I am using "radeonkms" on a machine with quite mixed, diverse and varying loads. Sometimes I get an X server crash like the following: kernel: [TTM] Unable to allocate page kernel: error: [drm:pid1815:radeon_gem_object_create] *ERROR* Failed to allocate GEM object (25591808, 2, 4096, -12) kernel: [TTM] Unable to allocate page kernel: [TTM] Buffer eviction failed kernel: vm_fault: pager read error, pid 1815 (Xorg) kernel: pid 1815 (Xorg), uid 0: exited on signal 11 (core dumped) At the same time there was the following in X servers stderr: Failed to allocate : size : 25589760 bytes alignment : 256 bytes domains : 2 I wonder if this is a generic problem for example caused by severe resource exhaustion or if this is something where FreeBSD specific code does not do its best. In particular, it caught my attention that ttm_get_pages() effectively has semantics of M_NOWAIT as it never retries allocation failures in vm_page_alloc_contig(). -- Andriy GaponReceived on Fri Feb 14 2014 - 12:57:12 UTC
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:40:46 UTC