if (kimage_voffset == NOT_FOUND_NUMBER ||
- (vaddr >= PAGE_OFFSET))
- return (vaddr - PAGE_OFFSET + info->phys_base);
+ if (kimage_voffset == NOT_FOUND_NUMBER || is_linear_addr(vaddr))
+ return ((vaddr & ~PAGE_OFFSET) + info->phys_base);
else
return (vaddr - kimage_voffset);
}
@@ -253,6 +262,7 @@ static int calculate_plat_config(void)
(PAGESIZE() == SZ_64K && va_bits == 42)) {
pgtable_level = 2;
} else if ((PAGESIZE() == SZ_64K && va_bits == 48) ||
+ (PAGESIZE() == SZ_64K && va_bits == 52) ||
(PAGESIZE() == SZ_4K && va_bits == 39) ||
(PAGESIZE() == SZ_16K && va_bits == 47)) {
pgtable_level = 3;
@@ -263,6 +273,7 @@ static int calculate_plat_config(void)
PAGESIZE(), va_bits);
return FALSE;
}
+ DEBUG_MSG("pgtable_level: %d\n", pgtable_level);
return TRUE;
}
@@ -383,22 +394,54 @@ get_va_bits_from_stext_arm64(void)
return TRUE;
}
+static void
+get_page_offset_arm64(void)
+{
+ ulong page_end;
+ int vabits_min;
+
+ /*
+ * See arch/arm64/include/asm/memory.h for more details of
+ * the PAGE_OFFSET calculation.
+ */
+ vabits_min = (va_bits > 48) ? 48 : va_bits;
+ page_end = -(1UL << (vabits_min - 1));
+
+ if (SYMBOL(_stext) > page_end) {
+ flipped_va = TRUE;
+ info->page_offset = -(1UL << vabits_actual);
+ } else {
+ flipped_va = FALSE;
+ info->page_offset = -(1UL << (vabits_actual - 1));
+ }
+
+ DEBUG_MSG("page_offset : %lx (from page_end check)\n",
+ info->page_offset);
+}
+
int
get_machdep_info_arm64(void)
{
+ /* Check if va_bits is still not initialized. If still 0, call
+ * get_versiondep_info() to initialize the same.
+ */
+ if (!va_bits)
+ get_versiondep_info_arm64();
+
/* Determine if the PA address range is 52-bits: ARMv8.2-LPA */
if (NUMBER(MAX_PHYSMEM_BITS) != NOT_FOUND_NUMBER) {
info->max_physmem_bits = NUMBER(MAX_PHYSMEM_BITS);
+ DEBUG_MSG("max_physmem_bits : %ld (vmcoreinfo)\n", info->max_physmem_bits);
if (info->max_physmem_bits == 52)
lpa_52_bit_support_available = 1;
- } else
- info->max_physmem_bits = 48;
+ } else {
+ if (va_bits == 52)
+ info->max_physmem_bits = 52; /* just guess */
+ else
+ info->max_physmem_bits = 48;
- /* Check if va_bits is still not initialized. If still 0, call
- * get_versiondep_info() to initialize the same.
- */
- if (!va_bits)
- get_versiondep_info_arm64();
+ DEBUG_MSG("max_physmem_bits : %ld (guess)\n", info->max_physmem_bits);
+ }
if (!calculate_plat_config()) {
ERRMSG("Can't determine platform config values\n");
@@ -409,7 +452,6 @@ get_machdep_info_arm64(void)
info->section_size_bits = SECTIONS_SIZE_BITS;
DEBUG_MSG("kimage_voffset : %lx\n", kimage_voffset);
- DEBUG_MSG("max_physmem_bits : %ld\n", info->max_physmem_bits);
DEBUG_MSG("section_size_bits: %ld\n", info->section_size_bits);
return TRUE;
@@ -444,10 +486,33 @@ get_versiondep_info_arm64(void)
return FALSE;
}
- info->page_offset = (0xffffffffffffffffUL) << (va_bits - 1);
+ /*
+ * See TCR_EL1, Translation Control Register (EL1) register
+ * description in the ARMv8 Architecture Reference Manual.
+ * Basically, we can use the TCR_EL1.T1SZ
+ * value to determine the virtual addressing range supported
+ * in the kernel-space (i.e. vabits_actual) since Linux 5.9.
+ */
+ if (NUMBER(TCR_EL1_T1SZ) != NOT_FOUND_NUMBER) {
+ vabits_actual = 64 - NUMBER(TCR_EL1_T1SZ);
+ DEBUG_MSG("vabits_actual : %d (vmcoreinfo)\n", vabits_actual);
+ } else if ((va_bits == 52) && (SYMBOL(mem_section) != NOT_FOUND_SYMBOL)) {
+ /*
+ * Linux 5.4 through 5.10 have the following linear space:
+ * 48-bit: 0xffff000000000000 - 0xffff7fffffffffff
+ * 58-bit: 0xfff0000000000000 - 0xfff7ffffffffffff
+ */
+ if (SYMBOL(mem_section) & (1UL << (52 - 1)))
+ vabits_actual = 48;
+ else
+ vabits_actual = 52;
+ DEBUG_MSG("vabits_actual : %d (guess from mem_section)\n", vabits_actual);
+ } else {
+ vabits_actual = va_bits;
+ DEBUG_MSG("vabits_actual : %d (same as va_bits)\n", vabits_actual);
+ }
- DEBUG_MSG("va_bits : %d\n", va_bits);
- DEBUG_MSG("page_offset : %lx\n", info->page_offset);
+ get_page_offset_arm64();
return TRUE;
}
diff --git a/makedumpfile.c b/makedumpfile.c
index fdfe437..37788ff 100644
--- a/makedumpfile.c
+++ b/makedumpfile.c
@@ -2396,6 +2396,7 @@ write_vmcoreinfo_data(void)
WRITE_NUMBER("HUGETLB_PAGE_DTOR", HUGETLB_PAGE_DTOR);
#ifdef __aarch64__
WRITE_NUMBER("VA_BITS", VA_BITS);
+ /* WRITE_NUMBER("TCR_EL1_T1SZ", TCR_EL1_T1SZ); should not exists */
WRITE_NUMBER_UNSIGNED("PHYS_OFFSET", PHYS_OFFSET);
WRITE_NUMBER_UNSIGNED("kimage_voffset", kimage_voffset);
#endif
@@ -2838,6 +2839,7 @@ read_vmcoreinfo(void)
READ_NUMBER("KERNEL_IMAGE_SIZE", KERNEL_IMAGE_SIZE);
#ifdef __aarch64__
READ_NUMBER("VA_BITS", VA_BITS);
+ READ_NUMBER("TCR_EL1_T1SZ", TCR_EL1_T1SZ);
READ_NUMBER_UNSIGNED("PHYS_OFFSET", PHYS_OFFSET);
READ_NUMBER_UNSIGNED("kimage_voffset", kimage_voffset);
#endif
diff --git a/makedumpfile.h b/makedumpfile.h
index 5f50080..a151da2 100644
--- a/makedumpfile.h
+++ b/makedumpfile.h
@@ -1999,6 +1999,7 @@ struct number_table {
long KERNEL_IMAGE_SIZE;
#ifdef __aarch64__
long VA_BITS;
+ long TCR_EL1_T1SZ;
unsigned long PHYS_OFFSET;
unsigned long kimage_voffset;
#endif
--
2.9.3
_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec
^ permalink raw reply related [flat|nested] 10+ messages in thread
* Re: [RFC PATCH 4/4] arm64: support flipped VA and 52-bit kernel VA
2021-01-14 8:25 ` [RFC PATCH 4/4] arm64: support flipped VA and 52-bit kernel VA kazuhito.hagio
@ 2021-01-14 10:03 ` piliu
2021-01-14 18:25 ` Bhupesh SHARMA
0 siblings, 1 reply; 10+ messages in thread
From: piliu @ 2021-01-14 10:03 UTC (permalink / raw)
To: kazuhito.hagio, bhsharma, bhupesh.linux, alexander.kamensky42, kexec
Cc: aadiga, ioanna-maria.alifieraki
On 1/14/21 4:25 PM, kazuhito.hagio@gmail.com wrote:
> From: Kazuhito Hagio <k-hagio-ab@nec.com>
>
> Based on Bhupesh's patch and contains Pingfan's idea.
