Re: [PATCHv9 2/2] powerpc/setup: Loosen the mapping between cpu logical id and its seq in dt

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 





On 09/01/24 9:57 am, Hari Bathini wrote:
Hi Michael,


Sorry, Michael.
I am just about getting back to work and I spoke too soon.
You already seem to have posted a set with the approach you had in mind:

  https://patchwork.ozlabs.org/project/linuxppc-dev/list/?series=388350

Thanks
Hari

I am fine with either approach. I was trying to address your concerns
in my way. Looking for your inputs here on how to go about this now..

On 29/11/23 7:00 am, Pingfan Liu wrote:
Hi Hari,


On Mon, Nov 27, 2023 at 12:30 PM Hari Bathini <hbathini@xxxxxxxxxxxxx> wrote:

Hi Pingfan, Michael,

On 17/10/23 4:03 pm, Hari Bathini wrote:


On 17/10/23 7:58 am, Pingfan Liu wrote:
*** Idea ***
For kexec -p, the boot cpu can be not the cpu0, this causes the problem
of allocating memory for paca_ptrs[]. However, in theory, there is no
requirement to assign cpu's logical id as its present sequence in the
device tree. But there is something like cpu_first_thread_sibling(),
which makes assumption on the mapping inside a core. Hence partially
loosening the mapping, i.e. unbind the mapping of core while keep the
mapping inside a core.

*** Implement ***
At this early stage, there are plenty of memory to utilize. Hence, this
patch allocates interim memory to link the cpu info on a list, then
reorder cpus by changing the list head. As a result, there is a rotate
shift between the sequence number in dt and the cpu logical number.

*** Result ***
After this patch, a boot-cpu's logical id will always be mapped into the
range [0,threads_per_core).

Besides this, at this phase, all threads in the boot core are forced to
be onlined. This restriction will be lifted in a later patch with
extra effort.

Signed-off-by: Pingfan Liu <piliu@xxxxxxxxxx>
Cc: Michael Ellerman <mpe@xxxxxxxxxxxxxx>
Cc: Nicholas Piggin <npiggin@xxxxxxxxx>
Cc: Christophe Leroy <christophe.leroy@xxxxxxxxxx>
Cc: Mahesh Salgaonkar <mahesh@xxxxxxxxxxxxx>
Cc: Wen Xiong <wenxiong@xxxxxxxxxx>
Cc: Baoquan He <bhe@xxxxxxxxxx>
Cc: Ming Lei <ming.lei@xxxxxxxxxx>
Cc: Sourabh Jain <sourabhjain@xxxxxxxxxxxxx>
Cc: Hari Bathini <hbathini@xxxxxxxxxxxxx>
Cc: kexec@xxxxxxxxxxxxxxxxxxx
To: linuxppc-dev@xxxxxxxxxxxxxxxx

Thanks for working on this, Pingfan.
Looks good to me.

Acked-by: Hari Bathini <hbathini@xxxxxxxxxxxxx>


On second thoughts, probably better off with no impact for
bootcpu < nr_cpu_ids case and changing only two cores logical
numbering otherwise. Something like the below (Please share
your thoughts):


I am afraid that it may not be as ideal as it looks, considering the
following factors:
-1. For the case of 'bootcpu < nr_cpu_ids', crash can happen evenly
across any cpu in the system, which seriously undermines the
protection intended here (Under the most optimistic scenario, there is
a 50% chance of success)

-2. For the re-ordering of logical numbering, IMHO, if there is
concern that re-ordering will break something, the partial re-ordering
can not avoid that.  We ought to spot probable hazards so as to ease
worries.


Thanks,

Pingfan

diff --git a/arch/powerpc/kernel/prom.c b/arch/powerpc/kernel/prom.c
index ec82f5bda908..78a8312aa8c4 100644
--- a/arch/powerpc/kernel/prom.c
+++ b/arch/powerpc/kernel/prom.c
@@ -76,7 +76,9 @@ u64 ppc64_rma_size;
   unsigned int boot_cpu_node_count __ro_after_init;
   #endif
   static phys_addr_t first_memblock_size;
+#ifdef CONFIG_SMP
   static int __initdata boot_cpu_count;
+#endif

