Re: [PATCH V2 1/3] riscv: Fixup static_obj() fail

From: Palmer Dabbelt
Date: Mon Oct 05 2020 - 17:17:48 EST


On Mon, 05 Oct 2020 14:12:44 PDT (-0700), atishp@xxxxxxxxxxxxxx wrote:
On Mon, Oct 5, 2020 at 12:46 PM Palmer Dabbelt <palmerdabbelt@xxxxxxxxxx> wrote:

On Mon, 05 Oct 2020 11:40:54 PDT (-0700), schwab@xxxxxxxxxxxxxx wrote:
> On Okt 05 2020, Palmer Dabbelt wrote:
>
>> On Mon, 05 Oct 2020 01:25:22 PDT (-0700), schwab@xxxxxxxxxxxxxx wrote:
>>> On Sep 14 2020, Aurelien Jarno wrote:
>>>
>>>> How should we proceed to get that fixed in time for 5.9? For the older
>>>> branches where it has been backported (so far 5.7 and 5.8), should we
>>>> just get that commit reverted instead?
>>>
>>> Why is this still broken?
>>
>> Sorry, I hadn't seen this. I'm not seeing a boot failure on 5.9-rc8 with just
>> CONFIG_HARDENED_USERCPOY=y in addition to defconfig (on QEMU, though I doubt
>> that's relevant here).
>
> I don't see a boot failure either, but eventually you will get crashes
> like this, and resources are not properly released:
>
> [ 4560.936645] usercopy: Kernel memory overwrite attempt detected to kernel text (offset 241626, size 16)!
> [ 4560.945324] ------------[ cut here ]------------
> [ 4560.949954] kernel BUG at mm/usercopy.c:99!
> [ 4560.954030] Kernel BUG [#1]
> [ 4560.956805] Modules linked in: nfsv3 nfs_acl rfkill mmc_block sf_pdma i2c_ocores virt_dma spi_sifive uio_pdrv_genirq uio loop drm drm_panel_orientation_quirks rpcsec_gss_krb5 auth_rpcgss nfsv4 dns_resolver nfs lockd grace fscache af_packet mscc macsec macb ptp pps_core phylink of_mdio fixed_phy libphy pwm_sifive mmc_spi crc_itu_t crc7 of_mmc_spi mmc_core spi_bitbang sunrpc sg dm_multipath dm_mod scsi_dh_rdac scsi_dh_emc scsi_dh_alua
> [ 4560.995103] CPU: 2 PID: 23806 Comm: nis Not tainted 5.8.10-1-default #1 openSUSE Tumbleweed (unreleased)
> [ 4561.004563] epc: ffffffe00036140e ra : ffffffe00036140e sp : ffffffe004bc7d60
> [ 4561.011679] gp : ffffffe00127ee60 tp : ffffffe1b05d0000 t0 : ffffffe001297ca0
> [ 4561.018886] t1 : ffffffe001297c30 t2 : 0000000000000000 s0 : ffffffe004bc7d80
> [ 4561.026093] s1 : ffffffe00003afda a0 : 000000000000005b a1 : ffffffe1f7d67588
> [ 4561.033298] a2 : ffffffe1f7d6c108 a3 : 0000000000000000 a4 : ffffffe000043e80
> [ 4561.040506] a5 : ffffffe1f7d6be80 a6 : 0000000000000144 a7 : 0000000000000000
> [ 4561.047712] s2 : 0000000000000010 s3 : 0000000000000000 s4 : ffffffe00003afea
> [ 4561.054918] s5 : ffffffe1f7e00e80 s6 : 0000002af4a2c2e0 s7 : fffffffffffff000
> [ 4561.062124] s8 : 0000003ffffff000 s9 : ffffffe19f985400 s10: 0000000000000010
> [ 4561.069329] s11: ffffffe1f7e00e80 t3 : 0000000000038fa8 t4 : 0000000000038fa8
> [ 4561.076533] t5 : 0000000000000001 t6 : ffffffe00128e062
> [ 4561.081832] status: 0000000200000120 badaddr: 0000000000000000 cause: 0000000000000003
> [ 4561.089821] ---[ end trace a7c93e7595e9c2cc ]---
> [ 4561.095589] BUG: Bad rss-counter state mm:00000000c54f4c29 type:MM_ANONPAGES val:1

Ah, I must have misunderstood. I guess I just assumed "init crashes" meant on
boot, not just at some time later. I just sent out a patch reverting this, LMK
if it fixes the issue. I have some work stuff to do, but I'll try to find some
time tonight to look into fixing both of the bugs -- otherwise I'll just take
the revert (assuming it does actually fix the issue for you and passes the
tests).

I saw Atish post after I started writing this: I agree we need to sort of the
kernel's memory map, I just think it's too late for 5.9.


Yes. It is definitely a for-next material. I will try to take a stab
at this if nobody else has an objection.

WFM. Thanks!


> Andreas.

_______________________________________________
linux-riscv mailing list
linux-riscv@xxxxxxxxxxxxxxxxxxx
http://lists.infradead.org/mailman/listinfo/linux-riscv