>
> Signed-off-by: Bhupesh Sharma <bhsharma@redhat.com>
> Signed-off-by: Kazuhito Hagio <k-hagio-ab@nec.com>
> ---
> arch/arm64.c | 95 ++++++++++++++++++++++++++++++++++++++++++++++++----------
> makedumpfile.c | 2 ++
> makedumpfile.h | 1 +
> 3 files changed, 83 insertions(+), 15 deletions(-)
>
> diff --git a/arch/arm64.c b/arch/arm64.c
> index 61ec89a..4ece19d 100644
> --- a/arch/arm64.c
> +++ b/arch/arm64.c
> @@ -47,6 +47,8 @@ typedef struct {
> static int lpa_52_bit_support_available;
> static int pgtable_level;
> static int va_bits;
> +static int vabits_actual;
> +static int flipped_va;
> static unsigned long kimage_voffset;
>
> #define SZ_4K 4096
> @@ -58,7 +60,6 @@ static unsigned long kimage_voffset;
> #define PAGE_OFFSET_42 ((0xffffffffffffffffUL) << 42)
> #define PAGE_OFFSET_47 ((0xffffffffffffffffUL) << 47)
> #define PAGE_OFFSET_48 ((0xffffffffffffffffUL) << 48)
> -#define PAGE_OFFSET_52 ((0xffffffffffffffffUL) << 52)
>
> #define pgd_val(x) ((x).pgd)
> #define pud_val(x) (pgd_val((x).pgd))
> @@ -218,12 +219,20 @@ pmd_page_paddr(pmd_t pmd)
> #define pte_index(vaddr) (((vaddr) >> PAGESHIFT()) & (PTRS_PER_PTE - 1))
> #define pte_offset(dir, vaddr) (pmd_page_paddr((*dir)) + pte_index(vaddr) * sizeof(pte_t))
>
> +/*
> + * The linear kernel range starts at the bottom of the virtual address
> + * space. Testing the top bit for the start of the region is a
> + * sufficient check and avoids having to worry about the tag.
> + */
> +#define is_linear_addr(addr) (flipped_va ? \
> + (!((unsigned long)(addr) & (1UL << (vabits_actual - 1)))) : \
> + (!!((unsigned long)(addr) & (1UL << (vabits_actual - 1)))))
> +
> static unsigned long long
> __pa(unsigned long vaddr)
> {
> - if (kimage_voffset == NOT_FOUND_NUMBER ||
> - (vaddr >= PAGE_OFFSET))
> - return (vaddr - PAGE_OFFSET + info->phys_base);
> + if (kimage_voffset == NOT_FOUND_NUMBER || is_linear_addr(vaddr))
> + return ((vaddr & ~PAGE_OFFSET) + info->phys_base);
> else
> return (vaddr - kimage_voffset);
> }
> @@ -253,6 +262,7 @@ static int calculate_plat_config(void)
> (PAGESIZE() == SZ_64K && va_bits == 42)) {
> pgtable_level = 2;
> } else if ((PAGESIZE() == SZ_64K && va_bits == 48) ||
> + (PAGESIZE() == SZ_64K && va_bits == 52) ||
> (PAGESIZE() == SZ_4K && va_bits == 39) ||
> (PAGESIZE() == SZ_16K && va_bits == 47)) {
> pgtable_level = 3;
> @@ -263,6 +273,7 @@ static int calculate_plat_config(void)
> PAGESIZE(), va_bits);
> return FALSE;
> }
> + DEBUG_MSG("pgtable_level: %d\n", pgtable_level);
>
> return TRUE;
> }
> @@ -383,22 +394,54 @@ get_va_bits_from_stext_arm64(void)
> return TRUE;
> }
>
> +static void
> +get_page_offset_arm64(void)
> +{
> + ulong page_end;
> + int vabits_min;
> +
> + /*
> + * See arch/arm64/include/asm/memory.h for more details of
> + * the PAGE_OFFSET calculation.
> + */
> + vabits_min = (va_bits > 48) ? 48 : va_bits;
> + page_end = -(1UL << (vabits_min - 1));
> +
> + if (SYMBOL(_stext) > page_end) {
> + flipped_va = TRUE;
> + info->page_offset = -(1UL << vabits_actual);
> + } else {
> + flipped_va = FALSE;
> + info->page_offset = -(1UL << (vabits_actual - 1));
> + }
> +
> + DEBUG_MSG("page_offset : %lx (from page_end check)\n",
> + info->page_offset);
> +}
> +
> int
> get_machdep_info_arm64(void)
> {
> + /* Check if va_bits is still not initialized. If still 0, call
> + * get_versiondep_info() to initialize the same.
> + */
> + if (!va_bits)
> + get_versiondep_info_arm64();
> +
> /* Determine if the PA address range is 52-bits: ARMv8.2-LPA */
> if (NUMBER(MAX_PHYSMEM_BITS) != NOT_FOUND_NUMBER) {
> info->max_physmem_bits = NUMBER(MAX_PHYSMEM_BITS);
> + DEBUG_MSG("max_physmem_bits : %ld (vmcoreinfo)\n", info->max_physmem_bits);
> if (info->max_physmem_bits == 52)
> lpa_52_bit_support_available = 1;
> - } else
> - info->max_physmem_bits = 48;
> + } else {
> + if (va_bits == 52)
> + info->max_physmem_bits = 52; /* just guess */
> + else
> + info->max_physmem_bits = 48;
>
> - /* Check if va_bits is still not initialized. If still 0, call
> - * get_versiondep_info() to initialize the same.
> - */
> - if (!va_bits)
> - get_versiondep_info_arm64();
> + DEBUG_MSG("max_physmem_bits : %ld (guess)\n", info->max_physmem_bits);
> + }
>
> if (!calculate_plat_config()) {
> ERRMSG("Can't determine platform config values\n");
> @@ -409,7 +452,6 @@ get_machdep_info_arm64(void)
> info->section_size_bits = SECTIONS_SIZE_BITS;
>
> DEBUG_MSG("kimage_voffset : %lx\n", kimage_voffset);
> - DEBUG_MSG("max_physmem_bits : %ld\n", info->max_physmem_bits);
> DEBUG_MSG("section_size_bits: %ld\n", info->section_size_bits);
>
> return TRUE;
> @@ -444,10 +486,33 @@ get_versiondep_info_arm64(void)
> return FALSE;
> }
>
> - info->page_offset = (0xffffffffffffffffUL) << (va_bits - 1);
> + /*
> + * See TCR_EL1, Translation Control Register (EL1) register
> + * description in the ARMv8 Architecture Reference Manual.
> + * Basically, we can use the TCR_EL1.T1SZ
> + * value to determine the virtual addressing range supported
> + * in the kernel-space (i.e. vabits_actual) since Linux 5.9.
> + */
> + if (NUMBER(TCR_EL1_T1SZ) != NOT_FOUND_NUMBER) {
> + vabits_actual = 64 - NUMBER(TCR_EL1_T1SZ);
> + DEBUG_MSG("vabits_actual : %d (vmcoreinfo)\n", vabits_actual);
> + } else if ((va_bits == 52) && (SYMBOL(mem_section) != NOT_FOUND_SYMBOL)) {
> + /*
> + * Linux 5.4 through 5.10 have the following linear space:
> + * 48-bit: 0xffff000000000000 - 0xffff7fffffffffff
> + * 58-bit: 0xfff0000000000000 - 0xfff7ffffffffffff
> + */
> + if (SYMBOL(mem_section) & (1UL << (52 - 1)))
Sorry but I do not think any SYMBOL(x) is inside the range of linear
mapping address. All of them should be beyond kimage_vaddr.
Having vabits_actual is introduced and precise to resolve
is_lm_address(), but if it is not available, could we survive?
_PAGE_OFFSET(52) < _PAGE_END(52) < _PAGE_OFFSET(48) < _PAGE_END(48)
Translating into numbers:
0xfff0000000000000 < 0xfff8000000000000 < 0xffff000000000000 <
0xffff800000000000
Refer to linux/Documentation/arm64/memory.rst, the line
ffffa00010000000 fffff81ffffeffff ~88TB vmalloc
It comes to the conclusion that any symbol > SYMBOL(_text) > _PAGE_END(48).
So is_lm_address() can looks like
if (addr > _PAGE_END(48)), it is kimage
else, it is linear mapping
So even more aggressive, we can exclude the need of vabits_actual
totally in this patch.
