<!doctype html public "-//w3c//dtd html 4.0 transitional//en"> <html> <tt></tt> <tt>Sharyathi Nagesh wrote:</tt><tt></tt> <tt> Dave</tt> <tt> You were right it should be buffer_size instead of objectsize.(As obj_size is the size of debugging information).</tt> <tt> I modified function vm_init() and do_slab_chain_percpu_v2() to make it work with current dumps. This is just a code</tt> <tt> part to show how to fix the problem but it is not a patch as such...</tt> <tt> Please go through and let me know would it be good idea to write a new function (as you have suggested earlier) or</tt> <tt> is it possible to accommodate by calling do_slab_chain_percpu_v2() iteratively with different nodes by passing them as</tt> <tt> parameter.</tt><tt></tt> <tt> Changes to vm_init()</tt><tt></tt> <tt> if(MEMBER_EXISTS("kmem_cache","objsize"))</tt> <tt> MEMBER_OFFSET_INIT(kmem_cache_s_objsize, "kmem_cache",</tt> <tt> "objsize");</tt> <tt> else if(MEMBER_EXISTS("kmem_cache","buffer_size"))</tt> <tt> MEMBER_OFFSET_INIT(kmem_cache_s_objsize, "kmem_cache",</tt> <tt> "buffer_size")</tt><tt></tt> <tt> Changes to do_slab_chain_percpu_v2()</tt><tt></tt> <tt> ulong start_address[MAX_NUMNODES];</tt><tt></tt> <tt> if (!readmem(si->cache+OFFSET(kmem_cache_s_lists),</tt> <tt> KVADDR, &start_address[0],</tt> <tt> sizeof(ulong) * MAX_NUMNODES,</tt> <tt> "array nodelist array", RETURN_ON_ERROR))</tt> <tt> {printf("\n Error encountered with readmem \n");exit(0);}</tt><tt></tt> <tt> for (i = 0; i < MAX_NUMNODES ; i++) {</tt> <tt> slab_chains[0] = start_address[i] + OFFSET(kmem_list3_slabs_partial);</tt> <tt> slab_chains[1] = start_address[i] + OFFSET(kmem_list3_slabs_full);</tt> <tt> slab_chains[2] = start_address[i] + OFFSET(kmem_list3_slabs_free);</tt> <tt> </tt><tt></tt> <tt> </tt><tt></tt> <tt>The vm_init() change is fine, but it would also need:</tt><tt></tt> <tt>(1) the same kind of offset initializer to handle the kmem_cache</tt> <tt> structure's "lists"-to-"nodelist" name change.</tt> <tt>(2) A new flag in vt->flags to signal this node-enhanced version</tt> <tt> of the kmem_cache struct.</tt> <tt>(3) a check of the dimension of some processor-neutral, always-there,</tt> <tt> kernel array indexed by MAX_NUMNODES, so that you can determine</tt> <tt> what MAX_NUMNODES is equal to, and then store that value in a</tt> <tt> new vm_table entry.</tt><tt></tt> <tt>With respect to do_slab_chain_percpu_v2(), I would prefer that</tt> <tt>it be left as is, and either</tt><tt></tt> <tt>(1) at the top of the function, check for your new "node-enhanced"</tt> <tt> flag, and call a new do_slab_chain_percpu_v2_nodes() function, or</tt> <tt>(2) alternatively, the two locations where do_slab_chain_percpu_v2()</tt> <tt> gets called could be qualified based upon the new flag, and the</tt> <tt> appropriate function called.</tt><tt></tt> <tt>In that new function, the usage of a static start_address[] array</tt> <tt>as you indicate above would have to be replaced with dynamically</tt> <tt>allocated and freed with GETBUF() and FREEBUF(), based upon</tt> <tt>the predetermined size of MAX_NUMNODES.</tt><tt></tt> <tt>Doing it that way leaves the current functionality untouched, makes</tt> <tt>me a lot less nervous, and should make your task a bit easier.</tt><tt></tt> <tt>Make sense?</tt><tt></tt> <tt>Dave</tt></html>