| Message ID | 20200717092929.19453-3-mhartmay@linux.ibm.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | Enable virtio-fs on s390x | expand |
On Fri, Jul 17, 2020 at 11:29:28AM +0200, Marc Hartmayer wrote: > Since virtio existed even before it got standardized, the virtio > standard defines the following types of virtio devices: > > + legacy device (pre-virtio 1.0) > + non-legacy or VIRTIO 1.0 device > + transitional device (which can act both as legacy and non-legacy) > > Virtio 1.0 defines the fields of the virtqueues as little endian, > while legacy uses guest's native endian [1]. Currently libvhost-user > does not handle virtio endianness at all, i.e. it works only if the > native endianness matches with whatever is actually needed. That means > things break spectacularly on big-endian targets. Let us handle virtio > endianness for non-legacy as required by the virtio specification > [1]. We will fence non-legacy virtio devices with the upcoming patch. > > [1] https://docs.oasis-open.org/virtio/virtio/v1.1/cs01/virtio-v1.1-cs01.html#x1-210003 > > Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com> > > --- > Note: As we don't support legacy virtio devices there is dead code in > libvhost-access.h that could be removed. But for the sake of > completeness, I left it in the code. Please remove the dead code. It is unlikely that legacy device support will be required in the future and it will just confuse people reading the code. > --- > contrib/libvhost-user/libvhost-access.h | 61 ++++++++++++ > contrib/libvhost-user/libvhost-user.c | 119 ++++++++++++------------ > 2 files changed, 121 insertions(+), 59 deletions(-) > create mode 100644 contrib/libvhost-user/libvhost-access.h > > diff --git a/contrib/libvhost-user/libvhost-access.h b/contrib/libvhost-user/libvhost-access.h > new file mode 100644 > index 000000000000..868ba3e7bfb8 > --- /dev/null > +++ b/contrib/libvhost-user/libvhost-access.h > @@ -0,0 +1,61 @@ > +#ifndef LIBVHOST_ACCESS_H License/copyright header?
On Fri, Jul 17, 2020 at 11:29:28AM +0200, Marc Hartmayer wrote: > Since virtio existed even before it got standardized, the virtio > standard defines the following types of virtio devices: > > + legacy device (pre-virtio 1.0) > + non-legacy or VIRTIO 1.0 device > + transitional device (which can act both as legacy and non-legacy) > > Virtio 1.0 defines the fields of the virtqueues as little endian, > while legacy uses guest's native endian [1]. Currently libvhost-user > does not handle virtio endianness at all, i.e. it works only if the > native endianness matches with whatever is actually needed. That means > things break spectacularly on big-endian targets. Let us handle virtio > endianness for non-legacy as required by the virtio specification > [1]. We will fence non-legacy virtio devices with the upcoming patch. > > [1] https://docs.oasis-open.org/virtio/virtio/v1.1/cs01/virtio-v1.1-cs01.html#x1-210003 > > Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com> > > --- > Note: As we don't support legacy virtio devices Who's "we" in this sentence? vhost user supports legacy generally ... > there is dead code in > libvhost-access.h that could be removed. But for the sake of > completeness, I left it in the code. > --- > contrib/libvhost-user/libvhost-access.h | 61 ++++++++++++ > contrib/libvhost-user/libvhost-user.c | 119 ++++++++++++------------ > 2 files changed, 121 insertions(+), 59 deletions(-) > create mode 100644 contrib/libvhost-user/libvhost-access.h > > diff --git a/contrib/libvhost-user/libvhost-access.h b/contrib/libvhost-user/libvhost-access.h > new file mode 100644 > index 000000000000..868ba3e7bfb8 > --- /dev/null > +++ b/contrib/libvhost-user/libvhost-access.h > @@ -0,0 +1,61 @@ > +#ifndef LIBVHOST_ACCESS_H > + > +#include "qemu/bswap.h" > + > +#include "libvhost-user.h" > + > +static inline bool vu_access_is_big_endian(VuDev *dev) > +{ > + /* Devices conforming to VIRTIO 1.0 or later are always LE. */ > + return false; > +} > + > +static inline void vu_stw_p(VuDev *vdev, void *ptr, uint16_t v) > +{ > + if (vu_access_is_big_endian(vdev)) { > + stw_be_p(ptr, v); > + } else { > + stw_le_p(ptr, v); > + } > +} > + > +static inline void vu_stl_p(VuDev *vdev, void *ptr, uint32_t v) > +{ > + if (vu_access_is_big_endian(vdev)) { > + stl_be_p(ptr, v); > + } else { > + stl_le_p(ptr, v); > + } > +} > + > +static inline void vu_stq_p(VuDev *vdev, void *ptr, uint64_t v) > +{ > + if (vu_access_is_big_endian(vdev)) { > + stq_be_p(ptr, v); > + } else { > + stq_le_p(ptr, v); > + } > +} > + > +static inline int vu_lduw_p(VuDev *vdev, const void *ptr) > +{ > + if (vu_access_is_big_endian(vdev)) > + return lduw_be_p(ptr); > + return lduw_le_p(ptr); > +} > + > +static inline int vu_ldl_p(VuDev *vdev, const void *ptr) > +{ > + if (vu_access_is_big_endian(vdev)) > + return ldl_be_p(ptr); > + return ldl_le_p(ptr); > +} > + > +static inline uint64_t vu_ldq_p(VuDev *vdev, const void *ptr) > +{ > + if (vu_access_is_big_endian(vdev)) > + return ldq_be_p(ptr); > + return ldq_le_p(ptr); > +} > + > +#endif /* LIBVHOST_ACCESS_H */ > diff --git a/contrib/libvhost-user/libvhost-user.c b/contrib/libvhost-user/libvhost-user.c > index d315db139606..0214b04c5291 100644 > --- a/contrib/libvhost-user/libvhost-user.c > +++ b/contrib/libvhost-user/libvhost-user.c > @@ -45,6 +45,7 @@ > #include "qemu/memfd.h" > > #include "libvhost-user.h" > +#include "libvhost-access.h" > > /* usually provided by GLib */ > #ifndef MIN > @@ -1074,7 +1075,7 @@ vu_set_vring_addr_exec(VuDev *dev, VhostUserMsg *vmsg) > return