Thanks,
Pingfan
_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec
^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [RFC PATCH 4/4] arm64: support flipped VA and 52-bit kernel VA
2021-01-14 10:03 ` piliu
@ 2021-01-14 18:25 ` Bhupesh SHARMA
2021-01-15 0:40 ` HAGIO KAZUHITO(萩尾 一仁)
0 siblings, 1 reply; 10+ messages in thread
From: Bhupesh SHARMA @ 2021-01-14 18:25 UTC (permalink / raw)
To: piliu
Cc: Bhupesh Sharma, kexec, aadiga, Alexander Kamensky,
Ioanna Alifieraki, kazuhito.hagio
Hi Kazu,
On Thu, Jan 14, 2021 at 3:33 PM piliu <piliu@redhat.com> wrote:
>
>
>
> On 1/14/21 4:25 PM, kazuhito.hagio@gmail.com wrote:
> > From: Kazuhito Hagio <k-hagio-ab@nec.com>
> >
> > Based on Bhupesh's patch and contains Pingfan's idea.
> >
> > Signed-off-by: Bhupesh Sharma <bhsharma@redhat.com>
> > Signed-off-by: Kazuhito Hagio <k-hagio-ab@nec.com>
> > ---
> > arch/arm64.c | 95 ++++++++++++++++++++++++++++++++++++++++++++++++----------
> > makedumpfile.c | 2 ++
> > makedumpfile.h | 1 +
> > 3 files changed, 83 insertions(+), 15 deletions(-)
> >
> > diff --git a/arch/arm64.c b/arch/arm64.c
> > index 61ec89a..4ece19d 100644
> > --- a/arch/arm64.c
> > +++ b/arch/arm64.c
> > @@ -47,6 +47,8 @@ typedef struct {
> > static int lpa_52_bit_support_available;
> > static int pgtable_level;
> > static int va_bits;
> > +static int vabits_actual;
> > +static int flipped_va;
> > static unsigned long kimage_voffset;
> >
> > #define SZ_4K 4096
> > @@ -58,7 +60,6 @@ static unsigned long kimage_voffset;
> > #define PAGE_OFFSET_42 ((0xffffffffffffffffUL) << 42)
> > #define PAGE_OFFSET_47 ((0xffffffffffffffffUL) << 47)
> > #define PAGE_OFFSET_48 ((0xffffffffffffffffUL) << 48)
> > -#define PAGE_OFFSET_52 ((0xffffffffffffffffUL) << 52)
> >
> > #define pgd_val(x) ((x).pgd)
> > #define pud_val(x) (pgd_val((x).pgd))
> > @@ -218,12 +219,20 @@ pmd_page_paddr(pmd_t pmd)
> > #define pte_index(vaddr) (((vaddr) >> PAGESHIFT()) & (PTRS_PER_PTE - 1))
> > #define pte_offset(dir, vaddr) (pmd_page_paddr((*dir)) + pte_index(vaddr) * sizeof(pte_t))
> >
> > +/*
> > + * The linear kernel range starts at the bottom of the virtual address
> > + * space. Testing the top bit for the start of the region is a
> > + * sufficient check and avoids having to worry about the tag.
> > + */
> > +#define is_linear_addr(addr) (flipped_va ? \
> > + (!((unsigned long)(addr) & (1UL << (vabits_actual - 1)))) : \
> > + (!!((unsigned long)(addr) & (1UL << (vabits_actual - 1)))))
> > +
> > static unsigned long long
> > __pa(unsigned long vaddr)
> > {
> > - if (kimage_voffset == NOT_FOUND_NUMBER ||
> > - (vaddr >= PAGE_OFFSET))
> > - return (vaddr - PAGE_OFFSET + info->phys_base);
> > + if (kimage_voffset == NOT_FOUND_NUMBER || is_linear_addr(vaddr))
> > + return ((vaddr & ~PAGE_OFFSET) + info->phys_base);
> > else
> > return (vaddr - kimage_voffset);
> > }
> > @@ -253,6 +262,7 @@ static int calculate_plat_config(void)
> > (PAGESIZE() == SZ_64K && va_bits == 42)) {
> > pgtable_level = 2;
> > } else if ((PAGESIZE() == SZ_64K && va_bits == 48) ||
> > + (PAGESIZE() == SZ_64K && va_bits == 52) ||
> > (PAGESIZE() == SZ_4K && va_bits == 39) ||
> > (PAGESIZE() == SZ_16K && va_bits == 47)) {
> > pgtable_level = 3;
> > @@ -263,6 +273,7 @@ static int calculate_plat_config(void)
> > PAGESIZE(), va_bits);
> > return FALSE;
> > }
> > + DEBUG_MSG("pgtable_level: %d\n", pgtable_level);
> >
> > return TRUE;
> > }
> > @@ -383,22 +394,54 @@ get_va_bits_from_stext_arm64(void)
> > return TRUE;
> > }
> >
> > +static void
> > +get_page_offset_arm64(void)
> > +{
> > + ulong page_end;
> > + int vabits_min;
> > +
> > + /*
> > + * See arch/arm64/include/asm/memory.h for more details of
> > + * the PAGE_OFFSET calculation.
> > + */
> > + vabits_min = (va_bits > 48) ? 48 : va_bits;
> > + page_end = -(1UL << (vabits_min - 1));
> > +
> > + if (SYMBOL(_stext) > page_end) {
> > + flipped_va = TRUE;
> > + info->page_offset = -(1UL << vabits_actual);
> > + } else {
> > + flipped_va = FALSE;
> > + info->page_offset = -(1UL << (vabits_actual - 1));
> > + }
> > +
> > + DEBUG_MSG("page_offset : %lx (from page_end check)\n",
> > + info->page_offset);
> > +}
> > +
> > int
> > get_machdep_info_arm64(void)
> > {
> > + /* Check if va_bits is still not initialized. If still 0, call
> > + * get_versiondep_info() to initialize the same.
> > + */
> > + if (!va_bits)
> > + get_versiondep_info_arm64();
> > +
> > /* Determine if the PA address range is 52-bits: ARMv8.2-LPA */
> > if (NUMBER(MAX_PHYSMEM_BITS) != NOT_FOUND_NUMBER) {
> > info->max_physmem_bits = NUMBER(MAX_PHYSMEM_BITS);
> > + DEBUG_MSG("max_physmem_bits : %ld (vmcoreinfo)\n", info->max_physmem_bits);
> > if (info->max_physmem_bits == 52)
> > lpa_52_bit_support_available = 1;
> > - } else
> > - info->max_physmem_bits = 48;
> > + } else {
> > + if (va_bits == 52)
> > + info->max_physmem_bits = 52; /* just guess */
> > + else
> > + info->max_physmem_bits = 48;
> >
> > - /* Check if va_bits is still not initialized. If still 0, call
> > - * get_versiondep_info() to initialize the same.
> > - */
> > - if (!va_bits)
> > - get_versiondep_info_arm64();
> > + DEBUG_MSG("max_physmem_bits : %ld (guess)\n", info->max_physmem_bits);
> > + }
> >
> > if (!calculate_plat_config()) {
> > ERRMSG("Can't determine platform config values\n");
> > @@ -409,7 +452,6 @@ get_machdep_info_arm64(void)
> > info->section_size_bits = SECTIONS_SIZE_BITS;
> >
> > DEBUG_MSG("kimage_voffset : %lx\n", kimage_voffset);
> > - DEBUG_MSG("max_physmem_bits : %ld\n", info->max_physmem_bits);
> > DEBUG_MSG("section_size_bits: %ld\n", info->section_size_bits);
> >
> > return TRUE;
> > @@ -444,10 +486,33 @@ get_versiondep_info_arm64(void)
> > return FALSE;
> > }
> >
> > - info->page_offset = (0xffffffffffffffffUL) << (va_bits - 1);
> > + /*
> > + * See TCR_EL1, Translation Control Register (EL1) register
> > + * description in the ARMv8 Architecture Reference Manual.
> > + * Basically, we can use the TCR_EL1.T1SZ
> > + * value to determine the virtual addressing range supported
> > + * in the kernel-space (i.e. vabits_actual) since Linux 5.9.
