This adds a mac= option to crosvm's --tap-fd option. The virtio-net driver in the guest will read the desired MAC from virtio configuration space. See the documentation for VIRTIO_NET_F_MAC in the Virtio spec[1]. [1]: https://docs.oasis-open.org/virtio/virtio/v1.1/virtio-v1.1.html Thanks-to: Puck Meerburg <puck@puckipedia.com> --- v1: https://spectrum-os.org/lists/archives/spectrum-devel/20210409222026.13886-1-hi@alyssa.is/ This is a new version of the patch, updated for a much more recent crosvm. I'm posting it again because I had to change it enough that I feel it could use a re-review. The biggest change is that I gave up on a NetOptions struct, since upstream has added a new option, and it seems that their approach is going to be just passing ever more function arguments around. No use fighting that for my single addition. Also, change from RawFd to crosvm's RawDescriptor. It would be nice if we could use SafeDescriptor earlier here (which represents fd ownership), but I don't think that's really easily doable. devices/src/virtio/net.rs | 12 ++++++-- src/crosvm.rs | 8 +++-- src/linux.rs | 12 ++++---- src/main.rs | 63 ++++++++++++++++++++++++++++++--------- 4 files changed, 71 insertions(+), 24 deletions(-) diff --git a/devices/src/virtio/net.rs b/devices/src/virtio/net.rs index 92368c440..4d0ea1560 100644 --- a/devices/src/virtio/net.rs +++ b/devices/src/virtio/net.rs @@ -419,6 +419,7 @@ where } pub struct Net<T: TapT> { + mac_address: Option<MacAddress>, queue_sizes: Box<[u16]>, workers_kill_evt: Vec<Event>, kill_evts: Vec<Event>, @@ -439,6 +440,7 @@ where ip_addr: Ipv4Addr, netmask: Ipv4Addr, mac_addr: MacAddress, + guest_mac_addr: Option<MacAddress>, vq_pairs: u16, ) -> Result<Net<T>, NetError> { let multi_queue = if vq_pairs > 1 { true } else { false }; @@ -450,12 +452,12 @@ where tap.enable().map_err(NetError::TapEnable)?; - Net::from(base_features, tap, vq_pairs) + Net::with_tap(base_features, tap, vq_pairs, guest_mac_addr) } /// Creates a new virtio network device from a tap device that has already been /// configured. - pub fn from(base_features: u64, tap: T, vq_pairs: u16) -> Result<Net<T>, NetError> { + pub fn with_tap(base_features: u64, tap: T, vq_pairs: u16, mac_address: Option<MacAddress>) -> Result<Net<T>, NetError> { let taps = tap.into_mq_taps(vq_pairs).map_err(NetError::TapOpen)?; // This would also validate a tap created by Self::new(), but that's a good thing as it @@ -488,7 +490,12 @@ where workers_kill_evt.push(worker_kill_evt); } + if mac_address.is_some() { + avail_features |= 1 << virtio_net::VIRTIO_NET_F_MAC; + } + Ok(Net { + mac_address, queue_sizes: vec![QUEUE_SIZE; (vq_pairs * 2 + 1) as usize].into_boxed_slice(), workers_kill_evt, kill_evts, @@ -503,6 +510,7 @@ where let vq_pairs = self.queue_sizes.len() as u16 / 2; VirtioNetConfig { + mac: self.mac_address.map(|m| m.octets()).unwrap_or_else(Default::default), max_vq_pairs: Le16::from(vq_pairs), // Other field has meaningful value when the corresponding feature // is enabled, but all these features aren't supported now. diff --git a/src/crosvm.rs b/src/crosvm.rs index 04e267d6c..1dde7d769 100644 --- a/src/crosvm.rs +++ b/src/crosvm.rs @@ -171,6 +171,10 @@ impl Default for SharedDir { } } +pub struct TapFdOptions { + pub mac: Option<net_util::MacAddress>, +} + /// Aggregate of all configurable options for a running VM. pub struct Config { pub vcpu_count: Option<usize>, @@ -194,7 +198,7 @@ pub struct Config { pub mac_address: Option<net_util::MacAddress>, pub net_vq_pairs: Option<u16>, pub vhost_net: bool, - pub tap_fd: Vec<RawFd>, + pub tap_fd: BTreeMap<RawFd, TapFdOptions>, pub cid: Option<u64>, pub wayland_socket_paths: BTreeMap<String, PathBuf>, pub wayland_dmabuf: bool, @@ -253,7 +257,7 @@ impl Default for Config { mac_address: None, net_vq_pairs: None, vhost_net: false, - tap_fd: Vec::new(), + tap_fd: BTreeMap::new(), cid: None, #[cfg(feature = "gpu")] gpu_parameters: None, diff --git a/src/linux.rs b/src/linux.rs index f452fef3f..42c07df4f 100644 --- a/src/linux.rs +++ b/src/linux.rs @@ -75,7 +75,7 @@ use vm_memory::{GuestAddress, GuestMemory}; #[cfg(all(target_arch = "x86_64", feature = "gdb"))] use crate::gdb::{gdb_thread, GdbStub}; -use crate::{Config, DiskOption, Executable, SharedDir, SharedDirKind, TouchDeviceOption}; +use crate::{Config, DiskOption, Executable, SharedDir, SharedDirKind, TapFdOptions, TouchDeviceOption}; use arch::{ self, LinuxArch, RunnableLinuxVm, SerialHardware, SerialParameters, VcpuAffinity, VirtioDeviceStub, VmComponents, VmImage, @@ -675,7 +675,7 @@ fn create_balloon_device(cfg: &Config, socket: BalloonControlResponseSocket) -> }) } -fn create_tap_net_device(cfg: &Config, tap_fd: RawDescriptor) -> DeviceResult { +fn create_tap_net_device(cfg: &Config, tap_fd: RawDescriptor, options: &TapFdOptions) -> DeviceResult { // Safe because we ensure that we get a unique handle to the fd. let tap = unsafe { Tap::from_raw_descriptor( @@ -691,7 +691,7 @@ fn create_tap_net_device(cfg: &Config, tap_fd: RawDescriptor) -> DeviceResult { vq_pairs = 1; } let features = virtio::base_features(cfg.protected_vm); - let dev = virtio::Net::from(features, tap, vq_pairs).map_err(Error::NetDeviceNew)?; + let dev = virtio::Net::with_tap(features, tap, vq_pairs, options.mac).map_err(Error::NetDeviceNew)?; Ok(VirtioDeviceStub { dev: Box::new(dev), @@ -725,7 +725,7 @@ fn create_net_device( .map_err(Error::VhostNetDeviceNew)?; Box::new(dev) as Box<dyn VirtioDevice> } else { - let dev = virtio::Net::<Tap>::new(features, host_ip, netmask, mac_address, vq_pairs) + let dev = virtio::Net::<Tap>::new(features, host_ip, netmask, mac_address, None, vq_pairs) .map_err(Error::NetDeviceNew)?; Box::new(dev) as Box<dyn VirtioDevice> }; @@ -1311,8 +1311,8 @@ fn create_virtio_devices( devs.push(create_balloon_device(cfg, balloon_device_socket)?); // We checked above that if the IP is defined, then the netmask is, too. - for tap_fd in &cfg.tap_fd { - devs.push(create_tap_net_device(cfg, *tap_fd)?); + for (tap_fd, options) in &cfg.tap_fd { + devs.push(create_tap_net_device(cfg, *tap_fd, options)?); } if let (Some(host_ip), Some(netmask), Some(mac_address)) = diff --git a/src/main.rs b/src/main.rs index 5d02af02f..f8bc0d14e 100644 --- a/src/main.rs +++ b/src/main.rs @@ -28,7 +28,7 @@ use base::{ }; use crosvm::{ argument::{self, print_help, set_arguments, Argument}, - platform, BindMount, Config, DiskOption, Executable, GidMap, SharedDir, TouchDeviceOption, + platform, BindMount, Config, DiskOption, Executable, GidMap, SharedDir, TapFdOptions, TouchDeviceOption, DISK_ID_LEN, }; #[cfg(feature = "gpu")] @@ -1319,17 +1319,52 @@ fn set_argument(cfg: &mut Config, name: &str, value: Option<&str>) -> argument:: } "vhost-net" => cfg.vhost_net = true, "tap-fd" => { - cfg.tap_fd.push( - value - .unwrap() - .parse() - .map_err(|_| argument::Error::InvalidValue { - value: value.unwrap().to_owned(), - expected: String::from( - "this value for `tap-fd` must be an unsigned integer", - ), - })?, - ); + let mut components = value.unwrap().split(','); + + let fd: RawDescriptor = components + .next() + .and_then(|x| x.parse().ok()) + .ok_or_else(|| argument::Error::InvalidValue { + value: value.unwrap().to_owned(), + expected: String::from("this value for `tap-fd` must be an unsigned integer"), + })?; + + let mut mac = None; + for c in components { + let mut kv = c.splitn(2, '='); + let (kind, value) = match (kv.next(), kv.next()) { + (Some(kind), Some(value)) => (kind, value), + _ => { + return Err(argument::Error::InvalidValue { + value: c.to_owned(), + expected: String::from("option must be of the form `kind=value`"), + }) + } + }; + match kind { + "mac" => { + mac = Some(value.parse().map_err(|_| argument::Error::InvalidValue { + value: value.to_owned(), + expected: String::from( + "`mac` needs to be in the form \"XX:XX:XX:XX:XX:XX\"", + ), + })?) + } + _ => { + return Err(argument::Error::InvalidValue { + value: kind.to_owned(), + expected: String::from("unrecognized option"), + }) + } + } + } + if cfg.tap_fd.contains_key(&fd) { + return Err(argument::Error::TooManyArguments(format!