* [PATCH] x86-64: fix combining of regions in init_memory_mapping() @ 2008-09-12 14:43 Jan Beulich 2008-09-14 18:20 ` Yinghai Lu 0 siblings, 1 reply; 5+ messages in thread From: Jan Beulich @ 2008-09-12 14:43 UTC (permalink / raw) To: mingo, tglx, hpa; +Cc: linux-kernel When nr_range gets decremented, the same slot must be considered for coalescing with its new successor again. Signed-off-by: Jan Beulich <jbeulich@novell.com> --- arch/x86/mm/init_64.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) --- linux-2.6.27-rc6/arch/x86/mm/init_64.c 2008-08-29 10:53:00.000000000 +0200 +++ 2.6.27-rc6-x86_64-mr-coalesce/arch/x86/mm/init_64.c 2008-09-12 11:58:45.000000000 +0200 @@ -636,7 +636,7 @@ unsigned long __init_refok init_memory_m old_start = mr[i].start; memmove(&mr[i], &mr[i+1], (nr_range - 1 - i) * sizeof (struct map_range)); - mr[i].start = old_start; + mr[i--].start = old_start; nr_range--; } ^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] x86-64: fix combining of regions in init_memory_mapping() 2008-09-12 14:43 [PATCH] x86-64: fix combining of regions in init_memory_mapping() Jan Beulich @ 2008-09-14 18:20 ` Yinghai Lu 2008-09-15 8:17 ` Jan Beulich 0 siblings, 1 reply; 5+ messages in thread From: Yinghai Lu @ 2008-09-14 18:20 UTC (permalink / raw) To: Jan Beulich; +Cc: mingo, tglx, hpa, linux-kernel On Fri, Sep 12, 2008 at 7:43 AM, Jan Beulich <jbeulich@novell.com> wrote: > When nr_range gets decremented, the same slot must be considered for > coalescing with its new successor again. > > Signed-off-by: Jan Beulich <jbeulich@novell.com> > > --- > arch/x86/mm/init_64.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > --- linux-2.6.27-rc6/arch/x86/mm/init_64.c 2008-08-29 10:53:00.000000000 +0200 > +++ 2.6.27-rc6-x86_64-mr-coalesce/arch/x86/mm/init_64.c 2008-09-12 11:58:45.000000000 +0200 > @@ -636,7 +636,7 @@ unsigned long __init_refok init_memory_m > old_start = mr[i].start; > memmove(&mr[i], &mr[i+1], > (nr_range - 1 - i) * sizeof (struct map_range)); > - mr[i].start = old_start; > + mr[i--].start = old_start; > nr_range--; > } > this patch seems not right. Ingo, please don't apply it. original code: /* try to merge same page size and continuous */ for (i = 0; nr_range > 1 && i < nr_range - 1; i++) { unsigned long old_start; if (mr[i].end != mr[i+1].start || mr[i].page_size_mask != mr[i+1].page_size_mask) continue; /* move it */ old_start = mr[i].start; memmove(&mr[i], &mr[i+1], (nr_range - 1 - i) * sizeof (struct map_range)); mr[i].start = old_start; nr_range--; } so it save old_start and first, and move entries forward (so old one is overwriten), and put back old_start ... YH ^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] x86-64: fix combining of regions in init_memory_mapping() 2008-09-14 18:20 ` Yinghai Lu @ 2008-09-15 8:17 ` Jan Beulich 2008-09-15 8:34 ` Yinghai Lu 0 siblings, 1 reply; 5+ messages in thread From: Jan Beulich @ 2008-09-15 8:17 UTC (permalink / raw) To: Yinghai Lu; +Cc: mingo, tglx, linux-kernel, hpa >>> "Yinghai Lu" <yhlu.kernel@gmail.com> 14.09.08 20:20 >>> >On Fri, Sep 12, 2008 at 7:43 AM, Jan Beulich <jbeulich@novell.com> wrote: >> When nr_range gets decremented, the same slot must be considered for >> coalescing with its new successor again. >> >> Signed-off-by: Jan Beulich <jbeulich@novell.com> >> >> --- >> arch/x86/mm/init_64.c | 2 +- >> 1 file changed, 1 insertion(+), 1 deletion(-) >> >> --- linux-2.6.27-rc6/arch/x86/mm/init_64.c 2008-08-29 10:53:00.000000000 +0200 >> +++ 2.6.27-rc6-x86_64-mr-coalesce/arch/x86/mm/init_64.c 2008-09-12 11:58:45.000000000 +0200 >> @@ -636,7 +636,7 @@ unsigned long __init_refok init_memory_m >> old_start = mr[i].start; >> memmove(&mr[i], &mr[i+1], >> (nr_range - 1 - i) * sizeof (struct map_range)); >> - mr[i].start = old_start; >> + mr[i--].start = old_start; >> nr_range--; >> } >> > >this patch seems not right. >Ingo, please don't apply it. > >original code: > /* try to merge same page size and continuous */ > for (i = 0; nr_range > 1 && i < nr_range - 1; i++) { > unsigned long old_start; > if (mr[i].end != mr[i+1].start || > mr[i].page_size_mask != mr[i+1].page_size_mask) > continue; > /* move it */ > old_start = mr[i].start; > memmove(&mr[i], &mr[i+1], > (nr_range - 1 - i) * sizeof (struct map_range)); > mr[i].start = old_start; > nr_range--; > } > >so it save old_start and first, and move entries forward (so old one >is overwriten), and put back old_start ... Old and new code are not different in any way in this respect - both overwrite the old entry at index i with the entry at index i+1 and then set the start of the i-th entry back to what it was before the overwrite, effectively combining them. The patch