| Message ID | 87msx99b9o.fsf_-_@gmail.froward.int.ebiederm.org (mailing list archive) |
|---|---|
| State | Accepted |
| Commit | 9eb5419799f08402ee3bd185c2d2c50ded669b06 |
| Headers | show |
| Series | [v3] bulk-checkin: only support blobs in index_bulk_checkin | expand |
"Eric W. Biederman" <ebiederm@gmail.com> writes: > As the code is written today index_bulk_checkin only accepts blobs. > Remove the enum object_type parameter and rename index_bulk_checkin to > index_blob_bulk_checkin, index_stream to index_blob_stream, > deflate_to_pack to deflate_blob_to_pack, stream_to_pack to > stream_blob_to_pack, to make this explicit. > > Not supporting commits, tags, or trees has no downside as it is not > currently supported now, and commits, tags, and trees being smaller by > design do not have the problem that the problem that index_bulk_checkin > was built to solve. > > Before we start adding code to support the hash function transition > supporting additional objects types in index_bulk_checkin has no real > additional cost, just an extra function parameter to know what the > object type is. Once we begin the hash function transition this is not > the case. > > The hash function transition document specifies that a repository with > compatObjectFormat enabled will compute and store both the SHA-1 and > SHA-256 hash of every object in the repository. > > What makes this a challenge is that it is not just an additional hash > over the same object. Instead the hash function transition document > specifies that the compatibility hash (specified with > compatObjectFormat) be computed over the equivalent object that another > git repository whose storage hash (specified with objectFormat) would > store. When comparing equivalent repositories built with different > storage hash functions, the oids embedded in objects used to refer to > other objects differ and the location of signatures within objects > differ. > > As blob objects have neither oids referring to other objects nor stored > signatures their storage hash and their compatibility hash are computed > over the same object. > > The other kinds of objects: trees, commits, and tags, all store oids > referring to other objects. Signatures are stored in commit and tag > objects. As oids and the tags to store signatures are not the same size > in repositories built with different storage hashes the size of the > equivalent objects are also different. > > A version of index_bulk_checkin that supports more than just blobs when > computing both the SHA-1 and the SHA-256 of every object added would > need a different, and more expensive structure. The structure is more > expensive because it would be required to temporarily buffering the > equivalent object the compatibility hash needs to be computed over. > > A temporary object is needed, because before a hash over an object can > computed it's object header needs to be computed. One of the members of > the object header is the entire size of the object. To know the size of > an equivalent object an entire pass over the original object needs to be > made, as trees, commits, and tags are composed of a variable number of > variable sized pieces. Unfortunately there is no formula to compute the > size of an equivalent object from just the size of the original object. > > Avoid all of those future complications by limiting index_bulk_checkin > to only work on blobs. Thanks. Will queue.
On Tue, Sep 26, 2023 at 02:48:31PM -0700, Junio C Hamano wrote: > > Avoid all of those future complications by limiting index_bulk_checkin > > to only work on blobs. > > Thanks. Will queue. Hmm. I wonder if retaining some flexibility in the bulk-checkin mechanism may be worthwhile. We discussed at the Contributor's Summit[^1] today that the bulk-checkin system may be a good fit for packing any blobs/trees created by `merge-tree` or `replay` instead of writing them out as loose objects. Being able to write trees in addition to blobs is definitely important there, so we may want to wait on merging this down until that direction solidifies a bit more. (FWIW, I started working on that today and hope to have patches on the list in the next day or two). Alternatively, if there is an urgency to merge these down, we can always come back to it in the future and revert it if need be. Either way :-). Thanks, Taylor [^1]: I'll clean up our notes in the next day or two and share them with the list here.
Taylor Blau <me@ttaylorr.com> writes: > Hmm. I wonder if retaining some flexibility in the bulk-checkin > mechanism may be worthwhile. We discussed at the Contributor's > Summit[^1] today that the bulk-checkin system may be a good fit for > packing any blobs/trees created by `merge-tree` or `replay` instead of > writing them out as loose objects. But see the last paragraph of my review comments for the earlier round upthread. This particular function implements logic that is only applicable to blob objects, and streaming trees, commits, and tags will need their own separate helper functions. And when they are written, the top-level stream_to_pack() function can be reintroduced, which will be a thin dispatcher to the four type-specific helpers.