false; > } > > - vq->used_idx = vq->vring.used->idx; > + vq->used_idx = vu_lduw_p(dev, &vq->vring.used->idx); > > if (vq->last_avail_idx != vq->used_idx) { > bool resume = dev->iface->queue_is_processed_in_order && > @@ -1191,7 +1192,7 @@ vu_check_queue_inflights(VuDev *dev, VuVirtq *vq) > return 0; > } > > - vq->used_idx = vq->vring.used->idx; > + vq->used_idx = vu_lduw_p(dev, &vq->vring.used->idx); > vq->resubmit_num = 0; > vq->resubmit_list = NULL; > vq->counter = 0; > @@ -2019,35 +2020,35 @@ vu_queue_started(const VuDev *dev, const VuVirtq *vq) > } > > static inline uint16_t > -vring_avail_flags(VuVirtq *vq) > +vring_avail_flags(VuDev *dev, VuVirtq *vq) > { > - return vq->vring.avail->flags; > + return vu_lduw_p(dev, &vq->vring.avail->flags); > } > > static inline uint16_t > -vring_avail_idx(VuVirtq *vq) > +vring_avail_idx(VuDev *dev, VuVirtq *vq) > { > - vq->shadow_avail_idx = vq->vring.avail->idx; > + vq->shadow_avail_idx = vu_lduw_p(dev, &vq->vring.avail->idx); > > return vq->shadow_avail_idx; > } > > static inline uint16_t > -vring_avail_ring(VuVirtq *vq, int i) > +vring_avail_ring(VuDev *dev, VuVirtq *vq, int i) > { > - return vq->vring.avail->ring[i]; > + return vu_lduw_p(dev, &vq->vring.avail->ring[i]); > } > > static inline uint16_t > -vring_get_used_event(VuVirtq *vq) > +vring_get_used_event(VuDev *dev, VuVirtq *vq) > { > - return vring_avail_ring(vq, vq->vring.num); > + return vring_avail_ring(dev, vq, vq->vring.num); > } > > static int > virtqueue_num_heads(VuDev *dev, VuVirtq *vq, unsigned int idx) > { > - uint16_t num_heads = vring_avail_idx(vq) - idx; > + uint16_t num_heads = vring_avail_idx(dev, vq) - idx; > > /* Check it isn't doing very strange things with descriptor numbers. */ > if (num_heads > vq->vring.num) { > @@ -2070,7 +2071,7 @@ virtqueue_get_head(VuDev *dev, VuVirtq *vq, > { > /* Grab the next descriptor number they're advertising, and increment > * the index we've seen. */ > - *head = vring_avail_ring(vq, idx % vq->vring.num); > + *head = vring_avail_ring(dev, vq, idx % vq->vring.num); > > /* If their number is silly, that's a fatal mistake. */ > if (*head >= vq->vring.num) { > @@ -2123,12 +2124,12 @@ virtqueue_read_next_desc(VuDev *dev, struct vring_desc *desc, > int i, unsigned int max, unsigned int *next) > { > /* If this descriptor says it doesn't chain, we're done. */ > - if (!(desc[i].flags & VRING_DESC_F_NEXT)) { > + if (!(vu_lduw_p(dev, &desc[i].flags) & VRING_DESC_F_NEXT)) { > return VIRTQUEUE_READ_DESC_DONE; > } > > /* Check they're not leading us off end of descriptors. */ > - *next = desc[i].next; > + *next = vu_lduw_p(dev, &desc[i].next); > /* Make sure compiler knows to grab that: we don't want it changing! */ > smp_wmb(); > > @@ -2171,8 +2172,8 @@ vu_queue_get_avail_bytes(VuDev *dev, VuVirtq *vq, unsigned int *in_bytes, > } > desc = vq->vring.desc; > > - if (desc[i].flags & VRING_DESC_F_INDIRECT) { > - if (desc[i].len % sizeof(struct vring_desc)) { > + if (vu_lduw_p(dev, &desc[i].flags) & VRING_DESC_F_INDIRECT) { > + if (vu_ldl_p(dev, &desc[i].len) % sizeof(struct vring_desc)) { > vu_panic(dev, "Invalid size for indirect buffer table"); > goto err; > } > @@ -2185,8 +2186,8 @@ vu_queue_get_avail_bytes(VuDev *dev, VuVirtq *vq, unsigned int *in_bytes, > > /* loop over the indirect descriptor table */ > indirect = 1; > - desc_addr = desc[i].addr; > - desc_len = desc[i].len; > + desc_addr = vu_ldq_p(dev, &desc[i].addr); > + desc_len = vu_ldl_p(dev, &desc[i].len); > max = desc_len / sizeof(struct vring_desc); > read_len = desc_len; > desc = vu_gpa_to_va(dev, &read_len, desc_addr); > @@ -2213,10 +2214,10 @@ vu_queue_get_avail_bytes(VuDev *dev, VuVirtq *vq, unsigned int *in_bytes, > goto err; > } > > - if (desc[i].flags & VRING_DESC_F_WRITE) { > - in_total += desc[i].len; > + if (vu_lduw_p(dev, &desc[i].flags) & VRING_DESC_F_WRITE) { > + in_total += vu_ldl_p(dev, &desc[i].len); > } else { > - out_total += desc[i].len; > + out_total += vu_ldl_p(dev, &desc[i].len); > } > if (in_total >= max_in_bytes && out_total >= max_out_bytes) { > goto done; > @@ -2277,7 +2278,7 @@ vu_queue_empty(VuDev *dev, VuVirtq *vq) > return false; > } > > - return vring_avail_idx(vq) == vq->last_avail_idx; > + return vring_avail_idx(dev, vq) == vq->last_avail_idx; > } > > static bool > @@ -2296,14 +2297,14 @@ vring_notify(VuDev *dev, VuVirtq *vq) > } > > if (!vu_has_feature(dev, VIRTIO_RING_F_EVENT_IDX)) { > - return !(vring_avail_flags(vq) & VRING_AVAIL_F_NO_INTERRUPT); > + return !(vring_avail_flags(dev, vq) & VRING_AVAIL_F_NO_INTERRUPT); > } > > v = vq->signalled_used_valid; > vq->signalled_used_valid = true; > old = vq->signalled_used; > new = vq->signalled_used = vq->used_idx; > - return !v || vring_need_event(vring_get_used_event(vq), new, old); > + return !v || vring_need_event(vring_get_used_event(dev, vq), new, old); > } > > static void _vu_queue_notify(VuDev *dev, VuVirtq *vq, bool sync) > @@ -2361,33 +2362,33 @@ void vu_queue_notify_sync(VuDev *dev, VuVirtq *vq) > } > > static inline void > -vring_used_flags_set_bit(VuVirtq *vq, int mask) > +vring_used_flags_set_bit(VuDev *dev, VuVirtq *vq, int mask) > +{ > + uint16_t *flags; > + > + flags = (uint16_t *)(char*)vq->vring.used + > + offsetof(struct vring_used, flags); > + vu_stw_p(dev, flags, vu_lduw_p(dev, flags) | mask); > +} > + > +static inline void > +vring_used_flags_unset_bit(VuDev *dev, VuVirtq *vq, int mask) > { > uint16_t *flags; > > flags = (uint16_t *)((char*)vq->vring.used + > offsetof(struct vring_used, flags)); > - *flags |= mask; > + vu_stw_p(dev, flags, vu_lduw_p(dev, flags) & ~mask); > } > > static