> > + */
> > + if (NUMBER(TCR_EL1_T1SZ) != NOT_FOUND_NUMBER) {
> > + vabits_actual = 64 - NUMBER(TCR_EL1_T1SZ);
> > + DEBUG_MSG("vabits_actual : %d (vmcoreinfo)\n", vabits_actual);
> > + } else if ((va_bits == 52) && (SYMBOL(mem_section) != NOT_FOUND_SYMBOL)) {
> > + /*
> > + * Linux 5.4 through 5.10 have the following linear space:
> > + * 48-bit: 0xffff000000000000 - 0xffff7fffffffffff
> > + * 58-bit: 0xfff0000000000000 - 0xfff7ffffffffffff
> > + */
> > + if (SYMBOL(mem_section) & (1UL << (52 - 1)))
>
> Sorry but I do not think any SYMBOL(x) is inside the range of linear
> mapping address. All of them should be beyond kimage_vaddr.
>
> Having vabits_actual is introduced and precise to resolve
> is_lm_address(), but if it is not available, could we survive?
>
> _PAGE_OFFSET(52) < _PAGE_END(52) < _PAGE_OFFSET(48) < _PAGE_END(48)
> Translating into numbers:
> 0xfff0000000000000 < 0xfff8000000000000 < 0xffff000000000000 <
> 0xffff800000000000
>
> Refer to linux/Documentation/arm64/memory.rst, the line
> ffffa00010000000 fffff81ffffeffff ~88TB vmalloc
>
> It comes to the conclusion that any symbol > SYMBOL(_text) > _PAGE_END(48).
>
> So is_lm_address() can looks like
> if (addr > _PAGE_END(48)), it is kimage
> else, it is linear mapping
>
> So even more aggressive, we can exclude the need of vabits_actual
> totally in this patch.
As Pingfan noted, this approach still has issues and it failed on my
arm64 board in the limited testing I did today.
I will do some more testing tomorrow and come back with an approach
which addresses the same and also takes into account Pingfan's
concerns.
Thanks,
Bhupesh
_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec
^ permalink raw reply [flat|nested] 10+ messages in thread
* RE: [RFC PATCH 4/4] arm64: support flipped VA and 52-bit kernel VA
2021-01-14 18:25 ` Bhupesh SHARMA
@ 2021-01-15 0:40 ` HAGIO KAZUHITO(萩尾 一仁)
2021-01-15 1:41 ` piliu
0 siblings, 1 reply; 10+ messages in thread
From: HAGIO KAZUHITO(萩尾 一仁) @ 2021-01-15 0:40 UTC (permalink / raw)
To: Bhupesh SHARMA, piliu
Cc: Bhupesh Sharma, kexec, aadiga, kazuhito.hagio, Ioanna Alifieraki,
Alexander Kamensky
Hi Pingfan, Bhupesh,
Thank you for your comments!
-----Original Message-----
> Hi Kazu,
>
> On Thu, Jan 14, 2021 at 3:33 PM piliu <piliu@redhat.com> wrote:
> >
> >
> >
> > On 1/14/21 4:25 PM, kazuhito.hagio@gmail.com wrote:
> > > From: Kazuhito Hagio <k-hagio-ab@nec.com>
> > >
> > > Based on Bhupesh's patch and contains Pingfan's idea.
> > >
> > > Signed-off-by: Bhupesh Sharma <bhsharma@redhat.com>
> > > Signed-off-by: Kazuhito Hagio <k-hagio-ab@nec.com>
> > > ---
> > > arch/arm64.c | 95 ++++++++++++++++++++++++++++++++++++++++++++++++----------
> > > makedumpfile.c | 2 ++
> > > makedumpfile.h | 1 +
> > > 3 files changed, 83 insertions(+), 15 deletions(-)
> > >
> > > diff --git a/arch/arm64.c b/arch/arm64.c
> > > index 61ec89a..4ece19d 100644
> > > --- a/arch/arm64.c
> > > +++ b/arch/arm64.c
> > > @@ -47,6 +47,8 @@ typedef struct {
> > > static int lpa_52_bit_support_available;
> > > static int pgtable_level;
> > > static int va_bits;
> > > +static int vabits_actual;
> > > +static int flipped_va;
> > > static unsigned long kimage_voffset;
> > >
> > > #define SZ_4K 4096
> > > @@ -58,7 +60,6 @@ static unsigned long kimage_voffset;
> > > #define PAGE_OFFSET_42 ((0xffffffffffffffffUL) << 42)
> > > #define PAGE_OFFSET_47 ((0xffffffffffffffffUL) << 47)
> > > #define PAGE_OFFSET_48 ((0xffffffffffffffffUL) << 48)
> > > -#define PAGE_OFFSET_52 ((0xffffffffffffffffUL) << 52)
> > >
> > > #define pgd_val(x) ((x).pgd)
> > > #define pud_val(x) (pgd_val((x).pgd))
> > > @@ -218,12 +219,20 @@ pmd_page_paddr(pmd_t pmd)
> > > #define pte_index(vaddr) (((vaddr) >> PAGESHIFT()) & (PTRS_PER_PTE - 1))
> > > #define pte_offset(dir, vaddr) (pmd_page_paddr((*dir)) + pte_index(vaddr) *
> sizeof(pte_t))
> > >
> > > +/*
> > > + * The linear kernel range starts at the bottom of the virtual address
> > > + * space. Testing the top bit for the start of the region is a
> > > + * sufficient check and avoids having to worry about the tag.
> > > + */
> > > +#define is_linear_addr(addr) (flipped_va ? \
> > > + (!((unsigned long)(addr) & (1UL << (vabits_actual - 1)))) : \
> > > + (!!((unsigned long)(addr) & (1UL << (vabits_actual - 1)))))
> > > +
> > > static unsigned long long
> > > __pa(unsigned long vaddr)
> > > {
> > > - if (kimage_voffset == NOT_FOUND_NUMBER ||
> > > - (vaddr >= PAGE_OFFSET))
> > > - return (vaddr - PAGE_OFFSET + info->phys_base);
> > > + if (kimage_voffset == NOT_FOUND_NUMBER || is_linear_addr(vaddr))
> > > + return ((vaddr & ~PAGE_OFFSET) + info->phys_base);
> > > else
> > > return (vaddr - kimage_voffset);
> > > }
> > > @@ -253,6 +262,7 @@ static int calculate_plat_config(void)
> > > (PAGESIZE() == SZ_64K && va_bits == 42)) {
> > > pgtable_level = 2;
> > > } else if ((PAGESIZE() == SZ_64K && va_bits == 48) ||
> > > + (PAGESIZE() == SZ_64K && va_bits == 52) ||
> > > (PAGESIZE() == SZ_4K && va_bits == 39) ||
> > > (PAGESIZE() == SZ_16K && va_bits == 47)) {
> > > pgtable_level = 3;
> > > @@ -263,6 +273,7 @@ static int calculate_plat_config(void)
> > > PAGESIZE(), va_bits);
> > > return FALSE;
> > > }
> > > + DEBUG_MSG("pgtable_level: %d\n", pgtable_level);
> > >
> > > return TRUE;
> > > }
> > > @@ -383,22 +394,54 @@ get_va_bits_from_stext_arm64(void)
> > > return TRUE;
> > > }
> > >
> > > +static void
> > > +get_page_offset_arm64(void)
> > > +{
> > > + ulong page_end;
> > > + int vabits_min;
> > > +
> > > + /*
> > > + * See arch/arm64/include/asm/memory.h for more details of
> > > + * the PAGE_OFFSET calculation.
> > > + */
> > > + vabits_min = (va_bits > 48) ? 48 : va_bits;
> > > + page_end = -(1UL << (vabits_min - 1));
> > > +
> > > + if (SYMBOL(_stext) > page_end) {
> > > + flipped_va = TRUE;
> > > + info->page_offset = -(1UL << vabits_actual);
> > > + } else {
> > > + flipped_va = FALSE;
> > > + info->page_offset = -(1UL << (vabits_actual - 1));
> > > + }
> > > +
> > > + DEBUG_MSG("page_offset : %lx (from page_end check)\n",
> > > + info->page_offset);
> > > +}
> > > +
> > > int
> > > get_machdep_info_arm64(void)
> > > {
> > > + /* Check if va_bits is still not initialized. If still 0, call
> > > + * get_versiondep_info() to initialize the same.
> > > + */
> > > + if (!va_bits)
> > > + get_versiondep_info_arm64();
> > > +
> > > /* Determine if the PA address range is 52-bits: ARMv8.2-LPA */
> > > if (NUMBER(MAX_PHYSMEM_BITS) != NOT_FOUND_NUMBER) {
> > > info->max_physmem_bits = NUMBER(MAX_PHYSMEM_BITS);
> > > + DEBUG_MSG("max_physmem_bits : %ld (vmcoreinfo)\n", info->max_physmem_bits);
> > > if (info->max_physmem_bits == 52)
> > > lpa_52_bit_support_available = 1;
> > > - } else
> > > - info->max_physmem_bits = 48;
> > > + } else {
> > > + if (va_bits == 52)
> > > + info->max_physmem_bits = 52; /* just guess */
> > > + else
> > > + info->max_physmem_bits = 48;
> > >
> > > - /* Check if va_bits is still not initialized. If still 0, call
> > > - * get_versiondep_info() to initialize the same.