( + "TAP FD already used: '{}'", + name + ))); + } + cfg.tap_fd.insert(fd, TapFdOptions { mac }); } #[cfg(feature = "gpu")] "gpu" => { @@ -1644,8 +1679,8 @@ writeback=BOOL - Indicates whether the VM can use writeback caching (default: fa Argument::value("plugin-gid-map-file", "PATH", "Path to the file listing supplemental GIDs that should be mapped in plugin jail. Can be given more than once."), Argument::flag("vhost-net", "Use vhost for networking."), Argument::value("tap-fd", - "fd", - "File descriptor for configured tap device. A different virtual network card will be added each time this argument is given."), + "FD[,mac=MAC]", + "File descriptor for configured tap device. A different virtual network card will be added each time this argument is given. MAC is the MAC address that will be set in the guest."), #[cfg(feature = "gpu")] Argument::flag_or_value("gpu", "[width=INT,height=INT]", base-commit: f35d2c43ff19520855cffee761dc8899c5a439a1 -- 2.31.1
On Mon May 17, 2021 at 11:57 AM PDT, Alyssa Ross wrote: > This adds a mac= option to crosvm's --tap-fd option. The virtio-net > driver in the guest will read the desired MAC from virtio > configuration space. > > See the documentation for VIRTIO_NET_F_MAC in the Virtio spec[1]. > > [1]: https://docs.oasis-open.org/virtio/virtio/v1.1/virtio-v1.1.html > > Thanks-to: Puck Meerburg <puck@puckipedia.com> > --- > devices/src/virtio/net.rs | 12 ++++++-- > src/crosvm.rs | 8 +++-- > src/linux.rs | 12 ++++---- > src/main.rs | 63 ++++++++++++++++++++++++++++++--------- > 4 files changed, 71 insertions(+), 24 deletions(-) [snip] > diff --git a/src/main.rs b/src/main.rs > index 5d02af02f..f8bc0d14e 100644 > --- a/src/main.rs > +++ b/src/main.rs > @@ -28,7 +28,7 @@ use base::{ > }; > use crosvm::{ > argument::{self, print_help, set_arguments, Argument}, > - platform, BindMount, Config, DiskOption, Executable, GidMap, SharedDir, TouchDeviceOption, > + platform, BindMount, Config, DiskOption, Executable, GidMap, SharedDir, TapFdOptions, TouchDeviceOption, > DISK_ID_LEN, > }; > #[cfg(feature = "gpu")] > @@ -1319,17 +1319,52 @@ fn set_argument(cfg: &mut Config, name: &str, value: Option<&str>) -> argument:: > } > "vhost-net" => cfg.vhost_net = true, > "tap-fd" => { > - cfg.tap_fd.push( > - value > - .unwrap() > - .parse() > - .map_err(|_| argument::Error::InvalidValue { > - value: value.unwrap().to_owned(), > - expected: String::from( > - "this value for `tap-fd` must be an unsigned integer", > - ), > - })?, > - ); > + let mut components = value.unwrap().split(','); > + > + let fd: RawDescriptor = components > + .next() > + .and_then(|x| x.parse().ok()) > + .ok_or_else(|| argument::Error::InvalidValue { > + value: value.unwrap().to_owned(), > + expected: String::from("this value for `tap-fd` must be an unsigned integer"), > + })?; > + > + let mut mac = None; > + for c in components { > + let mut kv = c.splitn(2, '='); > + let (kind, value) = match (kv.next(), kv.next()) { > + (Some(kind), Some(value)) => (kind, value), > + _ => { > + return Err(argument::Error::InvalidValue { > + value: c.to_owned(), > + expected: String::from("option must be of the form `kind=value`"), > + }) > + } > + }; > + match kind { > + "mac" => { > + mac = Some(value.parse().map_err(|_| argument::Error::InvalidValue { > + value: value.to_owned(), > + expected: String::from( > + "`mac` needs to be in the form \"XX:XX:XX:XX:XX:XX\"", > + ), > + })?) > + } > + _ => { > + return Err(argument::Error::InvalidValue { > + value: kind.to_owned(), > + expected: String::from("unrecognized option"), > + }) > + } > + } > + } > + if cfg.tap_fd.contains_key(&fd) { > + return Err(argument::Error::TooManyArguments(format!