inline void > -vring_used_flags_unset_bit(VuVirtq *vq, int mask) > -{ > - uint16_t *flags; > - > - flags = (uint16_t *)((char*)vq->vring.used + > - offsetof(struct vring_used, flags)); > - *flags &= ~mask; > -} > - > -static inline void > -vring_set_avail_event(VuVirtq *vq, uint16_t val) > +vring_set_avail_event(VuDev *dev, VuVirtq *vq, uint16_t val) > { > if (!vq->notification) { > return; > } > > - *((uint16_t *) &vq->vring.used->ring[vq->vring.num]) = val; > + vu_stw_p(dev, &vq->vring.used->ring[vq->vring.num], val); > } > > void > @@ -2395,11 +2396,11 @@ vu_queue_set_notification(VuDev *dev, VuVirtq *vq, int enable) > { > vq->notification = enable; > if (vu_has_feature(dev, VIRTIO_RING_F_EVENT_IDX)) { > - vring_set_avail_event(vq, vring_avail_idx(vq)); > + vring_set_avail_event(dev, vq, vring_avail_idx(dev, vq)); > } else if (enable) { > - vring_used_flags_unset_bit(vq, VRING_USED_F_NO_NOTIFY); > + vring_used_flags_unset_bit(dev, vq, VRING_USED_F_NO_NOTIFY); > } else { > - vring_used_flags_set_bit(vq, VRING_USED_F_NO_NOTIFY); > + vring_used_flags_set_bit(dev, vq, VRING_USED_F_NO_NOTIFY); > } > if (enable) { > /* Expose avail event/used flags before caller checks the avail idx. */ > @@ -2476,14 +2477,14 @@ vu_queue_map_desc(VuDev *dev, VuVirtq *vq, unsigned int idx, size_t sz) > struct vring_desc desc_buf[VIRTQUEUE_MAX_SIZE]; > int rc; > > - if (desc[i].flags & VRING_DESC_F_INDIRECT) { > - if (desc[i].len % sizeof(struct vring_desc)) { > + if (vu_lduw_p(dev, &desc[i].flags) & VRING_DESC_F_INDIRECT) { > + if (vu_ldl_p(dev, &desc[i].len) % sizeof(struct vring_desc)) { > vu_panic(dev, "Invalid size for indirect buffer table"); > } > > /* loop over the indirect descriptor table */ > - desc_addr = desc[i].addr; > - desc_len = desc[i].len; > + desc_addr = vu_ldq_p(dev, &desc[i].addr); > + desc_len = vu_ldl_p(dev, &desc[i].len); > max = desc_len / sizeof(struct vring_desc); > read_len = desc_len; > desc = vu_gpa_to_va(dev, &read_len, desc_addr); > @@ -2505,10 +2506,10 @@ vu_queue_map_desc(VuDev *dev, VuVirtq *vq, unsigned int idx, size_t sz) > > /* Collect all the descriptors */ > do { > - if (desc[i].flags & VRING_DESC_F_WRITE) { > + if (vu_lduw_p(dev, &desc[i].flags) & VRING_DESC_F_WRITE) { > virtqueue_map_desc(dev, &in_num, iov + out_num, > VIRTQUEUE_MAX_SIZE - out_num, true, > - desc[i].addr, desc[i].len); > + vu_ldq_p(dev, &desc[i].addr), vu_ldl_p(dev, &desc[i].len)); > } else { > if (in_num) { > vu_panic(dev, "Incorrect order for descriptors"); > @@ -2516,7 +2517,7 @@ vu_queue_map_desc(VuDev *dev, VuVirtq *vq, unsigned int idx, size_t sz) > } > virtqueue_map_desc(dev, &out_num, iov, > VIRTQUEUE_MAX_SIZE, false, > - desc[i].addr, desc[i].len); > + vu_ldq_p(dev, &desc[i].addr), vu_ldl_p(dev, &desc[i].len)); > } > > /* If we've got too many, that implies a descriptor loop. */ > @@ -2642,7 +2643,7 @@ vu_queue_pop(VuDev *dev, VuVirtq *vq, size_t sz) > } > > if (vu_has_feature(dev, VIRTIO_RING_F_EVENT_IDX)) { > - vring_set_avail_event(vq, vq->last_avail_idx); > + vring_set_avail_event(dev, vq, vq->last_avail_idx); > } > > elem = vu_queue_map_desc(dev, vq, head, sz); > @@ -2712,14 +2713,14 @@ vu_log_queue_fill(VuDev *dev, VuVirtq *vq, > max = vq->vring.num; > i = elem->index; > > - if (desc[i].flags & VRING_DESC_F_INDIRECT) { > - if (desc[i].len % sizeof(struct vring_desc)) { > + if (vu_lduw_p(dev, &desc[i].flags) & VRING_DESC_F_INDIRECT) { > + if (vu_ldl_p(dev, &desc[i].len) % sizeof(struct vring_desc)) { > vu_panic(dev, "Invalid size for indirect buffer table"); > } > > /* loop over the indirect descriptor table */ > - desc_addr = desc[i].addr; > - desc_len = desc[i].len; > + desc_addr = vu_ldq_p(dev, &desc[i].addr); > + desc_len = vu_ldl_p(dev, &desc[i].len); > max = desc_len / sizeof(struct vring_desc); > read_len = desc_len; > desc = vu_gpa_to_va(dev, &read_len, desc_addr); > @@ -2745,9 +2746,9 @@ vu_log_queue_fill(VuDev *dev, VuVirtq *vq, > return; > } > > - if (desc[i].flags & VRING_DESC_F_WRITE) { > - min = MIN(desc[i].len, len); > - vu_log_write(dev, desc[i].addr, min); > + if (vu_lduw_p(dev, &desc[i].flags) & VRING_DESC_F_WRITE) { > + min = MIN(vu_ldl_p(dev, &desc[i].len), len); > + vu_log_write(dev, vu_ldq_p(dev, &desc[i].addr), min); > len -= min; > } > > @@ -2772,15 +2773,15 @@ vu_queue_fill(VuDev *dev, VuVirtq *vq, > > idx = (idx + vq->used_idx) % vq->vring.num; > > - uelem.id = elem->index; > - uelem.len = len; > + vu_stl_p(dev, &uelem.id, elem->index); > + vu_stl_p(dev, &uelem.len, len); > vring_used_write(dev, vq, &uelem, idx); > } > > static inline > void vring_used_idx_set(VuDev *dev, VuVirtq *vq, uint16_t val) > { > - vq->vring.used->idx = val; > + vu_stw_p(dev, &vq->vring.used->idx, val); > vu_log_write(dev, > vq->vring.log_guest_addr + offsetof(struct vring_used, idx), > sizeof(vq->vring.used->idx)); > -- > 2.25.4