> > > - */
> > > - if (!va_bits)
> > > - get_versiondep_info_arm64();
> > > + DEBUG_MSG("max_physmem_bits : %ld (guess)\n", info->max_physmem_bits);
> > > + }
> > >
> > > if (!calculate_plat_config()) {
> > > ERRMSG("Can't determine platform config values\n");
> > > @@ -409,7 +452,6 @@ get_machdep_info_arm64(void)
> > > info->section_size_bits = SECTIONS_SIZE_BITS;
> > >
> > > DEBUG_MSG("kimage_voffset : %lx\n", kimage_voffset);
> > > - DEBUG_MSG("max_physmem_bits : %ld\n", info->max_physmem_bits);
> > > DEBUG_MSG("section_size_bits: %ld\n", info->section_size_bits);
> > >
> > > return TRUE;
> > > @@ -444,10 +486,33 @@ get_versiondep_info_arm64(void)
> > > return FALSE;
> > > }
> > >
> > > - info->page_offset = (0xffffffffffffffffUL) << (va_bits - 1);
> > > + /*
> > > + * See TCR_EL1, Translation Control Register (EL1) register
> > > + * description in the ARMv8 Architecture Reference Manual.
> > > + * Basically, we can use the TCR_EL1.T1SZ
> > > + * value to determine the virtual addressing range supported
> > > + * in the kernel-space (i.e. vabits_actual) since Linux 5.9.
> > > + */
> > > + if (NUMBER(TCR_EL1_T1SZ) != NOT_FOUND_NUMBER) {
> > > + vabits_actual = 64 - NUMBER(TCR_EL1_T1SZ);
> > > + DEBUG_MSG("vabits_actual : %d (vmcoreinfo)\n", vabits_actual);
> > > + } else if ((va_bits == 52) && (SYMBOL(mem_section) != NOT_FOUND_SYMBOL)) {
> > > + /*
> > > + * Linux 5.4 through 5.10 have the following linear space:
> > > + * 48-bit: 0xffff000000000000 - 0xffff7fffffffffff
> > > + * 58-bit: 0xfff0000000000000 - 0xfff7ffffffffffff
> > > + */
> > > + if (SYMBOL(mem_section) & (1UL << (52 - 1)))
> >
> > Sorry but I do not think any SYMBOL(x) is inside the range of linear
> > mapping address. All of them should be beyond kimage_vaddr.
As for this, SYMBOL(mem_section) is a bit special and shows the value in the
mem_section variable, not the address of it, by kernel commit a0b1280368d1.
Actually on my tiny arm64 board..
# strings /proc/kcore | less
...
OSRELEASE=5.4.0-1025-raspi
PAGESIZE=4096
SYMBOL(init_uts_ns)=ffffd6db0fa10568
SYMBOL(node_online_map)=ffffd6db0fa0a808
SYMBOL(swapper_pg_dir)=ffffd6db0f42b000
SYMBOL(_stext)=ffffd6db0e481000
SYMBOL(vmap_area_list)=ffffd6db0fa51680
SYMBOL(mem_map)=ffffd6db0fbc0240
SYMBOL(contig_page_data)=ffffd6db0faed6c0
SYMBOL(mem_section)=ffff0000eb809300 <<--
LENGTH(mem_section)=1024
SIZE(mem_section)=16
OFFSET(mem_section.section_mem_map)=0
...
NUMBER(VA_BITS)=48
So at least if the kernel is configured with CONFIG_SPARSEMEM_EXTREME,
I thought this would work. This is just a user tool effort to broaden
supported kernels, instead of necessary information from the kernel,
no need to support all kernels.
And the current makedumpfile's implementation, SYMBOL(mem_section) is set
to the address of the mem_section variable with -x option, so I wrote on
the cover letter that this patchset doesn't support "with -x option if
vabits_actual=52" case yet.
> >
> > Having vabits_actual is introduced and precise to resolve
> > is_lm_address(), but if it is not available, could we survive?
> >
> > _PAGE_OFFSET(52) < _PAGE_END(52) < _PAGE_OFFSET(48) < _PAGE_END(48)
> > Translating into numbers:
> > 0xfff0000000000000 < 0xfff8000000000000 < 0xffff000000000000 <
> > 0xffff800000000000
> >
> > Refer to linux/Documentation/arm64/memory.rst, the line
> > ffffa00010000000 fffff81ffffeffff ~88TB vmalloc
> >
> > It comes to the conclusion that any symbol > SYMBOL(_text) > _PAGE_END(48).
> >
> > So is_lm_address() can looks like
> > if (addr > _PAGE_END(48)), it is kimage
> > else, it is linear mapping
This sounds good and will work.
> >
> > So even more aggressive, we can exclude the need of vabits_actual
> > totally in this patch.
Interesting, but if vabits_actual is not available, I wonder how we can
get info->page_offset. Seems it's used mainly for p2v conversion and
is_kvaddr(), which still has a problem though.
>
> As Pingfan noted, this approach still has issues and it failed on my
> arm64 board in the limited testing I did today.
Can I have what configuration, error and debug messages you see,
for reference?
> I will do some more testing tomorrow and come back with an approach
> which addresses the same and also takes into account Pingfan's
> concerns.
ok, thanks!
Kazu
_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec
^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [RFC PATCH 4/4] arm64: support flipped VA and 52-bit kernel VA
2021-01-15 0:40 ` HAGIO KAZUHITO(萩尾 一仁)
@ 2021-01-15 1:41 ` piliu
2021-01-15 3:16 ` HAGIO KAZUHITO(萩尾 一仁)
0 siblings, 1 reply; 10+ messages in thread
From: piliu @ 2021-01-15 1:41 UTC (permalink / raw)
To: HAGIO KAZUHITO(萩尾 一仁), Bhupesh SHARMA
Cc: Bhupesh Sharma, kexec, aadiga, kazuhito.hagio, Ioanna Alifieraki,
Alexander Kamensky
On 1/15/21 8:40 AM, HAGIO KAZUHITO(萩尾 一仁) wrote:
> Hi Pingfan, Bhupesh,
>
> Thank you for your comments!
>
> -----Original Message-----
>> Hi Kazu,
>>
>> On Thu, Jan 14, 2021 at 3:33 PM piliu <piliu@redhat.com> wrote:
>>>
>>>
>>>
>>> On 1/14/21 4:25 PM, kazuhito.hagio@gmail.com wrote:
>>>> From: Kazuhito Hagio <k-hagio-ab@nec.com>
>>>>
>>>> Based on Bhupesh's patch and contains Pingfan's idea.
>>>>
>>>> Signed-off-by: Bhupesh Sharma <bhsharma@redhat.com>
>>>> Signed-off-by: Kazuhito Hagio <k-hagio-ab@nec.com>
>>>> ---
>>>> arch/arm64.c | 95 ++++++++++++++++++++++++++++++++++++++++++++++++----------
>>>> makedumpfile.c | 2 ++
>>>> makedumpfile.h | 1 +
>>>> 3 files changed, 83 insertions(+), 15 deletions(-)
>>>>
>>>> diff --git a/arch/arm64.c b/arch/arm64.c
>>>> index 61ec89a..4ece19d 100644
>>>> --- a/arch/arm64.c
>>>> +++ b/arch/arm64.c
>>>> @@ -47,6 +47,8 @@ typedef struct {
>>>> static int lpa_52_bit_support_available;
>>>> static int pgtable_level;
>>>> static int va_bits;
>>>> +static int vabits_actual;
>>>> +static int flipped_va;
>>>> static unsigned long kimage_voffset;
>>>>
>>>> #define SZ_4K 4096
>>>> @@ -58,7 +60,6 @@ static unsigned long kimage_voffset;
>>>> #define PAGE_OFFSET_42 ((0xffffffffffffffffUL) << 42)
>>>> #define PAGE_OFFSET_47 ((0xffffffffffffffffUL) << 47)
>>>> #define PAGE_OFFSET_48 ((0xffffffffffffffffUL) << 48)
>>>> -#define PAGE_OFFSET_52 ((0xffffffffffffffffUL) << 52)
>>>>
>>>> #define pgd_val(x) ((x).pgd)
>>>> #define pud_val(x) (pgd_val((x).pgd))
>>>> @@ -218,12 +219,20 @@ pmd_page_paddr(pmd_t pmd)
>>>> #define pte_index(vaddr) (((vaddr) >> PAGESHIFT()) & (PTRS_PER_PTE - 1))
>>>> #define pte_offset(dir, vaddr) (pmd_page_paddr((*dir)) + pte_index(vaddr) *
>> sizeof(pte_t))
>>>>
>>>> +/*
>>>> + * The linear kernel range starts at the bottom of the virtual address
>>>> + * space. Testing the top bit for the start of the region is a
>>>> + * sufficient check and avoids having to worry about the tag.