On Tue, Sep 26, 2023 at 09:08:59PM -0700, Junio C Hamano wrote: > Taylor Blau <me@ttaylorr.com> writes: > > > Hmm. I wonder if retaining some flexibility in the bulk-checkin > > mechanism may be worthwhile. We discussed at the Contributor's > > Summit[^1] today that the bulk-checkin system may be a good fit for > > packing any blobs/trees created by `merge-tree` or `replay` instead of > > writing them out as loose objects. > > But see the last paragraph of my review comments for the earlier > round upthread. This particular function implements logic that is > only applicable to blob objects, and streaming trees, commits, and > tags will need their own separate helper functions. And when they > are written, the top-level stream_to_pack() function can be > reintroduced, which will be a thin dispatcher to the four > type-specific helpers. I am not sure that I follow. If we have an address in memory from which we want to stream raw bytes directly to the packfile, that should work for all objects regardless of type, no? Having stream_to_pack() take a non-OBJ_BLOB 'type' argument would be OK provided that the file descriptor 'fd' contains the raw contents of an object which matches type 'type'. IIUC, for callers like in the ORT backend which assemble e.g. the raw bytes of a tree in its merge-ort.c::write_tree() function like so: for (i = 0; i < nr; i++) { struct merged_info *mi = versions->items[offset+i].util; struct version_info *ri = &mi->result; strbuf_addf(&buf, "%o %s%c", ri->mode, versions->items[offset+i].string, '\0'); strbuf_add(&buf, ri->oid.hash, hash_size); } we'd want some variant of stream_to_pack() that acts on a 'void *, size_t' pair rather than an 'int (fd), size_t' pair. Likely its signature would look something like: /* write raw bytes to a bulk-checkin pack */ static int write_to_pack(struct bulk_checkin_packfile *state, git_hash_ctx *ctx, off_t *already_hashed_to, void *ptr, size_t size, enum object_type type, unsigned flags); /* write an object from memory to a bulk-checkin pack */ static int deflate_to_pack_mem(struct bulk_checkin_packfile *state, struct object_id *result_oid, void *ptr, size_t size, enum object_type type, unsigned flags); , where the above are analogous to `stream_to_pack()` and `deflate_to_pack()`, respectively. ORT would be taught to conditionally replace calls like: write_object_file(buf.buf, buf.len, OBJ_TREE, result_oid); with: deflate_to_pack_mem(&state, result_oid, buf.buf, buf.len, OBJ_TREE, HASH_WRITE_OBJECT); I guess after writing all of that out, you'd never have any callers of the existing `deflate_to_pack()` function that pass a file descriptor containing the contents of a non-blob object. So in that sense, I don't think that my proposal would change anything about this patch. But I worry that I am missing something here, so having a sanity check would be appreciated ;-). Thanks, Taylor
Taylor Blau <me@ttaylorr.com> writes: > I am not sure that I follow. If we have an address in memory from which > we want to stream raw bytes directly to the packfile, that should work > for all objects regardless of type, no? For a single hash world, yes. For keeping track of "the other hash" and correspondence, you need to (1) interpret the contents of the object (e.g., if you received a tree contents for SHA-1 repository, you'd need to split them into tree entries and know which parts of the bytestream are SHA-1 hashes of the tree contebnts), (2) come up with the corresponding tree contents in the SHA-256 world (you should be able to do that now you know SHA-1 names of the objects directly referred to by the tree) and hash that using SHA-256, and (3) remember the SHA-1 and the SHA-256 name correspondence of the tree object you just hashed, in addition to the usual (4) hashing the contents using SHA-1 hash algorithm without caring what the byte stream represents.