On Tue, 21 Jul 2020 09:40:10 -0400 "Michael S. Tsirkin" <mst@redhat.com> wrote: > On Fri, Jul 17, 2020 at 11:29:28AM +0200, Marc Hartmayer wrote: > > Since virtio existed even before it got standardized, the virtio > > standard defines the following types of virtio devices: > > > > + legacy device (pre-virtio 1.0) > > + non-legacy or VIRTIO 1.0 device > > + transitional device (which can act both as legacy and non-legacy) > > > > Virtio 1.0 defines the fields of the virtqueues as little endian, > > while legacy uses guest's native endian [1]. Currently libvhost-user > > does not handle virtio endianness at all, i.e. it works only if the > > native endianness matches with whatever is actually needed. That means > > things break spectacularly on big-endian targets. Let us handle virtio > > endianness for non-legacy as required by the virtio specification > > [1]. We will fence non-legacy virtio devices with the upcoming patch. > > > > [1] https://docs.oasis-open.org/virtio/virtio/v1.1/cs01/virtio-v1.1-cs01.html#x1-210003 > > > > Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com> > > > > --- > > Note: As we don't support legacy virtio devices > > Who's "we" in this sentence? vhost user supports legacy generally ... In that sentence "we" is the library "libvhost-user". I would like to avoid s390x being an oddball regarding this. Marc's previous version made an attempt at correct endianness handling for legacy and non-legacy. That spawned a discussion on how we don't want legacy devices in this context. This series makes what seemed to be the consensus reached in that discussion explicit: namely that libvhost-user does not support legacy-virtio. Regards, Halil
On Tue, 21 Jul 2020 18:44:56 +0200 Halil Pasic <pasic@linux.ibm.com> wrote: > On Tue, 21 Jul 2020 09:40:10 -0400 > "Michael S. Tsirkin" <mst@redhat.com> wrote: > > > On Fri, Jul 17, 2020 at 11:29:28AM +0200, Marc Hartmayer wrote: > > > Since virtio existed even before it got standardized, the virtio > > > standard defines the following types of virtio devices: > > > > > > + legacy device (pre-virtio 1.0) > > > + non-legacy or VIRTIO 1.0 device > > > + transitional device (which can act both as legacy and non-legacy) > > > > > > Virtio 1.0 defines the fields of the virtqueues as little endian, > > > while legacy uses guest's native endian [1]. Currently libvhost-user > > > does not handle virtio endianness at all, i.e. it works only if the > > > native endianness matches with whatever is actually needed. That means > > > things break spectacularly on big-endian targets. Let us handle virtio > > > endianness for non-legacy as required by the virtio specification > > > [1]. We will fence non-legacy virtio devices with the upcoming patch. > > > > > > [1] https://docs.oasis-open.org/virtio/virtio/v1.1/cs01/virtio-v1.1-cs01.html#x1-210003 > > > > > > Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com> > > > > > > --- > > > Note: As we don't support legacy virtio devices > > > > Who's "we" in this sentence? vhost user supports legacy generally ... > > In that sentence "we" is the library "libvhost-user". I would like > to avoid s390x being an oddball regarding this. Marc's previous > version made an attempt at correct endianness handling for legacy > and non-legacy. That spawned a discussion on how we don't want > legacy devices in this context. This series makes what seemed to be the > consensus reached in that discussion explicit: namely that libvhost-user > does not support legacy-virtio. Can someone please clarify what libvhost-user actually supports? virtio-fs is clearly non-legacy only, but in the changelog I also see a mention of scsi, and that one probably does support legacy? Can we make a distinction along the lines of: - if we are non-legacy, we support any endianness; - if we are legacy, we support little endian host/guest only?
On Tue, Jul 21, 2020 at 06:44 PM +0200, Halil Pasic <pasic@linux.ibm.com> wrote: > On Tue, 21 Jul 2020 09:40:10 -0400 > "Michael S. Tsirkin" <mst@redhat.com> wrote: > >> On Fri, Jul 17, 2020 at 11:29:28AM +0200, Marc Hartmayer wrote: >> > Since virtio existed even before it got standardized, the virtio >> > standard defines the following types of virtio devices: >> > >> > + legacy device (pre-virtio 1.0) >> > + non-legacy or VIRTIO 1.0 device >> > + transitional device (which can act both as legacy and non-legacy) >> > >> > Virtio 1.0 defines the fields of the virtqueues as little endian, >> > while legacy uses guest's native endian [1]. Currently libvhost-user >> > does not handle virtio endianness at all, i.e. it works only if the >> > native endianness matches with whatever is actually needed. That means >> > things break spectacularly on big-endian targets. Let us handle virtio >> > endianness for non-legacy as required by the virtio specification >> > [1]. We will fence non-legacy virtio devices with the upcoming patch. >> > >> > [1] https://docs.oasis-open.org/virtio/virtio/v1.1/cs01/virtio-v1.1-cs01.html#x1-210003 >> > >> > Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com> >> > >> > --- >> > Note: As we don't support legacy virtio devices >> >> Who's "we" in this sentence? vhost user supports legacy generally ... > > In that sentence "we" is the library "libvhost-user". I would like > to avoid s390x being an oddball regarding this. Marc's previous > version made an attempt at correct endianness handling for legacy > and non-legacy. That spawned a discussion on how we don't want > legacy devices in this context. This series makes what seemed to be the > consensus reached in that discussion explicit: namely that libvhost-user > does not support legacy-virtio. Hi Michael, Polite ping - what’s your opinion? Thanks in advance. […snip]