>>>> + */
>>>> +#define is_linear_addr(addr) (flipped_va ? \
>>>> + (!((unsigned long)(addr) & (1UL << (vabits_actual - 1)))) : \
>>>> + (!!((unsigned long)(addr) & (1UL << (vabits_actual - 1)))))
>>>> +
>>>> static unsigned long long
>>>> __pa(unsigned long vaddr)
>>>> {
>>>> - if (kimage_voffset == NOT_FOUND_NUMBER ||
>>>> - (vaddr >= PAGE_OFFSET))
>>>> - return (vaddr - PAGE_OFFSET + info->phys_base);
>>>> + if (kimage_voffset == NOT_FOUND_NUMBER || is_linear_addr(vaddr))
>>>> + return ((vaddr & ~PAGE_OFFSET) + info->phys_base);
>>>> else
>>>> return (vaddr - kimage_voffset);
>>>> }
>>>> @@ -253,6 +262,7 @@ static int calculate_plat_config(void)
>>>> (PAGESIZE() == SZ_64K && va_bits == 42)) {
>>>> pgtable_level = 2;
>>>> } else if ((PAGESIZE() == SZ_64K && va_bits == 48) ||
>>>> + (PAGESIZE() == SZ_64K && va_bits == 52) ||
>>>> (PAGESIZE() == SZ_4K && va_bits == 39) ||
>>>> (PAGESIZE() == SZ_16K && va_bits == 47)) {
>>>> pgtable_level = 3;
>>>> @@ -263,6 +273,7 @@ static int calculate_plat_config(void)
>>>> PAGESIZE(), va_bits);
>>>> return FALSE;
>>>> }
>>>> + DEBUG_MSG("pgtable_level: %d\n", pgtable_level);
>>>>
>>>> return TRUE;
>>>> }
>>>> @@ -383,22 +394,54 @@ get_va_bits_from_stext_arm64(void)
>>>> return TRUE;
>>>> }
>>>>
>>>> +static void
>>>> +get_page_offset_arm64(void)
>>>> +{
>>>> + ulong page_end;
>>>> + int vabits_min;
>>>> +
>>>> + /*
>>>> + * See arch/arm64/include/asm/memory.h for more details of
>>>> + * the PAGE_OFFSET calculation.
>>>> + */
>>>> + vabits_min = (va_bits > 48) ? 48 : va_bits;
>>>> + page_end = -(1UL << (vabits_min - 1));
>>>> +
>>>> + if (SYMBOL(_stext) > page_end) {
>>>> + flipped_va = TRUE;
>>>> + info->page_offset = -(1UL << vabits_actual);
>>>> + } else {
>>>> + flipped_va = FALSE;
>>>> + info->page_offset = -(1UL << (vabits_actual - 1));
>>>> + }
>>>> +
>>>> + DEBUG_MSG("page_offset : %lx (from page_end check)\n",
>>>> + info->page_offset);
>>>> +}
>>>> +
>>>> int
>>>> get_machdep_info_arm64(void)
>>>> {
>>>> + /* Check if va_bits is still not initialized. If still 0, call
>>>> + * get_versiondep_info() to initialize the same.
>>>> + */
>>>> + if (!va_bits)
>>>> + get_versiondep_info_arm64();
>>>> +
>>>> /* Determine if the PA address range is 52-bits: ARMv8.2-LPA */
>>>> if (NUMBER(MAX_PHYSMEM_BITS) != NOT_FOUND_NUMBER) {
>>>> info->max_physmem_bits = NUMBER(MAX_PHYSMEM_BITS);
>>>> + DEBUG_MSG("max_physmem_bits : %ld (vmcoreinfo)\n", info->max_physmem_bits);
>>>> if (info->max_physmem_bits == 52)
>>>> lpa_52_bit_support_available = 1;
>>>> - } else
>>>> - info->max_physmem_bits = 48;
>>>> + } else {
>>>> + if (va_bits == 52)
>>>> + info->max_physmem_bits = 52; /* just guess */
>>>> + else
>>>> + info->max_physmem_bits = 48;
>>>>
>>>> - /* Check if va_bits is still not initialized. If still 0, call
>>>> - * get_versiondep_info() to initialize the same.
>>>> - */
>>>> - if (!va_bits)
>>>> - get_versiondep_info_arm64();
>>>> + DEBUG_MSG("max_physmem_bits : %ld (guess)\n", info->max_physmem_bits);
>>>> + }
>>>>
>>>> if (!calculate_plat_config()) {
>>>> ERRMSG("Can't determine platform config values\n");
>>>> @@ -409,7 +452,6 @@ get_machdep_info_arm64(void)
>>>> info->section_size_bits = SECTIONS_SIZE_BITS;
>>>>
>>>> DEBUG_MSG("kimage_voffset : %lx\n", kimage_voffset);
>>>> - DEBUG_MSG("max_physmem_bits : %ld\n", info->max_physmem_bits);
>>>> DEBUG_MSG("section_size_bits: %ld\n", info->section_size_bits);
>>>>
>>>> return TRUE;
>>>> @@ -444,10 +486,33 @@ get_versiondep_info_arm64(void)
>>>> return FALSE;
>>>> }
>>>>
>>>> - info->page_offset = (0xffffffffffffffffUL) << (va_bits - 1);
>>>> + /*
>>>> + * See TCR_EL1, Translation Control Register (EL1) register
>>>> + * description in the ARMv8 Architecture Reference Manual.
>>>> + * Basically, we can use the TCR_EL1.T1SZ
>>>> + * value to determine the virtual addressing range supported
>>>> + * in the kernel-space (i.e. vabits_actual) since Linux 5.9.
>>>> + */
>>>> + if (NUMBER(TCR_EL1_T1SZ) != NOT_FOUND_NUMBER) {
>>>> + vabits_actual = 64 - NUMBER(TCR_EL1_T1SZ);
>>>> + DEBUG_MSG("vabits_actual : %d (vmcoreinfo)\n", vabits_actual);
>>>> + } else if ((va_bits == 52) && (SYMBOL(mem_section) != NOT_FOUND_SYMBOL)) {
>>>> + /*
>>>> + * Linux 5.4 through 5.10 have the following linear space:
>>>> + * 48-bit: 0xffff000000000000 - 0xffff7fffffffffff
>>>> + * 58-bit: 0xfff0000000000000 - 0xfff7ffffffffffff
>>>> + */
>>>> + if (SYMBOL(mem_section) & (1UL << (52 - 1)))
>>>
>>> Sorry but I do not think any SYMBOL(x) is inside the range of linear
>>> mapping address. All of them should be beyond kimage_vaddr.
>
> As for this, SYMBOL(mem_section) is a bit special and shows the value in the
> mem_section variable, not the address of it, by kernel commit a0b1280368d1.
>
> Actually on my tiny arm64 board..
>
> # strings /proc/kcore | less
> ...
> OSRELEASE=5.4.0-1025-raspi
> PAGESIZE=4096
> SYMBOL(init_uts_ns)=ffffd6db0fa10568
> SYMBOL(node_online_map)=ffffd6db0fa0a808
> SYMBOL(swapper_pg_dir)=ffffd6db0f42b000
> SYMBOL(_stext)=ffffd6db0e481000
> SYMBOL(vmap_area_list)=ffffd6db0fa51680
> SYMBOL(mem_map)=ffffd6db0fbc0240
> SYMBOL(contig_page_data)=ffffd6db0faed6c0
> SYMBOL(mem_section)=ffff0000eb809300 <<--
> LENGTH(mem_section)=1024
> SIZE(mem_section)=16
> OFFSET(mem_section.section_mem_map)=0
> ...
