linux-kernel - Re: [PATCH -next] arm64/mm: fix a bogus GFP flag in pgd

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <20190613194427.GC25164@rapoport-lnx>
Date:   Thu, 13 Jun 2019 22:44:28 +0300
From:   Mike Rapoport <rppt@...ux.ibm.com>
To:     Qian Cai <cai@....pw>
Cc:     Mark Rutland <mark.rutland@....com>,
        Will Deacon <will.deacon@....com>, akpm@...ux-foundation.org,
        Roman Gushchin <guro@...com>, catalin.marinas@....com,
        linux-kernel@...r.kernel.org, mhocko@...nel.org,
        linux-mm@...ck.org, vdavydov.dev@...il.com, hannes@...xchg.org,
        cgroups@...r.kernel.org, linux-arm-kernel@...ts.infradead.org
Subject: Re: [PATCH -next] arm64/mm: fix a bogus GFP flag in pgd_alloc()

On Thu, Jun 13, 2019 at 09:22:36AM -0400, Qian Cai wrote:
> On Thu, 2019-06-13 at 15:11 +0300, Mike Rapoport wrote:
> > The log Qian Cai posted at [1] and partially cited below confirms that the
> > failure happens when *user* PGDs are allocated and the addition of
> > __GFP_ACCOUNT to gfp flags used by pgd_alloc() only uncovered another
> > issue.
> > 
> > I'm still failing to reproduce it with qemu and I'm not really familiar
> > with slub/memcg code to say anything smart about it. Will keep looking.
> > 
> > Note, that as failures start way after efi_virtmap_init() that allocates a
> > PGD for efi_mm, there are no real fixes required for the original series,
> > except that the check for mm == &init_mm I copied for some reason from
> > powerpc is bogus and can be removed.
> 
> Yes, there is more places are not happy with __GFP_ACCOUNT other than efi_mm.
> For example,

Here we allocate the pgd for a user process and it should be accounted.

Actually, the whole point of changing the gfp flags in arm64::pgd_alloc()
was to enable the accounting for memory occupied by user pgds, just like
x86 and powerpc do.

> [  132.786842][ T1501] kobject_add_internal failed for pgd_cache(49:systemd-
> udevd.service) (error: -2 parent: cgroup)
> [  132.795589][ T1889] CPU: 9 PID: 1889 Comm: systemd-udevd Tainted:
> G        W         5.2.0-rc4-next-20190613+ #8
> [  132.807356][ T1889] Hardware name: HPE Apollo
> 70             /C01_APACHE_MB         , BIOS L50_5.13_1.0.9 03/01/2019
> [  132.817872][ T1889] Call trace:
> [  132.821017][ T1889]  dump_backtrace+0x0/0x268
> [  132.825372][ T1889]  show_stack+0x20/0x2c
> [  132.829380][ T1889]  dump_stack+0xb4/0x108
> [  132.833475][ T1889]  pgd_alloc+0x34/0x5c
> [  132.837396][ T1889]  mm_init+0x27c/0x32c
> [  132.841315][ T1889]  dup_mm+0x84/0x7b4
> [  132.845061][ T1889]  copy_process+0xf20/0x24cc
> [  132.849500][ T1889]  _do_fork+0xa4/0x66c
> [  132.853420][ T1889]  __arm64_sys_clone+0x114/0x1b4
> [  132.858208][ T1889]  el0_svc_handler+0x198/0x260
> [  132.862821][ T1889]  el0_svc+0x8/0xc
> 
> > 
> > I surely can add pgd_alloc_kernel() to be used by the EFI code to make sure
> > we won't run into issues with memcg in the future.
> > 
> > [   82.125966] Freeing unused kernel memory: 28672K
> > [   87.940365] Checked W+X mappings: passed, no W+X pages found
> > [   87.946769] Run /init as init process
> > [   88.040040] systemd[1]: System time before build time, advancing clock.
> > [   88.054593] systemd[1]: Failed to insert module 'autofs4': No such file or
> > directory
> > [   88.374129] modprobe (1726) used greatest stack depth: 28464 bytes left
> > [   88.470108] systemd[1]: systemd 239 running in system mode. (+PAM +AUDIT
> > +SELINUX +IMA -APPARMOR +SMACK +SYSVINIT +UTMP +LIBCRYPTSETUP +GCRYPT
> > +GNUTLS +ACL +XZ +LZ4 +SECCOMP +BLKID +ELFUTILS +KMOD +IDN2 -IDN +PCRE2
> > default-hierarchy=legacy)
> > [   88.498398] systemd[1]: Detected architecture arm64.
> > [   88.506517] systemd[1]: Running in initial RAM disk.
> > [   89.621995] mkdir (1730) used greatest stack depth: 27872 bytes left
> > [   90.222658] random: systemd: uninitialized urandom read (16 bytes read)
> > [   90.230072] systemd[1]: Reached target Swap.
> > [   90.240205] random: systemd: uninitialized urandom read (16 bytes read)
> > [   90.251088] systemd[1]: Reached target Timers.
> > [   90.261303] random: systemd: uninitialized urandom read (16 bytes read)
> > [   90.271209] systemd[1]: Listening on udev Control Socket.
> > [   90.283238] systemd[1]: Reached target Local File Systems.
> > [   90.296232] systemd[1]: Reached target Slices.
> > [   90.307239] systemd[1]: Listening on udev Kernel Socket.
> > [   90.608597] kobject_add_internal failed for pgd_cache(13:init.scope)
> > (error: -2 parent: cgroup)
> > [   90.678007] kobject_add_internal failed for pgd_cache(13:init.scope)(error:
> > -2 parent: cgroup)
> > [   90.713260] kobject_add_internal failed for pgd_cache(21:systemd-tmpfiles-
> > setup.service) (error: -2 parent: cgroup)
> > [   90.820012] systemd-tmpfile (1759) used greatest stack depth: 27184 bytes
> > left
> > [   90.861942] kobject_add_internal failed for pgd_cache(13:init.scope) error:
> > -2 parent: cgroup)
> >  
> > > Thanks,
> > > Mark.
> > > 
> > 
> > [1] https://cailca.github.io/files/dmesg.txt
> > 

-- 
Sincerely yours,
Mike.