On Tue, 28 Jul 2020 12:48:50 +0200 Cornelia Huck <cohuck@redhat.com> wrote: > On Tue, 21 Jul 2020 18:44:56 +0200 > Halil Pasic <pasic@linux.ibm.com> wrote: > > > On Tue, 21 Jul 2020 09:40:10 -0400 > > "Michael S. Tsirkin" <mst@redhat.com> wrote: > > > > > On Fri, Jul 17, 2020 at 11:29:28AM +0200, Marc Hartmayer wrote: > > > > Since virtio existed even before it got standardized, the virtio > > > > standard defines the following types of virtio devices: > > > > > > > > + legacy device (pre-virtio 1.0) > > > > + non-legacy or VIRTIO 1.0 device > > > > + transitional device (which can act both as legacy and non-legacy) > > > > > > > > Virtio 1.0 defines the fields of the virtqueues as little endian, > > > > while legacy uses guest's native endian [1]. Currently libvhost-user > > > > does not handle virtio endianness at all, i.e. it works only if the > > > > native endianness matches with whatever is actually needed. That means > > > > things break spectacularly on big-endian targets. Let us handle virtio > > > > endianness for non-legacy as required by the virtio specification > > > > [1]. We will fence non-legacy virtio devices with the upcoming patch. > > > > > > > > [1] https://docs.oasis-open.org/virtio/virtio/v1.1/cs01/virtio-v1.1-cs01.html#x1-210003 > > > > > > > > Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com> > > > > > > > > --- > > > > Note: As we don't support legacy virtio devices > > > > > > Who's "we" in this sentence? vhost user supports legacy generally ... > > > > In that sentence "we" is the library "libvhost-user". I would like > > to avoid s390x being an oddball regarding this. Marc's previous > > version made an attempt at correct endianness handling for legacy > > and non-legacy. That spawned a discussion on how we don't want > > legacy devices in this context. This series makes what seemed to be the > > consensus reached in that discussion explicit: namely that libvhost-user > > does not support legacy-virtio. > > Can someone please clarify what libvhost-user actually supports? > > virtio-fs is clearly non-legacy only, but in the changelog I also see a > mention of scsi, and that one probably does support legacy? > > Can we make a distinction along the lines of: > - if we are non-legacy, we support any endianness; > - if we are legacy, we support little endian host/guest only? > I don't like the idea of this conditional support of legacy (with a condition being a little endian host/guest). The library is supposed to be portable in my opinion. Regards, Halil
On Tue, Jul 28, 2020 at 12:52:11PM +0200, Marc Hartmayer wrote: > On Tue, Jul 21, 2020 at 06:44 PM +0200, Halil Pasic <pasic@linux.ibm.com> wrote: > > On Tue, 21 Jul 2020 09:40:10 -0400 > > "Michael S. Tsirkin" <mst@redhat.com> wrote: > > > >> On Fri, Jul 17, 2020 at 11:29:28AM +0200, Marc Hartmayer wrote: > >> > Since virtio existed even before it got standardized, the virtio > >> > standard defines the following types of virtio devices: > >> > > >> > + legacy device (pre-virtio 1.0) > >> > + non-legacy or VIRTIO 1.0 device > >> > + transitional device (which can act both as legacy and non-legacy) > >> > > >> > Virtio 1.0 defines the fields of the virtqueues as little endian, > >> > while legacy uses guest's native endian [1]. Currently libvhost-user > >> > does not handle virtio endianness at all, i.e. it works only if the > >> > native endianness matches with whatever is actually needed. That means > >> > things break spectacularly on big-endian targets. Let us handle virtio > >> > endianness for non-legacy as required by the virtio specification > >> > [1]. We will fence non-legacy virtio devices with the upcoming patch. > >> > > >> > [1] https://docs.oasis-open.org/virtio/virtio/v1.1/cs01/virtio-v1.1-cs01.html#x1-210003 > >> > > >> > Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com> > >> > > >> > --- > >> > Note: As we don't support legacy virtio devices > >> > >> Who's "we" in this sentence? vhost user supports legacy generally ... > > > > In that sentence "we" is the library "libvhost-user". I would like > > to avoid s390x being an oddball regarding this. Marc's previous > > version made an attempt at correct endianness handling for legacy > > and non-legacy. That spawned a discussion on how we don't want > > legacy devices in this context. This series makes what seemed to be the > > consensus reached in that discussion explicit: namely that libvhost-user > > does not support legacy-virtio. > > Hi Michael, > > Polite ping - what’s your opinion? Thanks in advance. > > […snip] It's a reasonable limitation, but I also don't see anything that verifies that device is modern only. E.g. fail setting features if VIRTIO_F_VERSION_1 is not there? > -- > Kind regards / Beste Grüße > Marc Hartmayer > > IBM Deutschland Research & Development GmbH > Vorsitzender des Aufsichtsrats: Gregor Pillen > Geschäftsführung: Dirk Wittkopp > Sitz der Gesellschaft: Böblingen > Registergericht: Amtsgericht Stuttgart, HRB 243294
On Wed, Jul 29, 2020 at 10:13 AM -0400, "Michael S. Tsirkin" <mst@redhat.com> wrote: > On Tue, Jul 28, 2020 at 12:52:11PM +0200, Marc Hartmayer wrote: >> On Tue, Jul 21, 2020 at 06:44 PM +0200, Halil Pasic <pasic@linux.ibm.com> wrote: >> > On Tue, 21 Jul 2020 09:40:10 -0400 >> > "Michael S. Tsirkin" <mst@redhat.com> wrote: >> > >> >> On Fri, Jul 17, 2020 at 11:29:28AM +0200, Marc Hartmayer wrote: >> >> > Since virtio existed even before it got standardized, the virtio >> >> > standard defines the following types of virtio devices: >> >> > >> >> > + legacy device (pre-virtio 1.0) >> >> > + non-legacy or VIRTIO 1.0 device >> >> > + transitional device (which can act both as legacy and non-legacy) >> >> > >> >> > Virtio 1.0 defines the fields of the virtqueues as little endian, >> >> > while legacy uses guest's native endian [1]. Currently libvhost-user >> >> > does not handle virtio endianness at all, i.e. it works only if the >> >> > native endianness matches with whatever is actually needed. That means >> >> > things break spectacularly on big-endian targets. Let us handle virtio >> >> > endianness