Dne St 11. ledna 2012 15:37:50 Dave Anderson napsal(a):
> ----- Original Message -----
>
> > Dne St 11. ledna 2012 00:37:50 Petr Tesarik napsal(a):
> > > [...]
> > > I can see now that this is unnecessarily complicated, because the
> > > node_remap_* variables are static arrays of MAX_NUMNODES elements, so I
> > > can get their size from the debuginfo at POST_GDB init and initialize a
> > > machine-specific data type with it. I'll post another patch tomorrow.
> >
> > And here we go. Tested on my system and seems to work just fine.
> >
> > Petr Tesarik
> > SUSE Linux
>
> Hi Petr,
>
> This looks pretty good to me. However, just to clarify the
> chicken-and-egg situation here...
>
> When remap_init() does these 3 readmem() calls, they will pass
> through x86_kvtop_remap() -- which I guess would fail because
> the arrays would still be at least partially uninitialized?:
Hi Dave,
first of all, it's not a real issue, because kernel static data is always
below the remapped addresses, but I agree, it's ugly and making non-obvious
assumptions about memory layout. To make things clean, I can simply set
machdep->machspec->max_numnodes as the last thing, so the loop in
x86_kvtop_remap() cannot be executed until everything is initialized.
> + readmem(start_vaddr, KVADDR, ms->remap_start_vaddr,
> + ms->max_numnodes * sizeof(ulong), "node_remap_start_vaddr",
> + FAULT_ON_ERROR);
> + readmem(end_vaddr, KVADDR, ms->remap_end_vaddr,
> + ms->max_numnodes * sizeof(ulong), "node_remap_end_vaddr",
> + FAULT_ON_ERROR);
> + readmem(start_pfn, KVADDR, ms->remap_start_pfn,
> + ms->max_numnodes * sizeof(ulong), "node_remap_end_vaddr",
> + FAULT_ON_ERROR);
>
> To fortify this (and put my mind at ease), can we also set a "remap_state"
> flag in the machdep structure that describes the remap data structures in
> maybe two states:
>
> UNINITIALIZED - set at SETUP_ENV time, then cleared in x86_init() before
> remap_init() is called
Denoted by max_numnodes == 0
> NOT_USED - if remap_init() fails
Denoted by max_numnodes < 0
> Then have x86_kvtop_remap() -- which will get called on every readmem() or
> kvtop() for the rest of time, check for (UNINITIALIZED|NOT_USED), and
> return FALSE immediately if either are set?
>
> What do you think?
Patch attached.
Petr Tesarik
SUSE Linux
From: Petr Tesarik <ptesarik@xxxxxxx>
Subject: [x86] Add correct handling of regions allocated with the remap allocator
References: bnc#738742
Patch-mainline: no
For NUMA x86, the pgdat is remapped into the node's physical memory. Since
that physical memory may not be reachable through the identity mapping, a
small part of the identity mapping is used.
This special case has never been handled properly by crash.
Signed-off-by: Petr Tesarik <ptesarik@xxxxxxx>
---
defs.h | 4 ++
x86.c | 126 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++-----
2 files changed, 122 insertions(+), 8 deletions(-)
--- a/defs.h
+++ b/defs.h
@@ -4253,6 +4253,10 @@ struct machine_specific {
ulonglong last_pmd_read_PAE;
ulonglong last_ptbl_read_PAE;
ulong page_protnone;
+ int max_numnodes;
+ ulong *remap_start_vaddr;
+ ulong *remap_end_vaddr;
+ ulong *remap_start_pfn;
};
struct syment *x86_is_entry_tramp_address(ulong, ulong *);
--- a/x86.c
+++ b/x86.c
@@ -996,6 +996,7 @@ static int x86_uvtop_xen_wpt(struct task
static int x86_kvtop_xen_wpt(struct task_context *, ulong, physaddr_t *, int);
static int x86_uvtop_xen_wpt_PAE(struct task_context *, ulong, physaddr_t *, int);
static int x86_kvtop_xen_wpt_PAE(struct task_context *, ulong, physaddr_t *, int);
+static int x86_kvtop_remap(ulong, physaddr_t *);
static ulong x86_get_task_pgd(ulong);
static ulong x86_processor_speed(void);
static ulong x86_get_pc(struct bt_info *);
@@ -1006,6 +1007,7 @@ static uint64_t x86_memory_size(void);
static ulong x86_vmalloc_start(void);
static ulong *read_idt_table(int);
static void eframe_init(void);
+static int remap_init(void);
#define READ_IDT_INIT 1
#define READ_IDT_RUNTIME 2
static char *extract_idt_function(ulong *, char *, ulong *);
@@ -2014,6 +2016,9 @@ x86_init(int when)
"pr_reg");
STRUCT_SIZE_INIT(percpu_data, "percpu_data");
