Date: Tue, 17 Apr 2012 12:48:20 +0300 From: Konstantin Belousov <kostikbel@gmail.com> To: Ewart Tempest <etempest@juniper.net> Cc: alc@freebsd.org, "freebsd-hackers@freebsd.org" <freebsd-hackers@freebsd.org>, Tony Lanza <tlanza@juniper.net> Subject: Re: Corrupted pmap pm_vlist - pmap_remove_pte() Message-ID: <20120417094820.GK2358@deviant.kiev.zoral.com.ua> In-Reply-To: <A4C6A166C36F5F40A5767E6F66358FC0BB8B0B9681@EMBX01-WF.jnpr.net> References: <A4C6A166C36F5F40A5767E6F66358FC0BB8B0B9681@EMBX01-WF.jnpr.net>
next in thread | previous in thread | raw e-mail | index | archive | help
--nO+apMLAy/edto0B
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable
On Mon, Apr 16, 2012 at 03:08:25PM -0400, Ewart Tempest wrote:
> In FreeBSD 6.*, we have been seeing crashes in pmap_remove_pages() that o=
nly seem to occur in scaling scenarios:
>=20
> 2564 #ifdef PMAP_REMOVE_PAGES_CURPROC_ONLY
> 2565 pte =3D vtopte(pv->pv_va);
> 2566 #else
> 2567 pte =3D pmap_pte(pmap, pv->pv_va);
> 2568 #endif
> 2569 tpte =3D *pte; <=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D page fault here
>=20
> The suspicion is that the pmap's pm_pvlist list is getting corrupted. To =
this end, I have a question on the following logic in pmap_remove_pte() (se=
e in-line comment):
>=20
> 1533 static int
> 1534 pmap_remove_pte(pmap_t pmap, pt_entry_t *ptq, vm_offset_t va, pd_=
entry_t ptepde)
> 1535 {
> 1536 pt_entry_t oldpte;
> 1537 vm_page_t m;
> 1538=20
> 1539 PMAP_LOCK_ASSERT(pmap, MA_OWNED);
> 1540 oldpte =3D pte_load_clear(ptq);
> 1541 if (oldpte & PG_W)
> 1542 pmap->pm_stats.wired_count -=3D 1;
> 1543 /*
> 1544 * Machines that don't support invlpg, also don't support
> 1545 * PG_G.
> 1546 */
> 1547 if (oldpte & PG_G)
> 1548 pmap_invalidate_page(kernel_pmap, va);
> 1549 pmap->pm_stats.resident_count -=3D 1;
> 1550 if (oldpte & PG_MANAGED) {
> 1551 m =3D PHYS_TO_VM_PAGE(oldpte & PG_FRAME);
> 1552 if (oldpte & PG_M) {
> 1553 #if defined(PMAP_DIAGNOSTIC)
> 1554 if (pmap_nw_modified((pt_entry_t) oldpte)) {
> 1555 printf(
> 1556 "pmap_remove: modified page not writable: va: 0x%lx, pte: 0x%lx\=
n",
> 1557 va, oldpte);
> 1558 }
> 1559 #endif
> 1560 if (pmap_track_modified(va))
> 1561 vm_page_dirty(m);
> 1562 }
> 1563 if (oldpte & PG_A)
> 1564 vm_page_flag_set(m, PG_REFERENCED);
> 1565 pmap_remove_entry(pmap, m, va);
> 1566 }
> 1567 return (pmap_unuse_pt(pmap, va, ptepde)); <=3D=3D=3D=3D=3D=3D=3D=
*** under what circumstances is it valid to free the page but not remove i=
t from the pmap's pm_vlist? Even the code comment for pmap_unuse_pt() comme=
nces "After removing a page table entry ... ". ***
It is valid to not remove pv_entry when no pv_entry exists for the mapping.
The pv_entry is created if the page is managed, see pmap_enter() code.
The block above the return is executed when the page is managed, or at
least pmap thinks so.
The HEAD code will panic in pmap_pvh_free() if pmap_phv_remove() cannot
find the pv entry for given page and given pmap/va.
> 1568 }
>=20
> If the tail end of the above function is changed as follows:
>=20
> 1565 pmap_remove_entry(pmap, m, va);
> 1565.5 return (pmap_unuse_pt(pmap, va, ptepde));
> 1566 }
> 1567 return (0);
>=20
> Then we don't see any crashes ... but is it the right thing to do?
Should be not. Try to test this with some unmanaged mapping, like
/dev/mem pages mapped into the exiting process address space.
I am too new to know about any nuances of the RELENG_6 code.
--nO+apMLAy/edto0B
Content-Type: application/pgp-signature
Content-Disposition: inline
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.12 (FreeBSD)
iEYEARECAAYFAk+NPGQACgkQC3+MBN1Mb4jKjwCfZdJAjdcJ2d7SFP/NdhiTAZc/
kf4AnAmlE3ir/1n6zUackWmq8k5OTp81
=Gcsz
-----END PGP SIGNATURE-----
--nO+apMLAy/edto0B--
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20120417094820.GK2358>