( Is there a better Error variant for this? `TooManyArguments` seems not- completely-accurate when specifying an already-in-use FD. > + "TAP FD already used: '{}'", > + name > + ))); > + } > + cfg.tap_fd.insert(fd, TapFdOptions { mac }); > } > #[cfg(feature = "gpu")] > "gpu" => { > @@ -1644,8 +1679,8 @@ writeback=BOOL - Indicates whether the VM can use writeback caching (default: fa > Argument::value("plugin-gid-map-file", "PATH", "Path to the file listing supplemental GIDs that should be mapped in plugin jail. Can be given more than once."), > Argument::flag("vhost-net", "Use vhost for networking."), > Argument::value("tap-fd", > - "fd", > - "File descriptor for configured tap device. A different virtual network card will be added each time this argument is given."), > + "FD[,mac=MAC]", > + "File descriptor for configured tap device. A different virtual network card will be added each time this argument is given. MAC is the MAC address that will be set in the guest."), > #[cfg(feature = "gpu")] > Argument::flag_or_value("gpu", > "[width=INT,height=INT]", > > base-commit: f35d2c43ff19520855cffee761dc8899c5a439a1 > -- > 2.31.1 Else, LGTM. Reviewed-by: Cole Helbling <cole.e.helbling@outlook.com>
[-- Attachment #1: Type: text/plain, Size: 3106 bytes --] "Cole Helbling" <cole.e.helbling@outlook.com> writes: >> @@ -1319,17 +1319,52 @@ fn set_argument(cfg: &mut Config, name: &str, value: Option<&str>) -> argument:: >> } >> "vhost-net" => cfg.vhost_net = true, >> "tap-fd" => { >> - cfg.tap_fd.push( >> - value >> - .unwrap() >> - .parse() >> - .map_err(|_| argument::Error::InvalidValue { >> - value: value.unwrap().to_owned(), >> - expected: String::from( >> - "this value for `tap-fd` must be an unsigned integer", >> - ), >> - })?, >> - ); >> + let mut components = value.unwrap().split(','); >> + >> + let fd: RawDescriptor = components >> + .next() >> + .and_then(|x| x.parse().ok()) >> + .ok_or_else(|| argument::Error::InvalidValue { >> + value: value.unwrap().to_owned(), >> + expected: String::from("this value for `tap-fd` must be an unsigned integer"), >> + })?; >> + >> + let mut mac = None; >> + for c in components { >> + let mut kv = c.splitn(2, '='); >> + let (kind, value) = match (kv.next(), kv.next()) { >> + (Some(kind), Some(value)) => (kind, value), >> + _ => { >> + return Err(argument::Error::InvalidValue { >> + value: c.to_owned(), >> + expected: String::from("option must be of the form `kind=value`"), >> + }) >> + } >> + }; >> + match kind { >> + "mac" => { >> + mac = Some(value.parse().map_err(|_| argument::Error::InvalidValue { >> + value: value.to_owned(), >> + expected: String::from( >> + "`mac` needs to be in the form \"XX:XX:XX:XX:XX:XX\"", >> + ), >> + })?) >> + } >> + _ => { >> + return Err(argument::Error::InvalidValue { >> + value: kind.to_owned(), >> + expected: String::from("unrecognized option"), >> + }) >> + } >> + } >> + } >> + if cfg.tap_fd.contains_key(&fd) { >> + return Err(argument::Error::TooManyArguments(format!( > > Is there a better Error variant for this? `TooManyArguments` seems not- > completely-accurate when specifying an already-in-use FD. Here are all the variants. Do you think there's a better one? https://chromium.googlesource.com/chromiumos/platform/crosvm/+/f35d2c43ff19520855cffee761dc8899c5a439a1/src/argument.rs#49 The only other one I can see that might be applicable would be InvalidValue... [-- Attachment #2: signature.asc --] [-- Type: application/pgp-signature, Size: 832 bytes --]
On Tue May 18, 2021 at 1:36 AM PDT, Alyssa Ross wrote:
> "Cole Helbling" <cole.e.helbling@outlook.com> writes:
> > Is there a better Error variant for this? `TooManyArguments` seems not-
> > completely-accurate when specifying an already-in-use FD.