for non-legacy as required by the virtio specification >> >> > [1]. We will fence non-legacy virtio devices with the upcoming patch. >> >> > >> >> > [1] https://docs.oasis-open.org/virtio/virtio/v1.1/cs01/virtio-v1.1-cs01.html#x1-210003 >> >> > >> >> > Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com> >> >> > >> >> > --- >> >> > Note: As we don't support legacy virtio devices >> >> >> >> Who's "we" in this sentence? vhost user supports legacy generally ... >> > >> > In that sentence "we" is the library "libvhost-user". I would like >> > to avoid s390x being an oddball regarding this. Marc's previous >> > version made an attempt at correct endianness handling for legacy >> > and non-legacy. That spawned a discussion on how we don't want >> > legacy devices in this context. This series makes what seemed to be the >> > consensus reached in that discussion explicit: namely that libvhost-user >> > does not support legacy-virtio. >> >> Hi Michael, >> >> Polite ping - what’s your opinion? Thanks in advance. >> >> […snip] > > It's a reasonable limitation, but I also don't see anything > that verifies that device is modern only. > E.g. fail setting features if VIRTIO_F_VERSION_1 is not there? That’s implemented in patch 3 to emphasize how the fencing is done. For the next series I’ll merge patch 2 and 3. > > >> -- >> Kind regards / Beste Grüße >> Marc Hartmayer >> >> IBM Deutschland Research & Development GmbH >> Vorsitzender des Aufsichtsrats: Gregor Pillen >> Geschäftsführung: Dirk Wittkopp >> Sitz der Gesellschaft: Böblingen >> Registergericht: Amtsgericht Stuttgart, HRB 243294 >
On Tue, Jul 21, 2020 at 01:48 PM +0100, Stefan Hajnoczi <stefanha@redhat.com> wrote: > On Fri, Jul 17, 2020 at 11:29:28AM +0200, Marc Hartmayer wrote: >> Since virtio existed even before it got standardized, the virtio >> standard defines the following types of virtio devices: >> >> + legacy device (pre-virtio 1.0) >> + non-legacy or VIRTIO 1.0 device >> + transitional device (which can act both as legacy and non-legacy) >> >> Virtio 1.0 defines the fields of the virtqueues as little endian, >> while legacy uses guest's native endian [1]. Currently libvhost-user >> does not handle virtio endianness at all, i.e. it works only if the >> native endianness matches with whatever is actually needed. That means >> things break spectacularly on big-endian targets. Let us handle virtio >> endianness for non-legacy as required by the virtio specification >> [1]. We will fence non-legacy virtio devices with the upcoming patch. >> >> [1] https://docs.oasis-open.org/virtio/virtio/v1.1/cs01/virtio-v1.1-cs01.html#x1-210003 >> >> Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com> >> >> --- >> Note: As we don't support legacy virtio devices there is dead code in >> libvhost-access.h that could be removed. But for the sake of >> completeness, I left it in the code. > > Please remove the dead code. It is unlikely that legacy device support > will be required in the future and it will just confuse people reading > the code. Done. > >> --- >> contrib/libvhost-user/libvhost-access.h | 61 ++++++++++++ >> contrib/libvhost-user/libvhost-user.c | 119 ++++++++++++------------ >> 2 files changed, 121 insertions(+), 59 deletions(-) >> create mode 100644 contrib/libvhost-user/libvhost-access.h >> >> diff --git a/contrib/libvhost-user/libvhost-access.h b/contrib/libvhost-user/libvhost-access.h >> new file mode 100644 >> index 000000000000..868ba3e7bfb8 >> --- /dev/null >> +++ b/contrib/libvhost-user/libvhost-access.h >> @@ -0,0 +1,61 @@ >> +#ifndef LIBVHOST_ACCESS_H > > License/copyright header? With the dead code removed there is no reason for having libvhost-access.h so I’ve removed it.
diff --git a/contrib/libvhost-user/libvhost-access.h b/contrib/libvhost-user/libvhost-access.h new file mode 100644 index 000000000000..868ba3e7bfb8 --- /dev/null +++ b/contrib/libvhost-user/libvhost-access.h @@ -0,0 +1,61 @@ +#ifndef LIBVHOST_ACCESS_H + +#include "qemu/bswap.h" + +#include "libvhost-user.h" + +static inline bool vu_access_is_big_endian(VuDev *dev) +{ + /* Devices conforming to VIRTIO 1.0 or later are always LE. */ + return false; +} + +static inline void vu_stw_p(VuDev *vdev, void *ptr, uint16_t v) +{ + if (vu_access_is_big_endian(vdev)) { + stw_be_p(ptr, v); + } else { + stw_le_p(ptr, v); + } +} + +static inline void vu_stl_p(VuDev *vdev, void *ptr, uint32_t v) +{ + if (vu_access_is_big_endian(vdev)) { + stl_be_p(ptr, v); + } else { + stl_le_p(ptr, v); + } +} + +static inline void vu_stq_p(VuDev *vdev, void *ptr, uint64_t v) +{ + if (vu_access_is_big_endian(vdev)) { + stq_be_p(ptr, v); + } else { + stq_le_p(ptr, v); + } +} + +static inline int vu_lduw_p(VuDev *vdev, const void *ptr) +{ + if (vu_access_is_big_endian(vdev)) + return lduw_be_p(ptr); + return lduw_le_p(ptr); +} + +static inline int vu_ldl_p(VuDev *vdev, const void *ptr) +{ + if (vu_access_is_big_endian(vdev)) + return ldl_be_p(ptr); + return ldl_le_p(ptr); +} + +static inline uint64_t vu_ldq_p(VuDev *vdev, const void *ptr) +{ + if (vu_access_is_big_endian(vdev)) + return ldq_be_p(ptr); + return ldq_le_p(ptr); +} + +#endif /* LIBVHOST_ACCESS_H */ diff --git a/contrib/libvhost-user/libvhost-user.c b/contrib/libvhost-user/libvhost-user.c index d315db139606..0214b04c5291 100644 --- a/contrib/libvhost-user/libvhost-user.c +++ b/contrib/libvhost-user/libvhost-user.c @@ -45,6 +45,7 @@ #include "qemu/memfd.h" #include "libvhost-user.h" +#include "libvhost-access.h" /* usually provided by GLib */ #ifndef MIN @@ -1074,7 +1075,7 @@ vu_set_vring_addr_exec(VuDev *dev, VhostUserMsg *vmsg) return false; } - vq->used_idx = vq->vring.used->idx; + vq->used_idx = vu_lduw_p(dev, &vq->vring.used->idx); if (vq->last_avail_idx != vq->used_idx) { bool resume = dev->iface->queue_is_processed_in_order && @@ -1191,7 +1192,7 @@ vu_check_queue_inflights(VuDev *dev, VuVirtq *vq) return 0; } - vq->used_idx = vq->vring.used->idx; + vq->used_idx = vu_lduw_p(dev, &vq->vring.used->idx); vq->resubmit_num = 0; vq->resubmit_list = NULL; vq->counter = 0; @@ -2019,35 +2020,35 @@ vu_queue_started(const VuDev *dev, const VuVirtq *vq) } static inline uint16_t -vring_avail_flags(VuVirtq *vq) +vring_avail_flags(VuDev *dev, VuVirtq *vq) { - return vq->vring.avail->flags; + return vu_lduw_p(dev, &vq->vring.avail->flags); } static inline uint16_t -vring_avail_idx(VuVirtq *vq) +vring_avail_idx(VuDev *dev, VuVirtq *vq) { - vq->shadow_avail_idx = vq->vring.avail->idx; + vq->shadow_avail_idx = vu_lduw_p(dev, &vq->vring.avail->idx); return vq->shadow_avail_idx; } static inline uint16_t -vring_avail_ring(VuVirtq *vq, int i) +vring_avail_ring(VuDev *dev, VuVirtq *vq, int i) { - return vq->vring.avail->ring[i]; + return vu_lduw_p(dev, &vq->vring.avail->ring[i]); } static inline uint16_t -vring_get_used_event(VuVirtq *vq) +vring_get_used_event(VuDev *dev, VuVirtq *vq) { - return vring_avail_ring(vq, vq->vring.num); + return vring_avail_ring(dev, vq, vq->vring.num); } static int virtqueue_num_heads(VuDev *dev, VuVirtq *vq, unsigned int idx) { - uint16_t num_heads = vring_avail_idx(vq) - idx; + uint16_t num_heads = vring_avail_idx(dev, vq) - idx; /* Check it isn't doing very strange things with descriptor numbers. */ if (num_heads > vq->vring.num) { @@ -2070,7 +2071,7 @@ virtqueue_get_head(VuDev *dev, VuVirtq *vq, { /* Grab the next descriptor number they're advertising, and increment * the index we've seen. */ - *head = vring_avail_ring(vq, idx % vq->vring.num); + *head = vring_avail_ring(dev, vq, idx % vq->vring.num); /* If their number is silly, that's a fatal mistake. */ if (*head >= vq->vring.num) { @@ -2123,12 +2124,12 @@ virtqueue_read_next_desc(VuDev *dev, struct vring_desc *desc, int i, unsigned int max, unsigned int *next) { /* If this descriptor says it doesn't chain, we're done. */ - if (!(desc[i].flags & VRING_DESC_F_NEXT)) { + if (!(vu_lduw_p(dev, &desc[i].flags) & VRING_DESC_F_NEXT)) { return VIRTQUEUE_READ_DESC_DONE; } /* Check they're not leading us off end of descriptors. */ - *next = desc[i].next; + *next = vu_lduw_p(dev, &desc[i].next); /* Make sure compiler knows to grab that: we don't want it changing! */ smp_wmb(); @@ -2171,8 +2172,8 @@ vu_queue_get_avail_bytes(VuDev *dev, VuVirtq *vq, unsigned int *in_bytes, } desc = vq->vring.desc; - if (desc[i].flags & VRING_DESC_F_INDIRECT) { - if (desc[i].len % sizeof(struct vring_desc)) { + if (vu_lduw_p(dev, &desc[i].flags) & VRING_DESC_F_INDIRECT) { + if (vu_ldl_p(dev, &desc[i].len) % sizeof(struct vring_desc)) { vu_panic(dev, "Invalid size for indirect buffer table"); goto err; } @@ -2185,8 +2186,8 @@ vu_queue_get_avail_bytes(VuDev *dev, VuVirtq *vq, unsigned int *in_bytes, /* loop over the indirect descriptor table */ indirect = 1; - desc_addr = desc[i].addr; - desc_len = desc[i].len; + desc_addr = vu_ldq_p(dev, &desc[i].addr); + desc_len = vu_ldl_p(dev, &desc[i].len); max = desc_len / sizeof(struct vring_desc); read_len = desc_len; desc = vu_gpa_to_va(dev, &read_len, desc_addr); @@ -2213,10 +2214,10 @@ vu_queue_get_avail_bytes(VuDev *dev, VuVirtq *vq, unsigned int *in_bytes, goto err; } - if (desc[i].flags & VRING_DESC_F_WRITE) { - in_total += desc[i].len; + if (vu_lduw_p(dev, &desc[i].flags) & VRING_DESC_F_WRITE) { + in_total += vu_ldl_p(dev, &desc[i].len); } else { - out_total += desc[i].len; + out_total += vu_ldl_p(dev, &desc[i].len); } if (in_total >= max_in_bytes && out_total >= max_out_bytes) { goto done; @@ -2277,7 +2278,7 @@ vu_queue_empty(VuDev *dev, VuVirtq *vq) return false; } - return vring_avail_idx(vq) == vq->last_avail_idx; + return vring_avail_idx(dev, vq) == vq->last_avail_idx; } static bool @@ -2296,14 +2297,14 @@ vring_notify(VuDev *dev, VuVirtq *vq) } if (!vu_has_feature(dev, VIRTIO_RING_F_EVENT_IDX)) { - return !(vring_avail_flags(vq) & VRING_AVAIL_F_NO_INTERRUPT); + return !(vring_avail_flags(dev, vq) & VRING_AVAIL_F_NO_INTERRUPT); } v = vq->signalled_used_valid; vq->signalled_used_valid = true; old = vq->signalled_used; new = vq->signalled_used = vq->used_idx; - return !v || vring_need_event(vring_get_used_event(vq), new, old); + return !v || vring_need_event(vring_get_used_event(dev, vq), new, old); } static void _vu_queue_notify(VuDev *dev, VuVirtq *vq, bool sync) @@ -2361,33 +2362,33 @@ void vu_queue_notify_sync(VuDev *dev, VuVirtq *vq) } static inline void -vring_used_flags_set_bit(VuVirtq *vq, int mask) +vring_used_flags_set_bit(VuDev *dev, VuVirtq *vq, int mask) +{ + uint16_t *flags; + + flags = (uint16_t *)(char*)vq->vring.used + + offsetof(struct vring_used, flags); + vu_stw_p(dev, flags, vu_lduw_p(dev, flags) | mask); +} + +static inline void +vring_used_flags_unset_bit(VuDev *dev, VuVirtq *vq, int mask) { uint16_t *flags; flags = (uint16_t *)((char*)vq->vring.used + offsetof(struct vring_used, flags)); - *flags |= mask; + vu_stw_p(dev, flags, vu_lduw_p(dev, flags) & ~mask); } static inline void -vring_used_flags_unset_bit(VuVirtq *vq, int mask) -{ - uint16_t *flags; - - flags = (uint16_t *)((char*)vq->vring.used + - offsetof(struct vring_used, flags)); - *flags &= ~mask; -} - -static inline void -vring_set_avail_event(VuVirtq *vq, uint16_t