On 7/12/18 05:06, Stijn Tintel wrote: > On 11/11/18 18:37, Stijn Tintel wrote: >> Hi, >> >> I have just pushed support for the 4.14 kernel on the brcm2708 target to >> my staging tree [1], and would like to get some feedback before pushing >> it to master. It would also be nice if people could do runtime tests on >> bcm2709 and bcm2710, as I don't own such hardware. >> >> Thanks, >> Stijn >> >> [1] >> https://git.openwrt.org/?p=openwrt/staging/stintel.git;a=shortlog;h=refs/heads/brcm2708-4_14 >> > Updated [1]: > - based on 4.14.86 > - split kmod changes in separate commits > - kept lan78xx patches except those that are superseded by upstream > changes, added another one that attempts to fix broken ethernet and > "kevent 4 may have been dropped" flood > - added missing compatible strings using upstream names for compute modules > - no longer removing the sense HAT patch, I might add kmod packages for > that later > > Compile-tested all 3 subtargets again with CONFIG_ALL_KMODS=y, and > runtime-tested bcm2708 on RPi0W and bcm2710 on RPi3B+. The wired > interface on the latter seems to be usable now. > > I'm planning to push this to master as soon as Koen pushed the 4.14.86 > bump. If there are any objections, please speak up ASAP. >
After flashing one of my RPi0W's running headless, it didn't come back. When testing it on my desk, it works just fine. After some digging I found that not having HDMI plugged in during boot causes the problem. This happens because brcmfmac fails to load due to a "vmalloc: allocation failure": [ 17.527326] vmap allocation for size 253952 failed: use vmalloc=<size> to increase size [ 17.540014] kmodloader: vmalloc: allocation failure: 249856 bytes, mode:0x14000c0(GFP_KERNEL), nodemask=(null) [ 17.554778] CPU: 0 PID: 212 Comm: kmodloader Tainted: G C O 4.14.87 #0 [ 17.566819] Hardware name: BCM2835 [ 17.572486] [<c010e5cc>] (unwind_backtrace) from [<c010b0cc>] (show_stack+0x10/0x14) [ 17.584649] [<c010b0cc>] (show_stack) from [<c017fe64>] (warn_alloc+0xac/0x168) [ 17.594362] [<c017fe64>] (warn_alloc) from [<c01ab5d4>] (__vmalloc_node_range+0x1f0/0x238) [ 17.607085] [<c01ab5d4>] (__vmalloc_node_range) from [<c010c444>] (module_alloc+0x3c/0x4c) [ 17.620019] [<c010c444>] (module_alloc) from [<c015ff10>] (load_module+0x7fc/0x2160) [ 17.632592] [<c015ff10>] (load_module) from [<c01619ac>] (SyS_init_module+0x138/0x168) [ 17.645312] [<c01619ac>] (SyS_init_module) from [<c0107580>] (ret_fast_syscall+0x0/0x54) [ 17.658224] Mem-Info: [ 17.662897] active_anon:232 inactive_anon:0 isolated_anon:0 [ 17.662897] active_file:2793 inactive_file:1495 isolated_file:0 [ 17.662897] unevictable:0 dirty:104 writeback:0 unstable:0 [ 17.662897] slab_reclaimable:670 slab_unreclaimable:1282 [ 17.662897] mapped:258 shmem:3 pagetables:21 bounce:0 [ 17.662897] free:102979 free_pcp:75 free_cma:3749 [ 17.708759] Node 0 active_anon:928kB inactive_anon:0kB active_file:11172kB inactive_file:5980kB unevictable:0kB isolated(anon):0kB isolated(file):0kB mapped:1032kB dirty:416kB writeback:0kB shmem:12kB writeback_tmp:0kB unstable:0kB al l_unreclaimable? no [ 17.740097] Normal free:411916kB min:2624kB low:3280kB high:3936kB active_anon:928kB inactive_anon:0kB active_file:11172kB inactive_file:5980kB unevictable:0kB writepending:416kB present:458752kB managed:447900kB mlocked:0kB kernel_st ack:360kB pagetables:84kB bounce:0kB free_pcp:300kB local_pcp:300kB free_cma:14996kB [ 17.777379] lowmem_reserve[]: 0 0 [ 17.782966] Normal: 3*4kB (UMC) 32*8kB (UMEC) 18*16kB (MEC) 7*32kB (UMC) 2*64kB (UC) 3*128kB (MC) 2*256kB (EC) 1*512kB (U) 2*1024kB (ME) 1*2048kB (C) 99*4096kB (MC) = 411916kB [ 17.805336] 4291 total pagecache pages [ 17.811267] 0 pages in swap cache [ 17.816776] Swap cache stats: add 0, delete 0, find 0/0 [ 17.824208] Free swap = 0kB [ 17.829227] Total swap = 0kB [ 17.834223] 114688 pages RAM [ 17.839137] 0 pages HighMem/MovableOnly [ 17.845094] 2713 pages reserved [ 17.850284] 4096 pages cma reserved I'm running into this problem because I have both kmod-video-bcm2835 and kmod-drm-vc4 enabled in my config. If I remove the related files that load these modules from /etc/modules.d, brcmfmac loads fine without HDMI attached. If I manually load either of the disabled modules afterwards, this fails the same way. I'm going to report this problem upstream, and will hold off pushing this to master until a fix is available. Stijn _______________________________________________ openwrt-devel mailing list openwrt-devel@lists.openwrt.org https://lists.openwrt.org/mailman/listinfo/openwrt-devel