> NUMBER(VA_BITS)=48
>
> So at least if the kernel is configured with CONFIG_SPARSEMEM_EXTREME,
You are right. Exactly, your method works in this case.
> I thought this would work. This is just a user tool effort to broaden
> supported kernels, instead of necessary information from the kernel,
> no need to support all kernels.
I hold a private opinion against the solution based on kernel version,
since it raises extra effort to cope with different releases which back
port some upstream patches without advancing kernel version.
>
> And the current makedumpfile's implementation, SYMBOL(mem_section) is set
> to the address of the mem_section variable with -x option, so I wrote on
> the cover letter that this patchset doesn't support "with -x option if
> vabits_actual=52" case yet.
>
>>>
>>> Having vabits_actual is introduced and precise to resolve
>>> is_lm_address(), but if it is not available, could we survive?
>>>
>>> _PAGE_OFFSET(52) < _PAGE_END(52) < _PAGE_OFFSET(48) < _PAGE_END(48)
>>> Translating into numbers:
>>> 0xfff0000000000000 < 0xfff8000000000000 < 0xffff000000000000 <
>>> 0xffff800000000000
>>>
>>> Refer to linux/Documentation/arm64/memory.rst, the line
>>> ffffa00010000000 fffff81ffffeffff ~88TB vmalloc
>>>
>>> It comes to the conclusion that any symbol > SYMBOL(_text) > _PAGE_END(48).
>>>
>>> So is_lm_address() can looks like
>>> if (addr > _PAGE_END(48)), it is kimage
>>> else, it is linear mapping
>
> This sounds good and will work.
>
>>>
>>> So even more aggressive, we can exclude the need of vabits_actual
>>> totally in this patch.
>
> Interesting, but if vabits_actual is not available, I wonder how we can
> get info->page_offset. Seems it's used mainly for p2v conversion and
Ashamed. I will make the best of the fact that va_bits=52 kernel only
has _PAGE_OFFSET(52). It looks not elegant, but should work.
Regards,
Pingfan
_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec
^ permalink raw reply [flat|nested] 10+ messages in thread
* RE: [RFC PATCH 4/4] arm64: support flipped VA and 52-bit kernel VA
2021-01-15 1:41 ` piliu
@ 2021-01-15 3:16 ` HAGIO KAZUHITO(萩尾 一仁)
0 siblings, 0 replies; 10+ messages in thread
From: HAGIO KAZUHITO(萩尾 一仁) @ 2021-01-15 3:16 UTC (permalink / raw)
To: piliu, Bhupesh SHARMA
Cc: Bhupesh Sharma, kexec, aadiga, kazuhito.hagio, Ioanna Alifieraki,
Alexander Kamensky
Hi Pingfan,
-----Original Message-----
> >>> On 1/14/21 4:25 PM, kazuhito.hagio@gmail.com wrote:
> >>>> From: Kazuhito Hagio <k-hagio-ab@nec.com>
> >>>>
> >>>> Based on Bhupesh's patch and contains Pingfan's idea.
> >>>>
> >>>> Signed-off-by: Bhupesh Sharma <bhsharma@redhat.com>
> >>>> Signed-off-by: Kazuhito Hagio <k-hagio-ab@nec.com>
> >>>> ---
> >>>> arch/arm64.c | 95 ++++++++++++++++++++++++++++++++++++++++++++++++----------
> >>>> makedumpfile.c | 2 ++
> >>>> makedumpfile.h | 1 +
> >>>> 3 files changed, 83 insertions(+), 15 deletions(-)
> >>>>
> >>>> diff --git a/arch/arm64.c b/arch/arm64.c
> >>>> index 61ec89a..4ece19d 100644
> >>>> --- a/arch/arm64.c
> >>>> +++ b/arch/arm64.c
> >>>> @@ -47,6 +47,8 @@ typedef struct {
> >>>> static int lpa_52_bit_support_available;
> >>>> static int pgtable_level;
> >>>> static int va_bits;
> >>>> +static int vabits_actual;
> >>>> +static int flipped_va;
> >>>> static unsigned long kimage_voffset;
> >>>>
> >>>> #define SZ_4K 4096
> >>>> @@ -58,7 +60,6 @@ static unsigned long kimage_voffset;
> >>>> #define PAGE_OFFSET_42 ((0xffffffffffffffffUL) << 42)
> >>>> #define PAGE_OFFSET_47 ((0xffffffffffffffffUL) << 47)
> >>>> #define PAGE_OFFSET_48 ((0xffffffffffffffffUL) << 48)
> >>>> -#define PAGE_OFFSET_52 ((0xffffffffffffffffUL) << 52)
> >>>>
> >>>> #define pgd_val(x) ((x).pgd)
> >>>> #define pud_val(x) (pgd_val((x).pgd))
> >>>> @@ -218,12 +219,20 @@ pmd_page_paddr(pmd_t pmd)
> >>>> #define pte_index(vaddr) (((vaddr) >> PAGESHIFT()) & (PTRS_PER_PTE - 1))
> >>>> #define pte_offset(dir, vaddr) (pmd_page_paddr((*dir)) + pte_index(vaddr) *
> >> sizeof(pte_t))
> >>>>
> >>>> +/*
> >>>> + * The linear kernel range starts at the bottom of the virtual address
> >>>> + * space. Testing the top bit for the start of the region is a
> >>>> + * sufficient check and avoids having to worry about the tag.
> >>>> + */
> >>>> +#define is_linear_addr(addr) (flipped_va ? \
> >>>> + (!((unsigned long)(addr) & (1UL << (vabits_actual - 1)))) : \
> >>>> + (!!((unsigned long)(addr) & (1UL << (vabits_actual - 1)))))
> >>>> +
> >>>> static unsigned long long
> >>>> __pa(unsigned long vaddr)
> >>>> {
> >>>> - if (kimage_voffset == NOT_FOUND_NUMBER ||
> >>>> - (vaddr >= PAGE_OFFSET))
> >>>> - return (vaddr - PAGE_OFFSET + info->phys_base);
> >>>> + if (kimage_voffset == NOT_FOUND_NUMBER || is_linear_addr(vaddr))
> >>>> + return ((vaddr & ~PAGE_OFFSET) + info->phys_base);
> >>>> else
> >>>> return (vaddr - kimage_voffset);
> >>>> }
> >>>> @@ -253,6 +262,7 @@ static int calculate_plat_config(void)
> >>>> (PAGESIZE() == SZ_64K && va_bits == 42)) {
> >>>> pgtable_level = 2;
> >>>> } else if ((PAGESIZE() == SZ_64K && va_bits == 48) ||
> >>>> + (PAGESIZE() == SZ_64K && va_bits == 52) ||
> >>>> (PAGESIZE() == SZ_4K && va_bits == 39) ||
> >>>> (PAGESIZE() == SZ_16K && va_bits == 47)) {
> >>>> pgtable_level = 3;
> >>>> @@ -263,6 +273,7 @@ static int calculate_plat_config(void)
> >>>> PAGESIZE(), va_bits);
> >>>> return FALSE;
> >>>> }
> >>>> + DEBUG_MSG("pgtable_level: %d\n", pgtable_level);
> >>>>
> >>>> return TRUE;
> >>>> }
> >>>> @@ -383,22 +394,54 @@ get_va_bits_from_stext_arm64(void)
> >>>> return TRUE;
> >>>> }
> >>>>
> >>>> +static void
> >>>> +get_page_offset_arm64(void)
> >>>> +{
> >>>> + ulong page_end;
> >>>> + int vabits_min;
> >>>> +
> >>>> + /*
> >>>> + * See arch/arm64/include/asm/memory.h for more details of
> >>>> + * the PAGE_OFFSET calculation.
> >>>> + */
> >>>> + vabits_min = (va_bits > 48) ? 48 : va_bits;
> >>>> + page_end = -(1UL << (vabits_min - 1));
> >>>> +
> >>>> + if (SYMBOL(_stext) > page_end) {
> >>>> + flipped_va = TRUE;
> >>>> + info->page_offset = -(1UL << vabits_actual);
> >>>> + } else {
> >>>> + flipped_va = FALSE;
> >>>> + info->page_offset = -(1UL << (vabits_actual - 1));
> >>>> + }
> >>>> +
> >>>> + DEBUG_MSG("page_offset : %lx (from page_end check)\n",
> >>>> + info->page_offset);
> >>>> +}
> >>>> +
> >>>> int
> >>>> get_machdep_info_arm64(void)
> >>>> {
> >>>> + /* Check if va_bits is still not initialized. If still 0, call
> >>>> + * get_versiondep_info() to initialize the same.