val) +vring_set_avail_event(VuDev *dev, VuVirtq *vq, uint16_t val) { if (!vq->notification) { return; } - *((uint16_t *) &vq->vring.used->ring[vq->vring.num]) = val; + vu_stw_p(dev, &vq->vring.used->ring[vq->vring.num], val); } void @@ -2395,11 +2396,11 @@ vu_queue_set_notification(VuDev *dev, VuVirtq *vq, int enable) { vq->notification = enable; if (vu_has_feature(dev, VIRTIO_RING_F_EVENT_IDX)) { - vring_set_avail_event(vq, vring_avail_idx(vq)); + vring_set_avail_event(dev, vq, vring_avail_idx(dev, vq)); } else if (enable) { - vring_used_flags_unset_bit(vq, VRING_USED_F_NO_NOTIFY); + vring_used_flags_unset_bit(dev, vq, VRING_USED_F_NO_NOTIFY); } else { - vring_used_flags_set_bit(vq, VRING_USED_F_NO_NOTIFY); + vring_used_flags_set_bit(dev, vq, VRING_USED_F_NO_NOTIFY); } if (enable) { /* Expose avail event/used flags before caller checks the avail idx. */ @@ -2476,14 +2477,14 @@ vu_queue_map_desc(VuDev *dev, VuVirtq *vq, unsigned int idx, size_t sz) struct vring_desc desc_buf[VIRTQUEUE_MAX_SIZE]; int rc; - if (desc[i].flags & VRING_DESC_F_INDIRECT) { - if (desc[i].len % sizeof(struct vring_desc)) { + if (vu_lduw_p(dev, &desc[i].flags) & VRING_DESC_F_INDIRECT) { + if (vu_ldl_p(dev, &desc[i].len) % sizeof(struct vring_desc)) { vu_panic(dev, "Invalid size for indirect buffer table"); } /* loop over the indirect descriptor table */ - desc_addr = desc[i].addr; - desc_len = desc[i].len; + desc_addr = vu_ldq_p(dev, &desc[i].addr); + desc_len = vu_ldl_p(dev, &desc[i].len); max = desc_len / sizeof(struct vring_desc); read_len = desc_len; desc = vu_gpa_to_va(dev, &read_len, desc_addr); @@ -2505,10 +2506,10 @@ vu_queue_map_desc(VuDev *dev, VuVirtq *vq, unsigned int idx, size_t sz) /* Collect all the descriptors */ do { - if (desc[i].flags & VRING_DESC_F_WRITE) { + if (vu_lduw_p(dev, &desc[i].flags) & VRING_DESC_F_WRITE) { virtqueue_map_desc(dev, &in_num, iov + out_num, VIRTQUEUE_MAX_SIZE - out_num, true, - desc[i].addr, desc[i].len); + vu_ldq_p(dev, &desc[i].addr), vu_ldl_p(dev, &desc[i].len)); } else { if (in_num) { vu_panic(dev, "Incorrect order for descriptors"); @@ -2516,7 +2517,7 @@ vu_queue_map_desc(VuDev *dev, VuVirtq *vq, unsigned int idx, size_t sz) } virtqueue_map_desc(dev, &out_num, iov, VIRTQUEUE_MAX_SIZE, false, - desc[i].addr, desc[i].len); + vu_ldq_p(dev, &desc[i].addr), vu_ldl_p(dev, &desc[i].len)); } /* If we've got too many, that implies a descriptor loop. */ @@ -2642,7 +2643,7 @@ vu_queue_pop(VuDev *dev, VuVirtq *vq, size_t sz) } if (vu_has_feature(dev, VIRTIO_RING_F_EVENT_IDX)) { - vring_set_avail_event(vq, vq->last_avail_idx); + vring_set_avail_event(dev, vq, vq->last_avail_idx); } elem = vu_queue_map_desc(dev, vq, head, sz); @@ -2712,14 +2713,14 @@ vu_log_queue_fill(VuDev *dev, VuVirtq *vq, max = vq->vring.num; i = elem->index; - if (desc[i].flags & VRING_DESC_F_INDIRECT) { - if (desc[i].len % sizeof(struct vring_desc)) { + if (vu_lduw_p(dev, &desc[i].flags) & VRING_DESC_F_INDIRECT) { + if (vu_ldl_p(dev, &desc[i].len) % sizeof(struct vring_desc)) { vu_panic(dev, "Invalid size for indirect buffer table"); } /* loop over the indirect descriptor table */ - desc_addr = desc[i].addr; - desc_len = desc[i].len; + desc_addr = vu_ldq_p(dev, &desc[i].addr); + desc_len = vu_ldl_p(dev, &desc[i].len); max = desc_len / sizeof(struct vring_desc); read_len = desc_len; desc = vu_gpa_to_va(dev, &read_len, desc_addr); @@ -2745,9 +2746,9 @@ vu_log_queue_fill(VuDev *dev, VuVirtq *vq, return; } - if (desc[i].flags & VRING_DESC_F_WRITE) { - min = MIN(desc[i].len, len); - vu_log_write(dev, desc[i].addr, min); + if (vu_lduw_p(dev, &desc[i].flags) & VRING_DESC_F_WRITE) { + min = MIN(vu_ldl_p(dev, &desc[i].len), len); + vu_log_write(dev, vu_ldq_p(dev, &desc[i].addr), min); len -= min; } @@ -2772,15 +2773,15 @@ vu_queue_fill(VuDev *dev, VuVirtq *vq, idx = (idx + vq->used_idx) % vq->vring.num; - uelem.id = elem->index; - uelem.len = len; + vu_stl_p(dev, &uelem.id, elem->index); + vu_stl_p(dev, &uelem.len, len); vring_used_write(dev, vq, &uelem, idx); } static inline void vring_used_idx_set(VuDev *dev, VuVirtq *vq, uint16_t val) { - vq->vring.used->idx = val; + vu_stw_p(dev, &vq->vring.used->idx, val); vu_log_write(dev, vq->vring.log_guest_addr + offsetof(struct vring_used, idx), sizeof(vq->vring.used->idx));
Since virtio existed even before it got standardized, the virtio standard defines the following types of virtio devices: + legacy device (pre-virtio 1.0) + non-legacy or VIRTIO 1.0 device + transitional device (which can act both as legacy and non-legacy) Virtio 1.0 defines the fields of the virtqueues as little endian, while legacy uses guest's native endian [1]. Currently libvhost-user does not handle virtio endianness at all, i.e. it works only if the native endianness matches with whatever is actually needed. That means things break spectacularly on big-endian targets. Let us handle virtio endianness for non-legacy as required by the virtio specification [1]. We will fence non-legacy virtio devices with the upcoming patch. [1] https://docs.oasis-open.org/virtio/virtio/v1.1/cs01/virtio-v1.1-cs01.html#x1-210003 Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com> --- Note: As we don't support legacy virtio devices there is dead code in libvhost-access.h that could be removed. But for the sake of completeness, I left it in the code. --- contrib/libvhost-user/libvhost-access.h | 61 ++++++++++++ contrib/libvhost-user/libvhost-user.c | 119 ++++++++++++------------ 2 files changed, 121 insertions(+), 59 deletions(-) create mode 100644 contrib/libvhost-user/libvhost-access.h