Junio C Hamano <gitster@pobox.com> writes: > Taylor Blau <me@ttaylorr.com> writes: > >> I am not sure that I follow. If we have an address in memory from which >> we want to stream raw bytes directly to the packfile, that should work >> for all objects regardless of type, no? > > For a single hash world, yes. For keeping track of "the other hash" > and correspondence, you need to (1) interpret the contents of the > object (e.g., if you received a tree contents for SHA-1 repository, > you'd need to split them into tree entries and know which parts of > the bytestream are SHA-1 hashes of the tree contebnts), (2) come > up with the corresponding tree contents in the SHA-256 world (you > should be able to do that now you know SHA-1 names of the objects > directly referred to by the tree) and hash that using SHA-256, and > (3) remember the SHA-1 and the SHA-256 name correspondence of the > tree object you just hashed, in addition to the usual (4) hashing > the contents using SHA-1 hash algorithm without caring what the byte > stream represents. If it helps I just posted a patchset that implements what it takes to deal with objects small enough to live in-core. You can read object-file-convert.c to see what it takes to generate an object in the other hash function world. The exercise for the reader is how to apply this to objects that are too large to fit in memory. Eric
Taylor Blau <me@ttaylorr.com> writes: > On Tue, Sep 26, 2023 at 02:48:31PM -0700, Junio C Hamano wrote: >> > Avoid all of those future complications by limiting index_bulk_checkin >> > to only work on blobs. >> >> Thanks. Will queue. > > Hmm. I wonder if retaining some flexibility in the bulk-checkin > mechanism may be worthwhile. We discussed at the Contributor's > Summit[^1] today that the bulk-checkin system may be a good fit for > packing any blobs/trees created by `merge-tree` or `replay` instead of > writing them out as loose objects. > > Being able to write trees in addition to blobs is definitely important > there, so we may want to wait on merging this down until that direction > solidifies a bit more. (FWIW, I started working on that today and hope > to have patches on the list in the next day or two). > > Alternatively, if there is an urgency to merge these down, we can always > come back to it in the future and revert it if need be. Either way > :-). There are two things that index_bulk_checkin does. - Handle objects that are too large to fit into a memory - Place objects immediately in a pack. Do I read things correctly that you want to take an object that is small enough to fit into memory, and to immediately into a pack? If so you essentially want write_object_file that directly writes to a pack? A version of write_object_file that that directly writes to a pack is much easier than the chunking that index_bulk_checkin does. Perhaps your version could be called index_pack_checkin? Eric
just language nits on the commit message: On Tue, Sep 26, 2023 at 10:58:43AM -0500, Eric W. Biederman wrote: >Not supporting commits, tags, or trees has no downside as it is not >currently supported now, and commits, tags, and trees being smaller by >design do not have the problem that the problem that index_bulk_checkin ^^^^^^^^^^^^^^^^ duplicated! >was built to solve. >A version of index_bulk_checkin that supports more than just blobs when >computing both the SHA-1 and the SHA-256 of every object added would >need a different, and more expensive structure. The structure is more >expensive because it would be required to temporarily buffering the ^^^ no 'ing' here. >equivalent object the compatibility hash needs to be computed over. >A temporary object is needed, because before a hash over an object can >computed it's > "be computed, its" >object header needs to be computed. One of the members of regards
diff --git a/bulk-checkin.c b/bulk-checkin.c index 73bff3a23d27..223562b4e748 100644 --- a/bulk-checkin.c +++ b/bulk-checkin.c @@ -155,10 +155,10 @@ static int already_written(struct bulk_checkin_packfile *state, struct object_id * status before calling us just in case we ask it to call us again * with a new pack. */ -static int stream_to_pack(struct bulk_checkin_packfile *state, - git_hash_ctx *ctx, off_t *already_hashed_to, - int fd, size_t size, enum object_type type, - const char *path, unsigned flags) +static int stream_blob_to_pack(struct bulk_checkin_packfile *state, + git_hash_ctx *ctx, off_t *already_hashed_to, + int fd, size_t size, const char *path, + unsigned flags) { git_zstream s; unsigned char ibuf[16384]; @@ -170,7 +170,7 @@ static int stream_to_pack(struct bulk_checkin_packfile *state, git_deflate_init(&s, pack_compression_level); - hdrlen = encode_in_pack_object_header(obuf, sizeof(obuf), type, size); + hdrlen = encode_in_pack_object_header(obuf, sizeof(obuf), OBJ_BLOB, size); s.next_out = obuf + hdrlen; s.avail_out = sizeof(obuf) - hdrlen; @@ -247,11 +247,10 @@ static void prepare_to_stream(struct bulk_checkin_packfile *state, die_errno("unable to write pack header"); } -static int deflate_to_pack(struct bulk_checkin_packfile *state, - struct object_id *result_oid, - int fd, size_t size, - enum object_type type, const char *path, - unsigned flags) +static int deflate_blob_to_pack(struct bulk_checkin_packfile *state, + struct object_id *result_oid, + int fd, size_t size, + const char *path, unsigned flags) { off_t seekback, already_hashed_to; git_hash_ctx