crash-utility-bounces(a)redhat.com wrote on 11/16/2005 11:28:23 AM:
Badari Pulavarty wrote:
Hi Dave,
I am playing with "crash" on my machine with it fails with
CONFIG_SPARSE_MEM. It looks like "node_mem_map" doesn't
exist for SPARSE_MEM :(
crash: invalid structure member offset: pglist_data_node_mem_map
FILE: memory.c LINE: 10053 FUNCTION: dump_memory_nodes()
[./crash] error trace: 100afe98 => 100d12f4 => 100d01cc => 10142f64
10142f64: .OFFSET_verify+140
100d01cc: .dump_memory_nodes+520
100d12f4: .node_table_init+492
100afe98: .vm_init+7948
Thanks,
Badari
typedef struct pglist_data {
struct zone node_zones[MAX_NR_ZONES];
struct zonelist node_zonelists[GFP_ZONETYPES];
int nr_zones;
#ifdef CONFIG_FLAT_NODE_MEM_MAP
struct page *node_mem_map;
#endif
struct bootmem_data *bdata;
Hmmm,, it kind of looks like the global "mem_map" array should be
used in that case, but if that were true, then there could never
be multiple pglist_data structures in that case?
Here's what I mean -- in vm_init() there's this:
if (VALID_STRUCT(pglist_data)) {
vt->flags |= ZONES;
if (symbol_exists("pgdat_list"))
vt->flags |= NODES;
And later on in dump_memory_nodes(), which gathers the memory-related
information for each node, it looks at that "NODES" flag.
- If the NODES flag is *not* set, then it presumes that there is
just one node, and uses the global "mem_map".
- If the NODES flag is set, then starting at the pgdat_list list head,
it walks though each of the pglist_data structures (one per node),
and then collects all of the individual node_mem_map pointers
found in each node.
In this hybrid case, I don't know how exactly the mem_map
is found for each pglist_data node? You could quickly try
not setting the vt->flags NODES bit, and seeing what happens.
But I get the feeling that there's going to more to be done
than that -- presuming that you can have more than one pglist_data
node. In other words if the pglist_data list can have more
than one node when CONFIG_FLAT_NODE_MEM_MAP is turned off,
then we need to find out how to make the relationship between
each sparse node and its associated mem_map array.
Dave
As we know, recently, several config options (SPARSEMEM,
FLAT_NODE_MEM_MAP, DISCONTIG ..) got introduced. Badari, is
CONFIG_SPARSEMEM is enabled on your system (System.map should have the
symbol called mem_section). I think, we should also support
mm_section.section_mem_map.
Thanks
Haren
--
Crash-utility mailing list
Crash-utility(a)redhat.com
https://www.redhat.com/mailman/listinfo/crash-utility