[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [PATCH] x86/pv: Fix construction of 32bit dom0's


  • To: Jan Beulich <JBeulich@xxxxxxxx>
  • From: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>
  • Date: Thu, 7 Feb 2019 13:29:18 +0000
  • Autocrypt: addr=andrew.cooper3@xxxxxxxxxx; prefer-encrypt=mutual; keydata= mQINBFLhNn8BEADVhE+Hb8i0GV6mihnnr/uiQQdPF8kUoFzCOPXkf7jQ5sLYeJa0cQi6Penp VtiFYznTairnVsN5J+ujSTIb+OlMSJUWV4opS7WVNnxHbFTPYZVQ3erv7NKc2iVizCRZ2Kxn srM1oPXWRic8BIAdYOKOloF2300SL/bIpeD+x7h3w9B/qez7nOin5NzkxgFoaUeIal12pXSR Q354FKFoy6Vh96gc4VRqte3jw8mPuJQpfws+Pb+swvSf/i1q1+1I4jsRQQh2m6OTADHIqg2E ofTYAEh7R5HfPx0EXoEDMdRjOeKn8+vvkAwhviWXTHlG3R1QkbE5M/oywnZ83udJmi+lxjJ5 YhQ5IzomvJ16H0Bq+TLyVLO/VRksp1VR9HxCzItLNCS8PdpYYz5TC204ViycobYU65WMpzWe LFAGn8jSS25XIpqv0Y9k87dLbctKKA14Ifw2kq5OIVu2FuX+3i446JOa2vpCI9GcjCzi3oHV e00bzYiHMIl0FICrNJU0Kjho8pdo0m2uxkn6SYEpogAy9pnatUlO+erL4LqFUO7GXSdBRbw5 gNt25XTLdSFuZtMxkY3tq8MFss5QnjhehCVPEpE6y9ZjI4XB8ad1G4oBHVGK5LMsvg22PfMJ ISWFSHoF/B5+lHkCKWkFxZ0gZn33ju5n6/FOdEx4B8cMJt+cWwARAQABtClBbmRyZXcgQ29v cGVyIDxhbmRyZXcuY29vcGVyM0BjaXRyaXguY29tPokCOgQTAQgAJAIbAwULCQgHAwUVCgkI CwUWAgMBAAIeAQIXgAUCWKD95wIZAQAKCRBlw/kGpdefoHbdD/9AIoR3k6fKl+RFiFpyAhvO 59ttDFI7nIAnlYngev2XUR3acFElJATHSDO0ju+hqWqAb8kVijXLops0gOfqt3VPZq9cuHlh IMDquatGLzAadfFx2eQYIYT+FYuMoPZy/aTUazmJIDVxP7L383grjIkn+7tAv+qeDfE+txL4 SAm1UHNvmdfgL2/lcmL3xRh7sub3nJilM93RWX1Pe5LBSDXO45uzCGEdst6uSlzYR/MEr+5Z JQQ32JV64zwvf/aKaagSQSQMYNX9JFgfZ3TKWC1KJQbX5ssoX/5hNLqxMcZV3TN7kU8I3kjK mPec9+1nECOjjJSO/h4P0sBZyIUGfguwzhEeGf4sMCuSEM4xjCnwiBwftR17sr0spYcOpqET ZGcAmyYcNjy6CYadNCnfR40vhhWuCfNCBzWnUW0lFoo12wb0YnzoOLjvfD6OL3JjIUJNOmJy RCsJ5IA/Iz33RhSVRmROu+TztwuThClw63g7+hoyewv7BemKyuU6FTVhjjW+XUWmS/FzknSi dAG+insr0746cTPpSkGl3KAXeWDGJzve7/SBBfyznWCMGaf8E2P1oOdIZRxHgWj0zNr1+ooF /PzgLPiCI4OMUttTlEKChgbUTQ+5o0P080JojqfXwbPAyumbaYcQNiH1/xYbJdOFSiBv9rpt TQTBLzDKXok86LkCDQRS4TZ/ARAAkgqudHsp+hd82UVkvgnlqZjzz2vyrYfz7bkPtXaGb9H4 Rfo7mQsEQavEBdWWjbga6eMnDqtu+FC+qeTGYebToxEyp2lKDSoAsvt8w82tIlP/EbmRbDVn 7bhjBlfRcFjVYw8uVDPptT0TV47vpoCVkTwcyb6OltJrvg/QzV9f07DJswuda1JH3/qvYu0p vjPnYvCq4NsqY2XSdAJ02HrdYPFtNyPEntu1n1KK+gJrstjtw7KsZ4ygXYrsm/oCBiVW/OgU g/XIlGErkrxe4vQvJyVwg6YH653YTX5hLLUEL1NS4TCo47RP+wi6y+TnuAL36UtK/uFyEuPy wwrDVcC4cIFhYSfsO0BumEI65yu7a8aHbGfq2lW251UcoU48Z27ZUUZd2Dr6O/n8poQHbaTd 6bJJSjzGGHZVbRP9UQ3lkmkmc0+XCHmj5WhwNNYjgbbmML7y0fsJT5RgvefAIFfHBg7fTY/i kBEimoUsTEQz+N4hbKwo1hULfVxDJStE4sbPhjbsPCrlXf6W9CxSyQ0qmZ2bXsLQYRj2xqd1 bpA+1o1j2N4/au1R/uSiUFjewJdT/LX1EklKDcQwpk06Af/N7VZtSfEJeRV04unbsKVXWZAk uAJyDDKN99ziC0Wz5kcPyVD1HNf8bgaqGDzrv3TfYjwqayRFcMf7xJaL9xXedMcAEQEAAYkC HwQYAQgACQUCUuE2fwIbDAAKCRBlw/kGpdefoG4XEACD1Qf/er8EA7g23HMxYWd3FXHThrVQ HgiGdk5Yh632vjOm9L4sd/GCEACVQKjsu98e8o3ysitFlznEns5EAAXEbITrgKWXDDUWGYxd pnjj2u+GkVdsOAGk0kxczX6s+VRBhpbBI2PWnOsRJgU2n10PZ3mZD4Xu9kU2IXYmuW+e5KCA vTArRUdCrAtIa1k01sPipPPw6dfxx2e5asy21YOytzxuWFfJTGnVxZZSCyLUO83sh6OZhJkk b9rxL9wPmpN/t2IPaEKoAc0FTQZS36wAMOXkBh24PQ9gaLJvfPKpNzGD8XWR5HHF0NLIJhgg 4ZlEXQ2fVp3XrtocHqhu4UZR4koCijgB8sB7Tb0GCpwK+C4UePdFLfhKyRdSXuvY3AHJd4CP 4JzW0Bzq/WXY3XMOzUTYApGQpnUpdOmuQSfpV9MQO+/jo7r6yPbxT7CwRS5dcQPzUiuHLK9i nvjREdh84qycnx0/6dDroYhp0DFv4udxuAvt1h4wGwTPRQZerSm4xaYegEFusyhbZrI0U9tJ B8WrhBLXDiYlyJT6zOV2yZFuW47VrLsjYnHwn27hmxTC/7tvG3euCklmkn9Sl9IAKFu29RSo d5bD8kMSCYsTqtTfT6W4A3qHGvIDta3ptLYpIAOD2sY3GYq2nf3Bbzx81wZK14JdDDHUX2Rs 6+ahAA==
  • Cc: Xen-devel <xen-devel@xxxxxxxxxxxxx>, Wei Liu <wei.liu2@xxxxxxxxxx>, Roger Pau Monne <roger.pau@xxxxxxxxxx>
  • Delivery-date: Thu, 07 Feb 2019 13:29:30 +0000
  • List-id: Xen developer discussion <xen-devel.lists.xenproject.org>
  • Openpgp: preference=signencrypt

