struct zbi_debugdata_t

ZBI_TYPE_DEBUGDATA

This provides a data dump and associated logging from a boot loader,

shim, or earlier incarnation that wants its data percolated up by the

booting Zircon kernel. This is usually used for instrumented builds.

It's intended to provide the same information and formats that might be

published via the FIDL fuchsia.debugdata.Publisher protocol.

The payload is described by a trailer rather than a header, with the

dump contents themselves first in the payload. This is so that when the

whole payload is transferred to a VMO, the contents can be used there

and sent to fuchsia.debugdata.Publisher in place just by adjusting the

VMO's ZX_PROP_VMO_CONTENT_SIZE property.

The payload is organized into four chunks, followed by the trailer. The

whole payload size is always aligned to ZBI_ALIGNMENT bytes. The

zbi_debugdata_t trailer is read from the final bytes of the payload.

Each chunk has an arbitrary size that might not be aligned, and there is

no alignment padding between the chunks. The total size of all the

chunks together is padded to ZBI_ALIGNMENT, with the trailer at the end.

The sizes in the zbi_debugdata_t trailer must add up, rounded to alignment,

to exactly sizeof(zbi_debugdata_t) less than the whole payload size.

Layout:

* dump contents

This corresponds to the VMO contents in fuchsia.debugdata.Publisher.

This is arbitrary binary data, its format known by the ultimate consumer.

* sink name

This corresponds to the sink name in the fuchsia.debugdata.Publisher

FIDL protocol. It's a UTF-8 string with no NUL terminator. It's a

simple string that identifies what format the dump contents are in.

It can be used to route collected dumps to appropriate

post-processing. See the FIDL protocol for details.

* VMO name

This corresponds to the ZX_PROP_NAME property set on the VMO published.

It's usually used like a (shortened) file name for the particular dump,

perhaps incorporating a process name and PID or the like.

* log text

This corresponds to logging that was emitted while producing this

data. It might contain messages relevant to understanding the data,

including symbolizer markup with symbolization context. It's a UTF-8

string with no NUL terminator, expected to be log lines (with or

without timestamp and other such prefix text) separated by

characters.

The contents and log text can be of any size, but the sink name is

presumed to be possibly truncated to a small size such as 255 and the

VMO name is presumed to be possibly truncated to a very small size such

as ZX_MAX_NAME_LEN (32).