> >>>> + */
> >>>> + if (!va_bits)
> >>>> + get_versiondep_info_arm64();
> >>>> +
> >>>> /* Determine if the PA address range is 52-bits: ARMv8.2-LPA */
> >>>> if (NUMBER(MAX_PHYSMEM_BITS) != NOT_FOUND_NUMBER) {
> >>>> info->max_physmem_bits = NUMBER(MAX_PHYSMEM_BITS);
> >>>> + DEBUG_MSG("max_physmem_bits : %ld (vmcoreinfo)\n", info->max_physmem_bits);
> >>>> if (info->max_physmem_bits == 52)
> >>>> lpa_52_bit_support_available = 1;
> >>>> - } else
> >>>> - info->max_physmem_bits = 48;
> >>>> + } else {
> >>>> + if (va_bits == 52)
> >>>> + info->max_physmem_bits = 52; /* just guess */
> >>>> + else
> >>>> + info->max_physmem_bits = 48;
> >>>>
> >>>> - /* Check if va_bits is still not initialized. If still 0, call
> >>>> - * get_versiondep_info() to initialize the same.
> >>>> - */
> >>>> - if (!va_bits)
> >>>> - get_versiondep_info_arm64();
> >>>> + DEBUG_MSG("max_physmem_bits : %ld (guess)\n", info->max_physmem_bits);
> >>>> + }
> >>>>
> >>>> if (!calculate_plat_config()) {
> >>>> ERRMSG("Can't determine platform config values\n");
> >>>> @@ -409,7 +452,6 @@ get_machdep_info_arm64(void)
> >>>> info->section_size_bits = SECTIONS_SIZE_BITS;
> >>>>
> >>>> DEBUG_MSG("kimage_voffset : %lx\n", kimage_voffset);
> >>>> - DEBUG_MSG("max_physmem_bits : %ld\n", info->max_physmem_bits);
> >>>> DEBUG_MSG("section_size_bits: %ld\n", info->section_size_bits);
> >>>>
> >>>> return TRUE;
> >>>> @@ -444,10 +486,33 @@ get_versiondep_info_arm64(void)
> >>>> return FALSE;
> >>>> }
> >>>>
> >>>> - info->page_offset = (0xffffffffffffffffUL) << (va_bits - 1);
> >>>> + /*
> >>>> + * See TCR_EL1, Translation Control Register (EL1) register
> >>>> + * description in the ARMv8 Architecture Reference Manual.
> >>>> + * Basically, we can use the TCR_EL1.T1SZ
> >>>> + * value to determine the virtual addressing range supported
> >>>> + * in the kernel-space (i.e. vabits_actual) since Linux 5.9.
> >>>> + */
> >>>> + if (NUMBER(TCR_EL1_T1SZ) != NOT_FOUND_NUMBER) {
> >>>> + vabits_actual = 64 - NUMBER(TCR_EL1_T1SZ);
> >>>> + DEBUG_MSG("vabits_actual : %d (vmcoreinfo)\n", vabits_actual);
> >>>> + } else if ((va_bits == 52) && (SYMBOL(mem_section) != NOT_FOUND_SYMBOL)) {
> >>>> + /*
> >>>> + * Linux 5.4 through 5.10 have the following linear space:
> >>>> + * 48-bit: 0xffff000000000000 - 0xffff7fffffffffff
> >>>> + * 58-bit: 0xfff0000000000000 - 0xfff7ffffffffffff
> >>>> + */
> >>>> + if (SYMBOL(mem_section) & (1UL << (52 - 1)))
> >>>
> >>> Sorry but I do not think any SYMBOL(x) is inside the range of linear
> >>> mapping address. All of them should be beyond kimage_vaddr.
> >
> > As for this, SYMBOL(mem_section) is a bit special and shows the value in the
> > mem_section variable, not the address of it, by kernel commit a0b1280368d1.
> >
> > Actually on my tiny arm64 board..
> >
> > # strings /proc/kcore | less
> > ...
> > OSRELEASE=5.4.0-1025-raspi
> > PAGESIZE=4096
> > SYMBOL(init_uts_ns)=ffffd6db0fa10568
> > SYMBOL(node_online_map)=ffffd6db0fa0a808
> > SYMBOL(swapper_pg_dir)=ffffd6db0f42b000
> > SYMBOL(_stext)=ffffd6db0e481000
> > SYMBOL(vmap_area_list)=ffffd6db0fa51680
> > SYMBOL(mem_map)=ffffd6db0fbc0240
> > SYMBOL(contig_page_data)=ffffd6db0faed6c0
> > SYMBOL(mem_section)=ffff0000eb809300 <<--
> > LENGTH(mem_section)=1024
> > SIZE(mem_section)=16
> > OFFSET(mem_section.section_mem_map)=0
> > ...
> > NUMBER(VA_BITS)=48
> >
> > So at least if the kernel is configured with CONFIG_SPARSEMEM_EXTREME,
> You are right. Exactly, your method works in this case.
>
> > I thought this would work. This is just a user tool effort to broaden
> > supported kernels, instead of necessary information from the kernel,
> > no need to support all kernels.
>
> I hold a private opinion against the solution based on kernel version,
> since it raises extra effort to cope with different releases which back
> port some upstream patches without advancing kernel version.
I agree with you, we should avoid depending on kernel version if we can.
I left the 3/4 patch intact for possible use of kernel version method,
but for now, the patchset doesn't use kernel version with your idea! :)
> >
> > And the current makedumpfile's implementation, SYMBOL(mem_section) is set
> > to the address of the mem_section variable with -x option, so I wrote on
> > the cover letter that this patchset doesn't support "with -x option if
> > vabits_actual=52" case yet.
> >
> >>>
> >>> Having vabits_actual is introduced and precise to resolve
> >>> is_lm_address(), but if it is not available, could we survive?
> >>>
> >>> _PAGE_OFFSET(52) < _PAGE_END(52) < _PAGE_OFFSET(48) < _PAGE_END(48)
> >>> Translating into numbers:
> >>> 0xfff0000000000000 < 0xfff8000000000000 < 0xffff000000000000 <
> >>> 0xffff800000000000
> >>>
> >>> Refer to linux/Documentation/arm64/memory.rst, the line
> >>> ffffa00010000000 fffff81ffffeffff ~88TB vmalloc
> >>>
> >>> It comes to the conclusion that any symbol > SYMBOL(_text) > _PAGE_END(48).
> >>>
> >>> So is_lm_address() can looks like
> >>> if (addr > _PAGE_END(48)), it is kimage
> >>> else, it is linear mapping
> >
> > This sounds good and will work.
> >
> >>>
> >>> So even more aggressive, we can exclude the need of vabits_actual
> >>> totally in this patch.
> >
> > Interesting, but if vabits_actual is not available, I wonder how we can
> > get info->page_offset. Seems it's used mainly for p2v conversion and
>
> Ashamed. I will make the best of the fact that va_bits=52 kernel only
> has _PAGE_OFFSET(52). It looks not elegant, but should work.
ok, thanks for your help!
Kazu
_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec
^ permalink raw reply [flat|nested] 10+ messages in thread
end of thread, other threads:[~2021-01-15 3:16 UTC | newest]
Thread overview: 10+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-01-14 8:25 [RFC PATCH 0/4] makedumpfile/arm64: support flipped VA and 52-bit kernel VA kazuhito.hagio
2021-01-14 8:25 ` [RFC PATCH 1/4] Use ELF vmcoreinfo note for --mem-usage option kazuhito.hagio
2021-01-14 8:25 ` [RFC PATCH 2/4] arm64: use NUMBER(VA_BITS) in vmcoreinfo kazuhito.hagio
2021-01-14 8:25 ` [RFC PATCH 3/4] Get kernel version from OSRELEASE " kazuhito.hagio
2021-01-14 8:25 ` [RFC PATCH 4/4] arm64: support flipped VA and 52-bit kernel VA kazuhito.hagio
2021-01-14 10:03 ` piliu
2021-01-14 18:25 ` Bhupesh SHARMA
2021-01-15 0:40 ` HAGIO KAZUHITO(萩尾 一仁)
2021-01-15 1:41 ` piliu
2021-01-15 3:16 ` HAGIO KAZUHITO(萩尾 一仁)
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.