On 07/02/2019 12:58, Jan Beulich wrote:
On 06.02.19 at 21:41, <andrew.cooper3@xxxxxxxxxx> wrote:
Slightly RFC:

1) I've not worked out exactly what the

     v->vcpu_info = (void *)&d->shared_info->compat.vcpu_info[0];

   line is supposed to be doing and whether it is needed, but it doesn't
   appear to matter.  It is perhaps another redundant opencoding.
Afaict this is just to be independent of the fact that the vcpu_info
array is first in struct shared_info. I'd be fine with it getting replaced
by a respective BUILD_BUG_ON(), but I'd like to ask that it not be
dropped without replacement.

There is some ancillary logic with vcpu_info_mfn (which looks not to be initialised), which is why I suspect this is some more incorrect opencoding to begin with.


2) The reported

     Dom0 alloc.:   000000003e800000->000000003ec00000 (240470 pages to be allocated)

   line changes by 1 page because of the alloc_domheap_page() moving ahead of
   the printk(), but I'm fairly sure this is benign.  There is a matching
   reduction in the length of the constructed m2p which is perhaps less
   benign.
Well, the M2P of course has to be correctly sized. An off-by-one would
likely result in hard to repro bug reports.

The delta in output (with some of my own debugging) is:

@@ -22,13 +22,13 @@
 (XEN)     p2m_base         = 0xffffffffffffffff
 (XEN)  Xen  kernel: 64-bit, lsb, compat32
 (XEN)  Dom0 kernel: 32-bit, PAE, lsb, paddr 0x100000 -> 0x112000
-(XEN) ** nr_pages 241494
+(XEN) ** nr_pages 241493
 (XEN) PHYSICAL MEMORY ARRANGEMENT:
-(XEN)  Dom0 alloc.:   000000003e800000->000000003ec00000 (240470 pages to be allocated) (tot 1024, nr 241494)
+(XEN)  Dom0 alloc.:   000000003e800000->000000003ec00000 (240469 pages to be allocated) (tot 1024, nr 241493)
 (XEN) VIRTUAL MEMORY ARRANGEMENT:
 (XEN)  Loaded kernel: 0000000000100000->0000000000112000
 (XEN)  Init. ramdisk: 0000000000112000->0000000000112000
-(XEN)  Phys-Mach map: 0000000000112000->00000000001fdd58
+(XEN)  Phys-Mach map: 0000000000112000->00000000001fdd54
 (XEN)  Start info:    00000000001fe000->00000000001fe4b4
 (XEN)  Xenstore ring: 0000000000000000->0000000000000000
 (XEN)  Console ring:  0000000000000000->0000000000000000

I meant the P2M rather than M2P, and it is different by 1 entry which is expected, given the change by 1 page.  I've positively identified the 1-page change to be the alloc_domheap_page() for the monitor table moving.


@@ -606,23 +598,14 @@ int __init dom0_construct_pv(struct domain *d,
     {
         maddr_to_page(mpt_alloc)->u.inuse.type_info = PGT_l4_page_table;
         l4start = l4tab = __va(mpt_alloc); mpt_alloc += PAGE_SIZE;
+        clear_page(l4tab);
+        init_xen_l4_slots(l4tab, _mfn(virt_to_mfn(l4start)),
+                          d, INVALID_MFN, true);
+        v->arch.guest_table = pagetable_from_paddr(__pa(l4start));
     }
     else
-    {
-        page = alloc_domheap_page(d, MEMF_no_owner | MEMF_no_scrub);
-        if ( !page )
-            panic("Not enough RAM for domain 0 PML4\n");
-        page->u.inuse.type_info = PGT_l4_page_table|PGT_validated|1;
-        l4start = l4tab = page_to_virt(page);
-        maddr_to_page(mpt_alloc)->u.inuse.type_info = PGT_l3_page_table;
-        l3start = __va(mpt_alloc); mpt_alloc += PAGE_SIZE;
This one is lost without replacement, but is needed. Commit
7a9d764630 ("x86/32-on-64: adjust Dom0 initial page table layout")
specifically introduced it to make sure the guest-perceived top level
page table is allocated first (and hence marks the beginning of the
boot page tables, so Dom0 can later put all of them into general use).

I did call this out specifically in the commit message.  I had no idea about that commit when editing the code, but I still don't understand why it is important that the guests top level needs to be first.

~Andrew
_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel

 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.