just makes sure that the index isn't being updated at the same time as nr_range (because if you update both you effectively skip one). The issue is apparently pretty benign to native code, but surfaces as a boot time crash in our forward ported Xen tree (where the page table setup overall works differently than in native). Since the underlying issue was present in native (and since I assume if there is an attempt to merge subsequent regions, then it should work right), I nevertheless submitted the patch for native inclusion. Jan ^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] x86-64: fix combining of regions in init_memory_mapping() 2008-09-15 8:17 ` Jan Beulich @ 2008-09-15 8:34 ` Yinghai Lu 2008-09-15 12:51 ` Ingo Molnar 0 siblings, 1 reply; 5+ messages in thread From: Yinghai Lu @ 2008-09-15 8:34 UTC (permalink / raw) To: Jan Beulich; +Cc: mingo, tglx, linux-kernel, hpa On Mon, Sep 15, 2008 at 1:17 AM, Jan Beulich <jbeulich@novell.com> wrote: >>>> "Yinghai Lu" <yhlu.kernel@gmail.com> 14.09.08 20:20 >>> >>On Fri, Sep 12, 2008 at 7:43 AM, Jan Beulich <jbeulich@novell.com> wrote: >>> When nr_range gets decremented, the same slot must be considered for >>> coalescing with its new successor again. >>> >>> Signed-off-by: Jan Beulich <jbeulich@novell.com> >>> >>> --- >>> arch/x86/mm/init_64.c | 2 +- >>> 1 file changed, 1 insertion(+), 1 deletion(-) >>> >>> --- linux-2.6.27-rc6/arch/x86/mm/init_64.c 2008-08-29 10:53:00.000000000 +0200 >>> +++ 2.6.27-rc6-x86_64-mr-coalesce/arch/x86/mm/init_64.c 2008-09-12 11:58:45.000000000 +0200 >>> @@ -636,7 +636,7 @@ unsigned long __init_refok init_memory_m >>> old_start = mr[i].start; >>> memmove(&mr[i], &mr[i+1], >>> (nr_range - 1 - i) * sizeof (struct map_range)); >>> - mr[i].start = old_start; >>> + mr[i--].start = old_start; >>> nr_range--; >>> } >>> >> >>this patch seems not right. >>Ingo, please don't apply it. >> >>original code: >> /* try to merge same page size and continuous */ >> for (i = 0; nr_range > 1 && i < nr_range - 1; i++) { >> unsigned long old_start; >> if (mr[i].end != mr[i+1].start || >> mr[i].page_size_mask != mr[i+1].page_size_mask) >> continue; >> /* move it */ >> old_start = mr[i].start; >> memmove(&mr[i], &mr[i+1], >> (nr_range - 1 - i) * sizeof (struct map_range)); >> mr[i].start = old_start; >> nr_range--; >> } >> >>so it save old_start and first, and move entries forward (so old one >>is overwriten), and put back old_start ... > > Old and new code are not different in any way in this respect - both > overwrite the old entry at index i with the entry at index i+1 and then > set the start of the i-th entry back to what it was before the overwrite, > effectively combining them. The patch just makes sure that the index > isn't being updated at the same time as nr_range (because if you update > both you effectively skip one). > > The issue is apparently pretty benign to native code, but surfaces as a > boot time crash in our forward ported Xen tree (where the page table > setup overall works differently than in native). Since the underlying > issue was present in native (and since I assume if there is an attempt > to merge subsequent regions, then it should work right), I nevertheless > submitted the patch for native inclusion. yes. your patch fixed the skip... Acked-by: Yinghai Lu <yhlu.kernel@gmail.com> YH ^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] x86-64: fix combining of regions in init_memory_mapping() 2008-09-15 8:34 ` Yinghai Lu @ 2008-09-15 12:51 ` Ingo Molnar 0 siblings, 0 replies; 5+ messages in thread From: Ingo Molnar @ 2008-09-15 12:51 UTC (permalink / raw) To: Yinghai Lu; +Cc: Jan Beulich, tglx, linux-kernel, hpa * Yinghai Lu <yhlu.kernel@gmail.com> wrote: > > The issue is apparently pretty benign to native code, but surfaces > > as a boot time crash in our forward ported Xen tree (where the page > > table setup overall works differently than in native). Since the > > underlying issue was present in native (and since I assume if there > > is an attempt to merge subsequent regions, then it should work > > right), I nevertheless submitted the patch for native inclusion. > > yes. your patch fixed the skip... > > Acked-by: Yinghai Lu <yhlu.kernel@gmail.com> applied to tip/x86/core, thanks! Ingo ^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2008-09-15 12:52 UTC | newest] Thread overview: 5+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2008-09-12 14:43 [PATCH] x86-64: fix combining of regions in init_memory_mapping() Jan Beulich 2008-09-14 18:20 ` Yinghai Lu 2008-09-15 8:17 ` Jan Beulich 2008-09-15 8:34 ` Yinghai Lu 2008-09-15 12:51 ` Ingo Molnar
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox