- Nov 15, 2005
-
-
Andi Kleen authored
Add a new 4GB GFP_DMA32 zone between the GFP_DMA and GFP_NORMAL zones. As a bit of historical background: when the x86-64 port was originally designed we had some discussion if we should use a 16MB DMA zone like i386 or a 4GB DMA zone like IA64 or both. Both was ruled out at this point because it was in early 2.4 when VM is still quite shakey and had bad troubles even dealing with one DMA zone. We settled on the 16MB DMA zone mainly because we worried about older soundcards and the floppy. But this has always caused problems since then because device drivers had trouble getting enough DMA able memory. These days the VM works much better and the wide use of NUMA has proven it can deal with many zones successfully. So this patch adds both zones. This helps drivers who need a lot of memory below 4GB because their hardware is not accessing more (graphic drivers - proprietary and free ones, video frame buffer drivers, sound drivers etc.). Previously they could only use IOMMU+16MB GFP_DMA, which was not enough memory. Another common problem is that hardware who has full memory addressing for >4GB misses it for some control structures in memory (like transmit rings or other metadata). They tended to allocate memory in the 16MB GFP_DMA or the IOMMU/swiotlb then using pci_alloc_consistent, but that can tie up a lot of precious 16MB GFPDMA/IOMMU/swiotlb memory (even on AMD systems the IOMMU tends to be quite small) especially if you have many devices. With the new zone pci_alloc_consistent can just put this stuff into memory below 4GB which works better. One argument was still if the zone should be 4GB or 2GB. The main motivation for 2GB would be an unnamed not so unpopular hardware raid controller (mostly found in older machines from a particular four letter company) who has a strange 2GB restriction in firmware. But that one works ok with swiotlb/IOMMU anyways, so it doesn't really need GFP_DMA32. I chose 4GB to be compatible with IA64 and because it seems to be the most common restriction. The new zone is so far added only for x86-64. For other architectures who don't set up this new zone nothing changes. Architectures can set a compatibility define in Kconfig CONFIG_DMA_IS_DMA32 that will define GFP_DMA32 as GFP_DMA. Otherwise it's a nop because on 32bit architectures it's normally not needed because GFP_NORMAL (=0) is DMA able enough. One problem is still that GFP_DMA means different things on different architectures. e.g. some drivers used to have #ifdef ia64 use GFP_DMA (trusting it to be 4GB) #elif __x86_64__ (use other hacks like the swiotlb because 16MB is not enough) ... . This was quite ugly and is now obsolete. These should be now converted to use GFP_DMA32 unconditionally. I haven't done this yet. Or best only use pci_alloc_consistent/dma_alloc_coherent which will use GFP_DMA32 transparently. Signed-off-by:
Andi Kleen <ak@suse.de> Signed-off-by:
Linus Torvalds <torvalds@osdl.org>
-
- Sep 12, 2005
-
-
Jan Beulich authored
Rather than blindly re-enabling interrupts in oops_end(), save their state in oope_begin() and then restore that state. Signed-off-by:
Jan Beulich <jbeulich@novell.com> Signed-off-by:
-
Andi Kleen authored
The FLATMEM people added it, but there doesn't seem a good reason because end_pfn is identical. Signed-off-by:
-
- Sep 10, 2005
-
-
Adrian Bunk authored
get_cpu_vendor() no longer has any users in other files. Signed-off-by:
Adrian Bunk <bunk@stusta.de> Signed-off-by:
Andrew Morton <akpm@osdl.org> Signed-off-by:
-
- Jul 08, 2005
-
-
Shaohua Li authored
There has been some discuss about solving the SMP MTRR suspend/resume breakage, but I didn't find a patch for it. This is an intent for it. The basic idea is moving mtrr initializing into cpu_identify for all APs (so it works for cpu hotplug). For BP, restore_processor_state is responsible for restoring MTRR. Signed-off-by:
Shaohua Li <shaohua.li@intel.com> Acked-by:
Andi Kleen <ak@muc.de> Signed-off-by:
-
- May 17, 2005
-
-
Andi Kleen authored
There are unfortunately more and more multi processor Opteron systems which don't have HPET timer support in the southbridge. This covers in particular Nvidia and VIA chipsets. They also don't guarantee that the TSCs are synchronized between CPUs; and especially with MP powernow the systems are nearly unusable because the time gets very inconsistent between CPUs. The timer code for x86-64 was originally written under the assumption that we could fall back to the HPET timer on such systems. But this doesn't work there. Another alternative is to use the ACPI PM timer as primary time source. This patch does that. The kernel only uses PM timer when there is no other choice because it has some disadvantages. Ported over from i386. It should be faster than the i386 version because I dropped the "read three times" workaround, but is still considerable slower than HPET and also does not work together with vsyscalls which have to be disabled. Cc: <mark.langsdorf@amd.com> Signed-off-by:
-
- Apr 16, 2005
-
-
Andi Kleen authored
This will allow hotplug CPU in the future and in general cleans up a lot of crufty code. It also should plug some races that the old hackish way introduces. Remove one old race workaround in NMI watchdog setup that is not needed anymore. I removed the old total sum of bogomips reporting code. The brag value of BogoMips has been greatly devalued in the last years on the open market. Real CPU hotplug will need some more work, but the infrastructure for it is there now. One drawback: the new TSC sync algorithm is less accurate than before. The old way of zeroing TSCs is too intrusive to do later. Instead the TSC of the BP is duplicated now, which is less accurate. akpm: - sync_tsc_bp_init seems to have the sense of `init' inverted. - SPIN_LOCK_UNLOCKED is deprecated - use DEFINE_SPINLOCK. Cc: <rusty@rustcorp.com.au> Cc: <mingo@elte.hu> Signed-off-by:
-
Andi Kleen authored
Use a real VMA to map the 32bit vsyscall page This interacts better with Hugh's upcomming VMA walk optimization Also removes some ugly special cases. Code roughly modelled after the ppc64 vdso version from Ben Herrenschmidt. Signed-off-by:
-
Linus Torvalds authored
Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
-