   static int __init early_parse_mem(char *p)
   {
@@ -357,6 +359,25 @@ static int __init early_init_dt_scan_cpus(unsigned
long node,
                         fdt_boot_cpuid_phys(initial_boot_params)) {
                         found = boot_cpu_count;
                         found_thread = i;
+                       /*
+                        * Map boot-cpu logical id into the range
+                        * of [0, thread_per_core) if it can't be
+                        * accommodated within nr_cpu_ids.
+                        */
+                       if (i != boot_cpu_count && boot_cpu_count >= nr_cpu_ids) {
+                               boot_cpuid = i;
+                               DBG("Logical CPU number for boot CPU changed from %d to %d\n",
+                                   boot_cpu_count, i);
+                       } else {
+                               boot_cpuid = boot_cpu_count;
+                       }
+
+                       /* Ensure boot thread is acconted for in nr_cpu_ids */
+                       if (boot_cpuid >= nr_cpu_ids) {
+                               set_nr_cpu_ids(boot_cpuid + 1);
+                               DBG("Adjusted nr_cpu_ids to %u, to include boot CPU.\n",
+                                   nr_cpu_ids);
+                       }
                 }
   #ifdef CONFIG_SMP
                 /* logical cpu id is always 0 on UP kernels */
@@ -368,9 +389,8 @@ static int __init early_init_dt_scan_cpus(unsigned
long node,
         if (found < 0)
                 return 0;

-       DBG("boot cpu: logical %d physical %d\n", found,
+       DBG("boot cpu: logical %d physical %d\n", boot_cpuid,
             be32_to_cpu(intserv[found_thread]));
-       boot_cpuid = found;

         boot_cpu_hwid = be32_to_cpu(intserv[found_thread]);

diff --git a/arch/powerpc/kernel/setup-common.c
b/arch/powerpc/kernel/setup-common.c
index b7b733474b60..f7179525c774 100644
--- a/arch/powerpc/kernel/setup-common.c
+++ b/arch/powerpc/kernel/setup-common.c
@@ -409,6 +409,12 @@ static void __init cpu_init_thread_core_maps(int tpc)

   u32 *cpu_to_phys_id = NULL;

+struct interrupt_server_node {
+       bool    avail;
+       int     len;
+       __be32 intserv[];
+};
+
   /**
    * setup_cpu_maps - initialize the following cpu maps:
    *                  cpu_possible_mask
@@ -429,9 +435,13 @@ u32 *cpu_to_phys_id = NULL;
    */
   void __init smp_setup_cpu_maps(void)
   {
+       struct interrupt_server_node *core0_node = NULL, *bt_node = NULL;
+       int orig_boot_cpu = -1, orig_boot_thread = -1;
+       bool found_boot_cpu = false;
         struct device_node *dn;
-       int cpu = 0;
         int nthreads = 1;
+       int cpu = 0;
+       int j, len;

         DBG("smp_setup_cpu_maps()\n");

@@ -442,9 +452,9 @@ void __init smp_setup_cpu_maps(void)
                       __func__, nr_cpu_ids * sizeof(u32), __alignof__(u32));

         for_each_node_by_type(dn, "cpu") {
+               bool avail, skip = false;
                 const __be32 *intserv;
                 __be32 cpu_be;
-               int j, len;

                 DBG("  * %pOF...\n", dn);

@@ -466,29 +476,121 @@ void __init smp_setup_cpu_maps(void)

                 nthreads = len / sizeof(int);

-               for (j = 0; j < nthreads && cpu < nr_cpu_ids; j++) {
-                       bool avail;
+               avail = of_device_is_available(dn);
+               if (!avail)
+                       avail = !of_property_match_string(dn,
+                                       "enable-method", "spin-table");
+
+               for (j = 0; (cpu == 0 || !found_boot_cpu) && j < nthreads; j++) {
+                       if (be32_to_cpu(intserv[j]) == boot_cpu_hwid) {
+                               found_boot_cpu = true;
+                               if (cpu == 0)
+                                       break;
+
+                               /* Original logical CPU number of thread0 in boot core */
+                               orig_boot_cpu = cpu;
+                               orig_boot_thread = j;
+                               bt_node = memblock_alloc(sizeof(struct interrupt_server_node) + len,
+                                               __alignof__(u32));
+                               if (!bt_node)
+                                       panic("%s: Failed to allocate %zu bytes align=0x%zx\n",
+                                               __func__,
+                                               sizeof(struct interrupt_server_node) + len,
+                                               __alignof__(u32));
+                               bt_node->len = len;
+                               memcpy(bt_node->intserv, intserv, len);
+                               bt_node->avail = avail;
+                               skip = true;
+                               break;
+                       }
+               }

