Re: [PATCH 1/5] x86/smpboot: Add the missing description of possible_cpus
At 03/21/2018 05:34 PM, Dou Liyang wrote: Hi Peter, At 03/21/2018 05:08 PM, Peter Zijlstra wrote: On Wed, Mar 21, 2018 at 01:33:24PM +0800, Dou Liyang wrote: How about: possible_cpus= [s390,x86_64] Set the number of possible CPUs which are determined by the ACPI tables MADT or mptables by default. possible_cpus=n : n >= 1 enforces the possible number to be 'n'. While nr_cpus is also be set: nr_cpus=m, choice the minimum one for the number of possible CPUs. So what is the exact difference between possible_cpus and nr_cpus ? I the possible_cpus= can reset the number of possible CPUs, even bigger than 'num_processors + disabled_cpus', But nr_cpus= can't. ^^^ the maximum number kernel gets from ACPI/mptables, no matter what number nr_cpus= is, the number of possible CPUs will not bigger than it. konw maxcpus= limits the number of CPUs we bring up, and possible_cpus limits the possible_map, but I'm not entirely sure what nr_cpus does here. nr_cpus can limited the maximum CPUs that the kernel could support. Here is a double check in case of using them at the same time, even if I think just using possible_cpus= is enough. :-) Thanks, dou.
Re: [PATCH 1/5] x86/smpboot: Add the missing description of possible_cpus
At 03/21/2018 05:34 PM, Dou Liyang wrote: Hi Peter, At 03/21/2018 05:08 PM, Peter Zijlstra wrote: On Wed, Mar 21, 2018 at 01:33:24PM +0800, Dou Liyang wrote: How about: possible_cpus= [s390,x86_64] Set the number of possible CPUs which are determined by the ACPI tables MADT or mptables by default. possible_cpus=n : n >= 1 enforces the possible number to be 'n'. While nr_cpus is also be set: nr_cpus=m, choice the minimum one for the number of possible CPUs. So what is the exact difference between possible_cpus and nr_cpus ? I the possible_cpus= can reset the number of possible CPUs, even bigger than 'num_processors + disabled_cpus', But nr_cpus= can't. ^^^ the maximum number kernel gets from ACPI/mptables, no matter what number nr_cpus= is, the number of possible CPUs will not bigger than it. konw maxcpus= limits the number of CPUs we bring up, and possible_cpus limits the possible_map, but I'm not entirely sure what nr_cpus does here. nr_cpus can limited the maximum CPUs that the kernel could support. Here is a double check in case of using them at the same time, even if I think just using possible_cpus= is enough. :-) Thanks, dou.
Re: [PATCH 1/5] x86/smpboot: Add the missing description of possible_cpus
On Wed, 21 Mar 2018, Peter Zijlstra wrote: > On Wed, Mar 21, 2018 at 01:33:24PM +0800, Dou Liyang wrote: > > How about: > > > > possible_cpus= [s390,x86_64] Set the number of possible CPUs which > > are determined by the ACPI tables MADT or mptables by > > default. possible_cpus=n : n >= 1 enforces the possible > > number to be 'n'. > > While nr_cpus is also be set: nr_cpus=m, choice the > > minimum one for the number of possible CPUs. > > So what is the exact difference between possible_cpus and nr_cpus ? I > konw maxcpus= limits the number of CPUs we bring up, and possible_cpus > limits the possible_map, but I'm not entirely sure what nr_cpus does > here. nr_cpus limits the number of CPUs the kernel will handle. Think of it as a boot time override of NR_CPUs. Way too many commandline switches though. Thanks, tglx
Re: [PATCH 1/5] x86/smpboot: Add the missing description of possible_cpus
On Wed, 21 Mar 2018, Peter Zijlstra wrote: > On Wed, Mar 21, 2018 at 01:33:24PM +0800, Dou Liyang wrote: > > How about: > > > > possible_cpus= [s390,x86_64] Set the number of possible CPUs which > > are determined by the ACPI tables MADT or mptables by > > default. possible_cpus=n : n >= 1 enforces the possible > > number to be 'n'. > > While nr_cpus is also be set: nr_cpus=m, choice the > > minimum one for the number of possible CPUs. > > So what is the exact difference between possible_cpus and nr_cpus ? I > konw maxcpus= limits the number of CPUs we bring up, and possible_cpus > limits the possible_map, but I'm not entirely sure what nr_cpus does > here. nr_cpus limits the number of CPUs the kernel will handle. Think of it as a boot time override of NR_CPUs. Way too many commandline switches though. Thanks, tglx
Re: [PATCH 1/5] x86/smpboot: Add the missing description of possible_cpus
Hi Peter, At 03/21/2018 05:08 PM, Peter Zijlstra wrote: On Wed, Mar 21, 2018 at 01:33:24PM +0800, Dou Liyang wrote: How about: possible_cpus= [s390,x86_64] Set the number of possible CPUs which are determined by the ACPI tables MADT or mptables by default. possible_cpus=n : n >= 1 enforces the possible number to be 'n'. While nr_cpus is also be set: nr_cpus=m, choice the minimum one for the number of possible CPUs. So what is the exact difference between possible_cpus and nr_cpus ? I the possible_cpus= can reset the number of possible CPUs, even bigger than 'num_processors + disabled_cpus', But nr_cpus= can't. konw maxcpus= limits the number of CPUs we bring up, and possible_cpus limits the possible_map, but I'm not entirely sure what nr_cpus does here. nr_cpus can limited the maximum CPUs that the kernel could support. Here is a double check in case of using them at the same time, even if I think just using possible_cpus= is enough. :-) Thanks, dou.
