Re: Linux 4.10-rc2 arm: dmesg flooded with alloc_contig_range: [X, Y) PFNs busy
On 01/18/2017 01:23 AM, Shuah Khan wrote: > Hi, > > dmesg floods with PFNs busy messages. > > [10119.071455] alloc_contig_range: [bb900, bbc00) PFNs busy > [10119.071631] alloc_contig_range: [bba00, bbd00) PFNs busy > [10119.071762] alloc_contig_range: [bbb00, bbe00) PFNs busy > [10119.071940] alloc_contig_range: [bbc00, bbf00) PFNs busy > [10119.072039] alloc_contig_range: [bbd00, bc000) PFNs busy > [10119.072188] alloc_contig_range: [bbe00, bc100) PFNs busy > [10119.072301] alloc_contig_range: [bbf00, bc200) PFNs busy > [10119.072403] alloc_contig_range: [bc000, bc300) PFNs busy > [10119.072549] alloc_contig_range: [bc100, bc400) PFNs busy > [10119.072584] [drm:exynos_drm_gem_create] *ERROR* failed to allocate buffer. > > I think this is triggered when drm tries to allocate CMA buffers. > I might have seen one or two messages in 4.9, but since 4.10, it > just floods dmesg. > > Is this a known problem? I am seeing this on odroid-xu4 Hi, yes most likely it's the problem fixed by this patch in mmotm/linux-next: http://ozlabs.org/~akpm/mmots/broken-out/mm-alloc_contig-re-allow-cma-to-compact-fs-pages.patch > Linux odroid 4.10.0-rc2-00251-ge03c755-dirty #12 SMP PREEMPT > Wed Jan 11 23:12:52 UTC 2017 armv7l armv7l armv7l GNU/Linux > > thanks, > -- Shuah >
Linux 4.10-rc2 arm: dmesg flooded with alloc_contig_range: [X, Y) PFNs busy
Hi, dmesg floods with PFNs busy messages. [10119.071455] alloc_contig_range: [bb900, bbc00) PFNs busy [10119.071631] alloc_contig_range: [bba00, bbd00) PFNs busy [10119.071762] alloc_contig_range: [bbb00, bbe00) PFNs busy [10119.071940] alloc_contig_range: [bbc00, bbf00) PFNs busy [10119.072039] alloc_contig_range: [bbd00, bc000) PFNs busy [10119.072188] alloc_contig_range: [bbe00, bc100) PFNs busy [10119.072301] alloc_contig_range: [bbf00, bc200) PFNs busy [10119.072403] alloc_contig_range: [bc000, bc300) PFNs busy [10119.072549] alloc_contig_range: [bc100, bc400) PFNs busy [10119.072584] [drm:exynos_drm_gem_create] *ERROR* failed to allocate buffer. I think this is triggered when drm tries to allocate CMA buffers. I might have seen one or two messages in 4.9, but since 4.10, it just floods dmesg. Is this a known problem? I am seeing this on odroid-xu4 Linux odroid 4.10.0-rc2-00251-ge03c755-dirty #12 SMP PREEMPT Wed Jan 11 23:12:52 UTC 2017 armv7l armv7l armv7l GNU/Linux thanks, -- Shuah -- Shuah Khan Sr. Linux Kernel Developer Open Source Innovation Group Samsung Research America (Silicon Valley) shua...@osg.samsung.com
Linux 4.10-rc2
Hey, it's been a really slow week between Christmas Day and New Years Day, and I am not complaining at all. It does mean that rc2 is ridiculously and unrealistically small. I almost decided to skip rc2 entirely, but a small little meaningless release every once in a while never hurt anybody. So here it is. The only even remotely noticeable work here is the DAX fixups that really arguably should have been merge window material but depended on stuff during this merge window and were delayed until rc2 due to that. Even that wasn't big, and the rest is trivial small fixes. I'm expecting things to start picking up next week as people recover from the holidays. Linus --- Al Viro (1): arm64: don't pull uaccess.h into *.S Chun-Hao Lin (1): r8169: add support for RTL8168 series add-on card. Cihangir Akturk (1): Documentation/unaligned-memory-access.txt: fix incorrect comparison operator Daniel Borkmann (1): net, sched: fix soft lockup in tc_classify Eric Dumazet (1): ipvlan: fix various issues in ipvlan_process_multicast() Florian Fainelli (1): net: korina: Fix NAPI versus resources freeing Haishuang Yan (1): ipv4: Namespaceify tcp_tw_reuse knob Jan Kara (6): ext2: Return BH_New buffers for zeroed blocks mm: Invalidate DAX radix tree entries only if appropriate dax: Avoid page invalidation races and unnecessary radix tree traversals dax: Finish fault completely when loading holes dax: Call ->iomap_begin without entry lock during dax fault ext4: Simplify DAX fault path Jason Wang (1): net: xdp: remove unused bfp_warn_invalid_xdp_buffer() John Brooks (1): docs: Fix build failure Jon Paul Maloy (1): tipc: don't send FIN message from connectionless socket Kweh, Hock Leong (1): net: stmmac: fix incorrect bit set in gmac4 mdio addr register Laura Abbott (1): crypto: testmgr - Use heap buffer for acomp test input Linus Torvalds (2): mm: optimize PageWaiters bit use for unlock_page() Linux 4.10-rc2 Mahesh Bandewar (1): ipvlan: fix multicast processing Olof Johansson (1): mm/filemap: fix parameters to test_bit() Romain Perier (1): crypto: marvell - Copy IVDIG before launching partial DMA ahash requests Tariq Toukan (1): net/mlx4_en: Fix user prio field in XDP forward Thomas Gleixner (2): smp/hotplug: Undo tglxs brainfart x86/mce/AMD: Make the init code more robust pravin shelar (1): openvswitch: upcall: Fix vlan handling.