Re: kern/54817 (kernel stall in radeon_gem_object_create)

[Izumi Tsutsui <tsutsui%ceres.dti.ne.jp@localhost>]

The following reply was made to PR kern/54817; it has been noted by GNATS.

From: Izumi Tsutsui <tsutsui%ceres.dti.ne.jp@localhost>
To: mrg%netbsd.org@localhost, gnats-bugs%netbsd.org@localhost
Cc: tsutsui%ceres.dti.ne.jp@localhost
Subject: Re: kern/54817 (kernel stall in radeon_gem_object_create)
Date: Sun, 23 Feb 2020 15:35:38 +0900

 > > for the ENOMEM failures in PR/54817 can you try something with
 > > my latest change to include/drm/bus_dma_hacks.h (1.20) ?
 > > 
 > > this has a chance of allowing ENOMEM to be avoided, and if it
 > > doesn't help for you still, there is another method we can use
 > > to avoid any alloc at all i think.
 > 
 > I'll check this later.
 
 The "*ERROR* Failed to allocate GEM object" message didn't appear
 for ~three hour load, but now I get the following kernel fault:
 (typed from picture)
 
 ---
 kernel: supervisor trap page fault, code=0
 Stopped in pid 18756.1 (cc1plus) at     netbsd:amap_wipeout+0x77:       movl   
 4(%edx),%eax
 db{1}> bt
 amap_wipeout(d73e5f54,1,1b,0,c78e394c,0,c78e3958,e1aa5e70,c0862b5c,d33f5500) at
 netbsd:amap_wipeout+0x77
 uvm_unmap_detach(d33f5500,0,bfbff000,e1aa5e60,0,d33f5500,c9387d94,c888b9c0,e1aa5
 e9c,e1aa5f1c) at netbsd:uvm_unmap_detach+0x41
 uvmspace_free(c78e394c,0,e1aa5e9c,c3a8e1a0,c3a7f1c4,c3a97d48,ffffffff,ffffffff,c
 9387da8,0) at netbsd:uvmspace_free+0xf1
 exit1(c888b9c0,0,0,c888b9c0,e1aa5fa8,e1aa5f9c,c0153d67,c888b9c0,e1aa5f68,e1aa5f6
 0) at netbsd:ext1+0x178
 sys_exit(c888b9c0,e1aa5f68,e1aa5f60,8,213,e1aa5f60,1,e1aa5f68,0,0) at netbsd:sys
 _ext+0x3b
 syscall() at netbsd:syscall+0x127
 --- syscall (number 1) ---
 bb5b1937:
 db{1}> 
 ---
 
 But maybe this is yet another problem on memory shortage?
 
 I'll continue to check if the original radeon gem error occurs or not.
 
 Thanks,
 ---
 Izumi Tsutsui