Re: [PATCH 1/5] x86/smpboot: Add the missing description of possible_cpus
Hi Peter, At 03/21/2018 05:08 PM, Peter Zijlstra wrote: On Wed, Mar 21, 2018 at 01:33:24PM +0800, Dou Liyang wrote: How about: possible_cpus= [s390,x86_64] Set the number of possible CPUs which are determined by the ACPI tables MADT or mptables by default. possible_cpus=n : n >= 1 enforces the possible number to be 'n'. While nr_cpus is also be set: nr_cpus=m, choice the minimum one for the number of possible CPUs. So what is the exact difference between possible_cpus and nr_cpus ? I the possible_cpus= can reset the number of possible CPUs, even bigger than 'num_processors + disabled_cpus', But nr_cpus= can't. konw maxcpus= limits the number of CPUs we bring up, and possible_cpus limits the possible_map, but I'm not entirely sure what nr_cpus does here. nr_cpus can limited the maximum CPUs that the kernel could support. Here is a double check in case of using them at the same time, even if I think just using possible_cpus= is enough. :-) Thanks, dou.
Re: [PATCH 1/5] x86/smpboot: Add the missing description of possible_cpus
On Wed, Mar 21, 2018 at 01:33:24PM +0800, Dou Liyang wrote: > How about: > > possible_cpus=[s390,x86_64] Set the number of possible CPUs which > are determined by the ACPI tables MADT or mptables by > default. possible_cpus=n : n >= 1 enforces the possible > number to be 'n'. > While nr_cpus is also be set: nr_cpus=m, choice the > minimum one for the number of possible CPUs. So what is the exact difference between possible_cpus and nr_cpus ? I konw maxcpus= limits the number of CPUs we bring up, and possible_cpus limits the possible_map, but I'm not entirely sure what nr_cpus does here.
Re: [PATCH 1/5] x86/smpboot: Add the missing description of possible_cpus
On Wed, Mar 21, 2018 at 01:33:24PM +0800, Dou Liyang wrote: > How about: > > possible_cpus=[s390,x86_64] Set the number of possible CPUs which > are determined by the ACPI tables MADT or mptables by > default. possible_cpus=n : n >= 1 enforces the possible > number to be 'n'. > While nr_cpus is also be set: nr_cpus=m, choice the > minimum one for the number of possible CPUs. So what is the exact difference between possible_cpus and nr_cpus ? I konw maxcpus= limits the number of CPUs we bring up, and possible_cpus limits the possible_map, but I'm not entirely sure what nr_cpus does here.
Re: [PATCH 1/5] x86/smpboot: Add the missing description of possible_cpus
Hi Peter, At 03/20/2018 08:37 PM, Peter Zijlstra wrote: On Tue, Mar 20, 2018 at 07:04:28PM +0800, Dou Liyang wrote: + possible_cpus= [s390,x86_64] Use this to set hotpluggable cpus. + This option sets possible_cpus bits in cpu_possible_map. + Thus keeping the numbers of bits set constant even if + the machine gets rebooted. That description, esp. the last sentence, doesn't make any kind of sense to me. What? Ah, sure enough, I can't be lazy. :-) I stole that from the commit 3b11ce7f542e ("x86: use possible_cpus=NUM to extend the possible cpus allowed") How about: possible_cpus= [s390,x86_64] Set the number of possible CPUs which are determined by the ACPI tables MADT or mptables by default. possible_cpus=n : n >= 1 enforces the possible number to be 'n'. While nr_cpus is also be set: nr_cpus=m, choice the minimum one for the number of possible CPUs. Thank, dou
Re: [PATCH 1/5] x86/smpboot: Add the missing description of possible_cpus
Hi Peter, At 03/20/2018 08:37 PM, Peter Zijlstra wrote: On Tue, Mar 20, 2018 at 07:04:28PM +0800, Dou Liyang wrote: + possible_cpus= [s390,x86_64] Use this to set hotpluggable cpus. + This option sets possible_cpus bits in cpu_possible_map. + Thus keeping the numbers of bits set constant even if + the machine gets rebooted. That description, esp. the last sentence, doesn't make any kind of sense to me. What? Ah, sure enough, I can't be lazy. :-) I stole that from the commit 3b11ce7f542e ("x86: use possible_cpus=NUM to extend the possible cpus allowed") How about: possible_cpus= [s390,x86_64] Set the number of possible CPUs which are determined by the ACPI tables MADT or mptables by default. possible_cpus=n : n >= 1 enforces the possible number to be 'n'. While nr_cpus is also be set: nr_cpus=m, choice the minimum one for the number of possible CPUs. Thank, dou
Re: [PATCH 1/5] x86/smpboot: Add the missing description of possible_cpus
On Tue, Mar 20, 2018 at 07:04:28PM +0800, Dou Liyang wrote: > + possible_cpus= [s390,x86_64] Use this to set hotpluggable cpus. > + This option sets possible_cpus bits in cpu_possible_map. > + Thus keeping the numbers of bits set constant even if > + the machine gets rebooted. That description, esp. the last sentence, doesn't make any kind of sense to me. What?
Re: [PATCH 1/5] x86/smpboot: Add the missing description of possible_cpus
On Tue, Mar 20, 2018 at 07:04:28PM +0800, Dou Liyang wrote: > + possible_cpus= [s390,x86_64] Use this to set hotpluggable cpus. > + This option sets possible_cpus bits in cpu_possible_map. > + Thus keeping the numbers of bits set constant even if > + the machine gets rebooted. That description, esp. the last sentence, doesn't make any kind of sense to me. What?