1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
|
debug: 0.924
device: 0.921
permissions: 0.918
semantic: 0.917
assembly: 0.917
register: 0.916
performance: 0.914
user-level: 0.913
PID: 0.913
architecture: 0.912
boot: 0.909
system: 0.908
network: 0.907
graphic: 0.906
operating system: 0.906
arm: 0.901
KVM: 0.901
socket: 0.899
risc-v: 0.896
kernel: 0.895
virtual: 0.893
hypervisor: 0.889
TCG: 0.887
mistranslation: 0.882
files: 0.877
alpha: 0.876
vnc: 0.869
ppc: 0.864
VMM: 0.853
peripherals: 0.836
x86: 0.814
i386: 0.733
[Bug Report] Possible Missing Endianness Conversion
The virtio packed virtqueue support patch[1] suggests converting
endianness by lines:
virtio_tswap16s(vdev, &e->off_wrap);
virtio_tswap16s(vdev, &e->flags);
Though both of these conversion statements aren't present in the
latest qemu code here[2]
Is this intentional?
[1]:
https://mail.gnu.org/archive/html/qemu-block/2019-10/msg01492.html
[2]:
https://elixir.bootlin.com/qemu/latest/source/hw/virtio/virtio.c#L314
CCing Jason.
On Mon, Jun 24, 2024 at 4:30â¯PM Xoykie <xoykie@gmail.com> wrote:
>
>
The virtio packed virtqueue support patch[1] suggests converting
>
endianness by lines:
>
>
virtio_tswap16s(vdev, &e->off_wrap);
>
virtio_tswap16s(vdev, &e->flags);
>
>
Though both of these conversion statements aren't present in the
>
latest qemu code here[2]
>
>
Is this intentional?
Good catch!
It looks like it was removed (maybe by mistake) by commit
d152cdd6f6 ("virtio: use virtio accessor to access packed event")
Jason can you confirm that?
Thanks,
Stefano
>
>
[1]:
https://mail.gnu.org/archive/html/qemu-block/2019-10/msg01492.html
>
[2]:
https://elixir.bootlin.com/qemu/latest/source/hw/virtio/virtio.c#L314
>
On Mon, 24 Jun 2024 at 16:11, Stefano Garzarella <sgarzare@redhat.com> wrote:
>
>
CCing Jason.
>
>
On Mon, Jun 24, 2024 at 4:30â¯PM Xoykie <xoykie@gmail.com> wrote:
>
>
>
> The virtio packed virtqueue support patch[1] suggests converting
>
> endianness by lines:
>
>
>
> virtio_tswap16s(vdev, &e->off_wrap);
>
> virtio_tswap16s(vdev, &e->flags);
>
>
>
> Though both of these conversion statements aren't present in the
>
> latest qemu code here[2]
>
>
>
> Is this intentional?
>
>
Good catch!
>
>
It looks like it was removed (maybe by mistake) by commit
>
d152cdd6f6 ("virtio: use virtio accessor to access packed event")
That commit changes from:
- address_space_read_cached(cache, off_off, &e->off_wrap,
- sizeof(e->off_wrap));
- virtio_tswap16s(vdev, &e->off_wrap);
which does a byte read of 2 bytes and then swaps the bytes
depending on the host endianness and the value of
virtio_access_is_big_endian()
to this:
+ e->off_wrap = virtio_lduw_phys_cached(vdev, cache, off_off);
virtio_lduw_phys_cached() is a small function which calls
either lduw_be_phys_cached() or lduw_le_phys_cached()
depending on the value of virtio_access_is_big_endian().
(And lduw_be_phys_cached() and lduw_le_phys_cached() do
the right thing for the host-endianness to do a "load
a specifically big or little endian 16-bit value".)
Which is to say that because we use a load/store function that's
explicit about the size of the data type it is accessing, the
function itself can handle doing the load as big or little
endian, rather than the calling code having to do a manual swap after
it has done a load-as-bag-of-bytes. This is generally preferable
as it's less error-prone.
(Explicit swap-after-loading still has a place where the
code is doing a load of a whole structure out of the
guest and then swapping each struct field after the fact,
because it means we can do a single load-from-guest-memory
rather than a whole sequence of calls all the way down
through the memory subsystem.)
thanks
-- PMM
On Mon, Jun 24, 2024 at 04:19:52PM GMT, Peter Maydell wrote:
On Mon, 24 Jun 2024 at 16:11, Stefano Garzarella <sgarzare@redhat.com> wrote:
CCing Jason.
