Re: [Qemu-ppc] [Qemu-devel] [PATCH] ppc: Huge page detection mechanism fixes - Episode III

[Greg Kurz]

On Tue, 19 Jul 2016 08:23:46 +0200
Thomas Huth <address@hidden> wrote:

> On 18.07.2016 17:18, Greg Kurz wrote:
> > On Mon, 18 Jul 2016 15:19:04 +0200
> > Thomas Huth <address@hidden> wrote:
> >   
> >> After already fixing two issues with the huge page detection mechanism
> >> (see commit 159d2e39a860 and 86b50f2e1bef), Greg Kurz noticed another
> >> case that caused the guest to crash where QEMU announces huge pages
> >> though they should not be available for the guest:
> >>
> >> qemu-system-ppc64 -enable-kvm ... -mem-path /dev/hugepages \
> >>  -m 1G,slots=4,maxmem=32G
> >>  -object memory-backend-ram,policy=default,size=1G,id=mem-mem1 \
> >>  -device pc-dimm,id=dimm-mem1,memdev=mem-mem1 -smp 2 \
> >>  -numa node,nodeid=0 -numa node,nodeid=1
> >>
> >> That means if there is a global mem-path option, we still have
> >> to look at the memory-backend objects that have been specified
> >> additionally and return their minimum page size if that value
> >> is smaller than the page size of the main memory.
> >>
> >> Reported-by: Greg Kurz <address@hidden>
> >> Signed-off-by: Thomas Huth <address@hidden>
> >> ---  
> > 
> > Just one remark, see below, but apart from that:
> > 
> > Reviewed-by: Greg Kurz <address@hidden>
> > Tested-by: Greg Kurz <address@hidden>
> >   
> >>  target-ppc/kvm.c | 27 ++++++++++++++-------------
> >>  1 file changed, 14 insertions(+), 13 deletions(-)
> >>
> >> diff --git a/target-ppc/kvm.c b/target-ppc/kvm.c
> >> index 7a8f555..97ab450 100644
> >> --- a/target-ppc/kvm.c
> >> +++ b/target-ppc/kvm.c
> >> @@ -366,10 +366,13 @@ static int find_max_supported_pagesize(Object *obj, 
> >> void *opaque)
> >>  static long getrampagesize(void)
> >>  {
> >>      long hpsize = LONG_MAX;
> >> +    long mainrampagesize;
> >>      Object *memdev_root;
> >>  
> >>      if (mem_path) {
> >> -        return gethugepagesize(mem_path);
> >> +        mainrampagesize = gethugepagesize(mem_path);
> >> +    } else {
> >> +        mainrampagesize = getpagesize();
> >>      }
> >>  
> >>      /* it's possible we have memory-backend objects with
> >> @@ -383,28 +386,26 @@ static long getrampagesize(void)
> >>       * backend isn't backed by hugepages.
> >>       */
> >>      memdev_root = object_resolve_path("/objects", NULL);
> >> -    if (!memdev_root) {
> >> -        return getpagesize();
> >> +    if (memdev_root) {
> >> +        object_child_foreach(memdev_root, find_max_supported_pagesize, 
> >> &hpsize);
> >>      }
> >> -
> >> -    object_child_foreach(memdev_root, find_max_supported_pagesize, 
> >> &hpsize);
> >> -
> >> -    if (hpsize == LONG_MAX || hpsize == getpagesize()) {
> >> -        return getpagesize();
> >> +    if (hpsize == LONG_MAX) {
> >> +        /* No additional memory regions found ==> Report main RAM page 
> >> size */
> >> +        return mainrampagesize;
> >>      }
> >>  
> >>      /* If NUMA is disabled or the NUMA nodes are not backed with a
> >> -     * memory-backend, then there is at least one node using "normal"
> >> -     * RAM. And since normal RAM has not been configured with "-mem-path"
> >> -     * (what we've checked earlier here already), we can not use huge 
> >> pages!
> >> +     * memory-backend, then there is at least one node using "normal" RAM,
> >> +     * so if its page size is smaller we have got to report that size 
> >> instead.
> >>       */
> >> -    if (nb_numa_nodes == 0 || numa_info[0].node_memdev == NULL) {
> >> +    if (hpsize > mainrampagesize &&
> >> +        (nb_numa_nodes == 0 || numa_info[0].node_memdev == NULL)) {
> >>          static bool warned;
> >>          if (!warned) {
> >>              error_report("Huge page support disabled (n/a for main 
> >> memory).");  
> > 
> > Maybe update the error message since we have another condition ?
> > 
> > Something like:
> > 
> > "Huge page support disabled (at least one numa uses standard page size)"  
> 
> That sounds also a little bit confusing since the error message could
> occur when there is no numa configured at all. I think refering to "main
> memory" is better here so that the users have a chance to know that they
> might need to specify the global "-mem-path" parameter here, too.
> 
>  Thomas
> 

Fair enough. And anyway, David has already applied the patch.

Cheers.

--
Greg