+ if (!remap_init())
+ machdep->machspec->max_numnodes = -1;
+
break;
case POST_INIT:
@@ -2084,6 +2089,74 @@ eframe_init(void)
}
/*
+ * Locate regions remapped by the remap allocator
+ */
+static int
+remap_init(void)
+{
+ ulong start_vaddr, end_vaddr, start_pfn;
+ int max_numnodes;
+ struct machine_specific *ms;
+ struct syment *sp;
+
+ if (! (sp = symbol_search("node_remap_start_vaddr")) )
+ return FALSE;
+ start_vaddr = sp->value;
+
+ if (! (sp = symbol_search("node_remap_end_vaddr")) )
+ return FALSE;
+ end_vaddr = sp->value;
+
+ if (! (sp = symbol_search("node_remap_start_pfn")) )
+ return FALSE;
+ start_pfn = sp->value;
+
+ max_numnodes = get_array_length("node_remap_start_pfn", NULL,
+ sizeof(ulong));
+ if (max_numnodes < 1)
+ max_numnodes = 1;
+
+ ms = machdep->machspec;
+ ms->remap_start_vaddr = calloc(3 * max_numnodes, sizeof(ulong));
+ if (!ms->remap_start_vaddr)
+ error(FATAL, "cannot malloc remap array");
+ ms->remap_end_vaddr = ms->remap_start_vaddr + max_numnodes;
+ ms->remap_start_pfn = ms->remap_end_vaddr + max_numnodes;
+
+ readmem(start_vaddr, KVADDR, ms->remap_start_vaddr,
+ max_numnodes * sizeof(ulong), "node_remap_start_vaddr",
+ FAULT_ON_ERROR);
+ readmem(end_vaddr, KVADDR, ms->remap_end_vaddr,
+ max_numnodes * sizeof(ulong), "node_remap_end_vaddr",
+ FAULT_ON_ERROR);
+ readmem(start_pfn, KVADDR, ms->remap_start_pfn,
+ max_numnodes * sizeof(ulong), "node_remap_end_vaddr",
+ FAULT_ON_ERROR);
+ ms->max_numnodes = max_numnodes;
+
+ return TRUE;
+}
+
+static int
+x86_kvtop_remap(ulong kvaddr, physaddr_t *paddr)
+{
+ struct machine_specific *ms;
+ int i;
+
+ ms = machdep->machspec;
+
+ for (i = 0; i < ms->max_numnodes; ++i) {
+ if (kvaddr >= ms->remap_start_vaddr[i] &&
+ kvaddr < ms->remap_end_vaddr[i]) {
+ *paddr = PTOB(ms->remap_start_pfn[i]) +
+ kvaddr - ms->remap_start_vaddr[i];
+ return TRUE;
+ }
+ }
+ return FALSE;
+}
+
+/*
* Needs to be done this way because of potential 4G/4G split.
*/
static int
@@ -2768,12 +2841,13 @@ x86_kvtop(struct task_context *tc, ulong
}
pgd = (ulong *)symbol_value("idle_pg_table_l2");
} else {
- if (!vt->vmalloc_start) {
+ if (x86_kvtop_remap(kvaddr, paddr)) {
+ if (!verbose)
+ return TRUE;
+ } else if (!vt->vmalloc_start) {
*paddr = VTOP(kvaddr);
return TRUE;
- }
-
- if (!IS_VMALLOC_ADDR(kvaddr)) {
+ } else if (!IS_VMALLOC_ADDR(kvaddr)) {
*paddr = VTOP(kvaddr);
if (!verbose)
return TRUE;
@@ -3023,12 +3097,13 @@ x86_kvtop_PAE(struct task_context *tc, u
else
pgd = (ulonglong *)symbol_value("idle_pg_table");
} else {
- if (!vt->vmalloc_start) {
+ if (x86_kvtop_remap(kvaddr, paddr)) {
+ if (!verbose)
+ return TRUE;
+ } else if (!vt->vmalloc_start) {
*paddr = VTOP(kvaddr);
return TRUE;
- }
-
- if (!IS_VMALLOC_ADDR(kvaddr)) {
+ } else if (!IS_VMALLOC_ADDR(kvaddr)) {
*paddr = VTOP(kvaddr);
if (!verbose)
return TRUE;
@@ -3354,6 +3429,8 @@ x86_dump_machdep_table(ulong arg)
int others;
ulong xen_wpt;
char buf[BUFSIZE];
+ struct machine_specific *ms;
+ int i, j, max_numnodes;
switch (arg) {
default:
@@ -3489,6 +3566,39 @@ x86_dump_machdep_table(ulong arg)
machdep->machspec->last_ptbl_read_PAE);
fprintf(fp, " page_protnone: %lx\n",
machdep->machspec->page_protnone);
+
+ ms = machdep->machspec;
+ max_numnodes = ms->max_numnodes;
+ fprintf(fp, " MAX_NUMNODES: ");
+ if (max_numnodes < 0) {
+ fprintf(fp, "(unused)\n");
+ } else {
+ fprintf(fp, "%ld\n", max_numnodes);
+
+ fprintf(fp, " remap_start_vaddr:");
+ for (i = 0; i < max_numnodes; ++i) {
+ if ((i % 8) == 0)
+ fprintf(fp, "\n ");
+ fprintf(fp, "%08lx ", ms->remap_start_vaddr[i]);
+ }
+ fprintf(fp, "\n");
+
+ fprintf(fp, " remap_end_vaddr:");
+ for (i = 0; i < max_numnodes; ++i) {
+ if ((i % 8) == 0)
+ fprintf(fp, "\n ");
+ fprintf(fp, "%08lx ", ms->remap_end_vaddr[i]);
+ }
+ fprintf(fp, "\n");
+
+ fprintf(fp, " remap_start_pfn:");
+ for (i = 0; i < max_numnodes; ++i) {
+ if ((i % 8) == 0)
+ fprintf(fp, "\n ");
+ fprintf(fp, "%08lx ", ms->remap_start_pfn[i]);
+ }
+ fprintf(fp, "\n");
+ }
}
/*
--
Crash-utility mailing list
Crash-utility@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/crash-utility