On Mon, Jun 24, 2024 at 4:30â¯PM Xoykie <xoykie@gmail.com> wrote:
>
> The virtio packed virtqueue support patch[1] suggests converting
> endianness by lines:
>
> virtio_tswap16s(vdev, &e->off_wrap);
> virtio_tswap16s(vdev, &e->flags);
>
> Though both of these conversion statements aren't present in the
> latest qemu code here[2]
>
> Is this intentional?
Good catch!
It looks like it was removed (maybe by mistake) by commit
d152cdd6f6 ("virtio: use virtio accessor to access packed event")
That commit changes from:
- address_space_read_cached(cache, off_off, &e->off_wrap,
- sizeof(e->off_wrap));
- virtio_tswap16s(vdev, &e->off_wrap);
which does a byte read of 2 bytes and then swaps the bytes
depending on the host endianness and the value of
virtio_access_is_big_endian()
to this:
+ e->off_wrap = virtio_lduw_phys_cached(vdev, cache, off_off);
virtio_lduw_phys_cached() is a small function which calls
either lduw_be_phys_cached() or lduw_le_phys_cached()
depending on the value of virtio_access_is_big_endian().
(And lduw_be_phys_cached() and lduw_le_phys_cached() do
the right thing for the host-endianness to do a "load
a specifically big or little endian 16-bit value".)
Which is to say that because we use a load/store function that's
explicit about the size of the data type it is accessing, the
function itself can handle doing the load as big or little
endian, rather than the calling code having to do a manual swap after
it has done a load-as-bag-of-bytes. This is generally preferable
as it's less error-prone.
Thanks for the details!
So, should we also remove `virtio_tswap16s(vdev, &e->flags);` ?
I mean:
diff --git a/hw/virtio/virtio.c b/hw/virtio/virtio.c
index 893a072c9d..2e5e67bdb9 100644
--- a/hw/virtio/virtio.c
+++ b/hw/virtio/virtio.c
@@ -323,7 +323,6 @@ static void vring_packed_event_read(VirtIODevice *vdev,
/* Make sure flags is seen before off_wrap */
smp_rmb();
e->off_wrap = virtio_lduw_phys_cached(vdev, cache, off_off);
- virtio_tswap16s(vdev, &e->flags);
}
static void vring_packed_off_wrap_write(VirtIODevice *vdev,
Thanks,
Stefano
(Explicit swap-after-loading still has a place where the
code is doing a load of a whole structure out of the
guest and then swapping each struct field after the fact,
because it means we can do a single load-from-guest-memory
rather than a whole sequence of calls all the way down
through the memory subsystem.)
thanks
-- PMM
On Tue, 25 Jun 2024 at 08:18, Stefano Garzarella <sgarzare@redhat.com> wrote:
>
>
On Mon, Jun 24, 2024 at 04:19:52PM GMT, Peter Maydell wrote:
>
>On Mon, 24 Jun 2024 at 16:11, Stefano Garzarella <sgarzare@redhat.com> wrote:
>
>>
>
>> CCing Jason.
>
>>
>
>> On Mon, Jun 24, 2024 at 4:30â¯PM Xoykie <xoykie@gmail.com> wrote:
>
>> >
>
>> > The virtio packed virtqueue support patch[1] suggests converting
>
>> > endianness by lines:
>
>> >
>
>> > virtio_tswap16s(vdev, &e->off_wrap);
>
>> > virtio_tswap16s(vdev, &e->flags);
>
>> >
>
>> > Though both of these conversion statements aren't present in the
>
>> > latest qemu code here[2]
>
>> >
>
>> > Is this intentional?
>
>>
>
>> Good catch!
>
>>
>
>> It looks like it was removed (maybe by mistake) by commit
>
>> d152cdd6f6 ("virtio: use virtio accessor to access packed event")
>
>
>
>That commit changes from:
>
>
>
>- address_space_read_cached(cache, off_off, &e->off_wrap,
>
>- sizeof(e->off_wrap));
>
>- virtio_tswap16s(vdev, &e->off_wrap);
>
>
>
>which does a byte read of 2 bytes and then swaps the bytes
>
>depending on the host endianness and the value of
>
>virtio_access_is_big_endian()
>
>
>
>to this:
>
>
>
>+ e->off_wrap = virtio_lduw_phys_cached(vdev, cache, off_off);
>
>
>
>virtio_lduw_phys_cached() is a small function which calls
>
>either lduw_be_phys_cached() or lduw_le_phys_cached()
>
>depending on the value of virtio_access_is_big_endian().