+               /*
+                * Boot CPU not on core0.
+                * Hold off adding core0 until boot core is found as core0 +                * may have to be replaced with boot core if boot core can +                * not be accommodated within nr_cpu_ids with its original
+                * logical CPU numbering.
+                */
+               if (cpu == 0 && !found_boot_cpu) {
+                       core0_node = memblock_alloc(sizeof(struct interrupt_server_node) + len,
+                                       __alignof__(u32));
+                       if (!core0_node)
+                               panic("%s: Failed to allocate %zu bytes align=0x%zx\n",
+                                       __func__,
+                                       sizeof(struct interrupt_server_node) + len,
+                                       __alignof__(u32));
+                       core0_node->len = len;
+                       memcpy(core0_node->intserv, intserv, len);
+                       core0_node->avail = avail;
+                       skip = true;
+               }
+
+               if (skip) {
+                       /* Assumes same number of threads for all cores */
+                       cpu += nthreads;
+                       continue;
+               }
+
+               for (j = 0; j < nthreads && cpu < nr_cpu_ids; j++) {
                         DBG("    thread %d -> cpu %d (hard id %d)\n",
                             j, cpu, be32_to_cpu(intserv[j]));

-                       avail = of_device_is_available(dn);
-                       if (!avail)
-                               avail = !of_property_match_string(dn,
-                                               "enable-method", "spin-table");
-
                         set_cpu_present(cpu, avail);
                         set_cpu_possible(cpu, true);
                         cpu_to_phys_id[cpu] = be32_to_cpu(intserv[j]);
                         cpu++;
                 }

-               if (cpu >= nr_cpu_ids) {
+               if (found_boot_cpu && cpu >= nr_cpu_ids) {
                         of_node_put(dn);
                         break;
                 }
         }

+       /*
+        * Boot CPU not on core0.
+        *
+        * If nr_cpu_ids does not accommodate the original logical CPU
numbering for
+        * boot CPU core, use logical CPU numbers 0 to nthreads for boot CPU core. +        * Note that boot cpu is already assigned with logical CPU number
somewhere
+        * between 0 to nthreads (depending on the boot thread within the core) in
+        * early_init_dt_scan_cpus() for this case.
+        *
+        * Otherwise, stick with the original logical CPU numbering.
+        */
+       if (bt_node) {
+               int core0_cpu;
+
+               if (orig_boot_cpu + orig_boot_thread >= nr_cpu_ids) {
+                       cpu = 0;
+                       core0_cpu = orig_boot_cpu;
+               } else {
+                       cpu = orig_boot_cpu;
+                       core0_cpu = 0;
+               }
+
+               for (j = 0; j < nthreads && core0_cpu < nr_cpu_ids; j++) {
+                       DBG("    thread %d -> cpu %d (hard id %d)\n",
+                           j, core0_cpu, be32_to_cpu(core0_node->intserv[j]));
+
+                       set_cpu_present(core0_cpu, core0_node->avail);
+                       set_cpu_possible(core0_cpu, true);
+                       cpu_to_phys_id[core0_cpu] = be32_to_cpu(core0_node->intserv[j]);
+                       core0_cpu++;
+               }
+
+               for (j = 0; j < nthreads && cpu < nr_cpu_ids; j++) {
+                       DBG("    thread %d -> cpu %d (hard id %d)\n",
+                           j, cpu, be32_to_cpu(bt_node->intserv[j]));
+
+                       set_cpu_present(cpu, bt_node->avail);
+                       set_cpu_possible(cpu, true);
+                       cpu_to_phys_id[cpu] = be32_to_cpu(bt_node->intserv[j]);
+                       cpu++;
+               }
+       }
+
         /* If no SMT supported, nthreads is forced to 1 */
         if (!cpu_has_feature(CPU_FTR_SMT)) {
                 DBG("  SMT disabled ! nthreads forced to 1\n");



Thanks
Hari

_______________________________________________
kexec mailing list
kexec@xxxxxxxxxxxxxxxxxxx
http://lists.infradead.org/mailman/listinfo/kexec




[Index of Archives]     [LM Sensors]     [Linux Sound]     [ALSA Users]     [ALSA Devel]     [Linux Audio Users]     [Linux Media]     [Kernel]     [Gimp]     [Yosemite News]     [Linux Media]

  Powered by Linux