>
> Here are all the variants. Do you think there's a better one?
>
> https://chromium.googlesource.com/chromiumos/platform/crosvm/+/f35d2c43ff19520855cffee761dc8899c5a439a1/src/argument.rs#49
>
> The only other one I can see that might be applicable would be
> InvalidValue...
Yeah, I think InvalidValue would be a better choice (IMHO). The user hasn't
provided too many arguments; they just provided a invalid TAP FD because it's
already in use.
[-- Attachment #1: Type: text/plain, Size: 1141 bytes --] "Cole Helbling" <cole.e.helbling@outlook.com> writes: > On Tue May 18, 2021 at 1:36 AM PDT, Alyssa Ross wrote: >> "Cole Helbling" <cole.e.helbling@outlook.com> writes: >> > Is there a better Error variant for this? `TooManyArguments` seems not- >> > completely-accurate when specifying an already-in-use FD. >> >> Here are all the variants. Do you think there's a better one? >> >> https://chromium.googlesource.com/chromiumos/platform/crosvm/+/f35d2c43ff19520855cffee761dc8899c5a439a1/src/argument.rs#49 >> >> The only other one I can see that might be applicable would be >> InvalidValue... > > Yeah, I think InvalidValue would be a better choice (IMHO). The user hasn't > provided too many arguments; they just provided a invalid TAP FD because it's > already in use. Hmm, I agree with you, but I just went to make this change, and noticed that everywhere else in the file, they use TooManyArguments for this sort of case. e.g. "wayland socket name already used". So I think it's probably best to stick with TooManyArguments for consistency (and that's probably why I did it like that in the first place, although I don't remember). [-- Attachment #2: signature.asc --] [-- Type: application/pgp-signature, Size: 832 bytes --]
On Sat May 29, 2021 at 7:08 AM PDT, Alyssa Ross wrote:
> "Cole Helbling" <cole.e.helbling@outlook.com> writes:
>
> > On Tue May 18, 2021 at 1:36 AM PDT, Alyssa Ross wrote:
> >> "Cole Helbling" <cole.e.helbling@outlook.com> writes:
> >> > Is there a better Error variant for this? `TooManyArguments` seems not-
> >> > completely-accurate when specifying an already-in-use FD.
> >>
> >> Here are all the variants. Do you think there's a better one?
> >>
> >> https://chromium.googlesource.com/chromiumos/platform/crosvm/+/f35d2c43ff19520855cffee761dc8899c5a439a1/src/argument.rs#49
> >>
> >> The only other one I can see that might be applicable would be
> >> InvalidValue...
> >
> > Yeah, I think InvalidValue would be a better choice (IMHO). The user hasn't
> > provided too many arguments; they just provided a invalid TAP FD because it's
> > already in use.
>
> Hmm, I agree with you, but I just went to make this change, and noticed
> that everywhere else in the file, they use TooManyArguments for this
> sort of case. e.g. "wayland socket name already used". So I think it's
> probably best to stick with TooManyArguments for consistency (and that's
> probably why I did it like that in the first place, although I don't
> remember).
OK, that's fair.
Reviewed-by: Cole Helbling <cole.e.helbling@outlook.com>