>
>(And lduw_be_phys_cached() and lduw_le_phys_cached() do
>
>the right thing for the host-endianness to do a "load
>
>a specifically big or little endian 16-bit value".)
>
>
>
>Which is to say that because we use a load/store function that's
>
>explicit about the size of the data type it is accessing, the
>
>function itself can handle doing the load as big or little
>
>endian, rather than the calling code having to do a manual swap after
>
>it has done a load-as-bag-of-bytes. This is generally preferable
>
>as it's less error-prone.
>
>
Thanks for the details!
>
>
So, should we also remove `virtio_tswap16s(vdev, &e->flags);` ?
>
>
I mean:
>
diff --git a/hw/virtio/virtio.c b/hw/virtio/virtio.c
>
index 893a072c9d..2e5e67bdb9 100644
>
--- a/hw/virtio/virtio.c
>
+++ b/hw/virtio/virtio.c
>
@@ -323,7 +323,6 @@ static void vring_packed_event_read(VirtIODevice *vdev,
>
/* Make sure flags is seen before off_wrap */
>
smp_rmb();
>
e->off_wrap = virtio_lduw_phys_cached(vdev, cache, off_off);
>
- virtio_tswap16s(vdev, &e->flags);
>
}
That definitely looks like it's probably not correct...
-- PMM
On Fri, Jun 28, 2024 at 03:53:09PM GMT, Peter Maydell wrote:
On Tue, 25 Jun 2024 at 08:18, Stefano Garzarella <sgarzare@redhat.com> wrote:
On Mon, Jun 24, 2024 at 04:19:52PM GMT, Peter Maydell wrote:
>On Mon, 24 Jun 2024 at 16:11, Stefano Garzarella <sgarzare@redhat.com> wrote:
>>
>> CCing Jason.
>>
>> On Mon, Jun 24, 2024 at 4:30â¯PM Xoykie <xoykie@gmail.com> wrote:
>> >
>> > The virtio packed virtqueue support patch[1] suggests converting
>> > endianness by lines:
>> >
>> > virtio_tswap16s(vdev, &e->off_wrap);
>> > virtio_tswap16s(vdev, &e->flags);
>> >
>> > Though both of these conversion statements aren't present in the
>> > latest qemu code here[2]
>> >
>> > Is this intentional?
>>
>> Good catch!
>>
>> It looks like it was removed (maybe by mistake) by commit
>> d152cdd6f6 ("virtio: use virtio accessor to access packed event")
>
>That commit changes from:
>
>- address_space_read_cached(cache, off_off, &e->off_wrap,
>- sizeof(e->off_wrap));
>- virtio_tswap16s(vdev, &e->off_wrap);
>
>which does a byte read of 2 bytes and then swaps the bytes
>depending on the host endianness and the value of
>virtio_access_is_big_endian()
>
>to this:
>
>+ e->off_wrap = virtio_lduw_phys_cached(vdev, cache, off_off);
>
>virtio_lduw_phys_cached() is a small function which calls
>either lduw_be_phys_cached() or lduw_le_phys_cached()
>depending on the value of virtio_access_is_big_endian().
>(And lduw_be_phys_cached() and lduw_le_phys_cached() do
>the right thing for the host-endianness to do a "load
>a specifically big or little endian 16-bit value".)
>
>Which is to say that because we use a load/store function that's
>explicit about the size of the data type it is accessing, the
>function itself can handle doing the load as big or little
>endian, rather than the calling code having to do a manual swap after
>it has done a load-as-bag-of-bytes. This is generally preferable
>as it's less error-prone.
Thanks for the details!
So, should we also remove `virtio_tswap16s(vdev, &e->flags);` ?
I mean:
diff --git a/hw/virtio/virtio.c b/hw/virtio/virtio.c
index 893a072c9d..2e5e67bdb9 100644
--- a/hw/virtio/virtio.c
+++ b/hw/virtio/virtio.c
@@ -323,7 +323,6 @@ static void vring_packed_event_read(VirtIODevice *vdev,
/* Make sure flags is seen before off_wrap */
smp_rmb();
e->off_wrap = virtio_lduw_phys_cached(vdev, cache, off_off);
- virtio_tswap16s(vdev, &e->flags);
}
That definitely looks like it's probably not correct...
Yeah, I just sent that patch:
20240701075208.19634-1-sgarzare@redhat.com
">https://lore.kernel.org/qemu-devel/
20240701075208.19634-1-sgarzare@redhat.com
We can continue the discussion there.
Thanks,
Stefano
|