ctx; @@ -265,7 +264,7 @@ static int deflate_to_pack(struct bulk_checkin_packfile *state, return error("cannot find the current offset"); header_len = format_object_header((char *)obuf, sizeof(obuf), - type, size); + OBJ_BLOB, size); the_hash_algo->init_fn(&ctx); the_hash_algo->update_fn(&ctx, obuf, header_len); @@ -282,8 +281,8 @@ static int deflate_to_pack(struct bulk_checkin_packfile *state, idx->offset = state->offset; crc32_begin(state->f); } - if (!stream_to_pack(state, &ctx, &already_hashed_to, - fd, size, type, path, flags)) + if (!stream_blob_to_pack(state, &ctx, &already_hashed_to, + fd, size, path, flags)) break; /* * Writing this object to the current pack will make @@ -350,12 +349,12 @@ void fsync_loose_object_bulk_checkin(int fd, const char *filename) } } -int index_bulk_checkin(struct object_id *oid, - int fd, size_t size, enum object_type type, - const char *path, unsigned flags) +int index_blob_bulk_checkin(struct object_id *oid, + int fd, size_t size, + const char *path, unsigned flags) { - int status = deflate_to_pack(&bulk_checkin_packfile, oid, fd, size, type, - path, flags); + int status = deflate_blob_to_pack(&bulk_checkin_packfile, oid, fd, size, + path, flags); if (!odb_transaction_nesting) flush_bulk_checkin_packfile(&bulk_checkin_packfile); return status; diff --git a/bulk-checkin.h b/bulk-checkin.h index 48fe9a6e9171..aa7286a7b3e1 100644 --- a/bulk-checkin.h +++ b/bulk-checkin.h @@ -9,9 +9,9 @@ void prepare_loose_object_bulk_checkin(void); void fsync_loose_object_bulk_checkin(int fd, const char *filename); -int index_bulk_checkin(struct object_id *oid, - int fd, size_t size, enum object_type type, - const char *path, unsigned flags); +int index_blob_bulk_checkin(struct object_id *oid, + int fd, size_t size, + const char *path, unsigned flags); /* * Tell the object database to optimize for adding diff --git a/object-file.c b/object-file.c index 7dc0c4bfbba8..7c7afe579364 100644 --- a/object-file.c +++ b/object-file.c @@ -2446,11 +2446,11 @@ static int index_core(struct index_state *istate, * binary blobs, they generally do not want to get any conversion, and * callers should avoid this code path when filters are requested. */ -static int index_stream(struct object_id *oid, int fd, size_t size, - enum object_type type, const char *path, - unsigned flags) +static int index_blob_stream(struct object_id *oid, int fd, size_t size, + const char *path, + unsigned flags) { - return index_bulk_checkin(oid, fd, size, type, path, flags); + return index_blob_bulk_checkin(oid, fd, size, path, flags); } int index_fd(struct index_state *istate, struct object_id *oid, @@ -2472,8 +2472,8 @@ int index_fd(struct index_state *istate, struct object_id *oid, ret = index_core(istate, oid, fd, xsize_t(st->st_size), type, path, flags); else - ret = index_stream(oid, fd, xsize_t(st->st_size), type, path, - flags); + ret = index_blob_stream(oid, fd, xsize_t(st->st_size), path, + flags); close(fd); return ret; }
As the code is written today index_bulk_checkin only accepts blobs. Remove the enum object_type parameter and rename index_bulk_checkin to index_blob_bulk_checkin, index_stream to index_blob_stream, deflate_to_pack to deflate_blob_to_pack, stream_to_pack to stream_blob_to_pack, to make this explicit. Not supporting commits, tags, or trees has no downside as it is not currently supported now, and commits, tags, and trees being smaller by design do not have the problem that the problem that index_bulk_checkin was built to solve. Before we start adding code to support the hash function transition supporting additional objects types in index_bulk_checkin has no real additional cost, just an extra function parameter to know what the object type is. Once we begin the hash function transition this is not the case. The hash function transition document specifies that a repository with compatObjectFormat enabled will compute and store both the SHA-1 and SHA-256 hash of every object in the repository. What makes this a challenge is that it is not just an additional hash over the same object. Instead the hash function transition document specifies that the compatibility hash (specified with compatObjectFormat) be computed over the equivalent object that another git repository whose storage hash (specified with objectFormat) would store. When comparing equivalent repositories built with different storage hash functions, the oids embedded in objects used to refer to other objects differ and the location of signatures within objects differ. As blob objects have neither oids referring to other objects nor stored signatures their storage hash and their compatibility hash are computed over the same object. The other kinds of objects: trees, commits, and tags, all store oids referring to other objects. Signatures are stored in commit and tag objects. As oids and the tags to store signatures are not the same size in repositories built with different storage hashes the size of the equivalent objects are also different. A version of index_bulk_checkin that supports more than just blobs when computing both the SHA-1 and the SHA-256 of every object added would need a different, and more expensive structure. The structure is more expensive because it would be required to temporarily buffering the equivalent object the compatibility hash needs to be computed over. A temporary object is needed, because before a hash over an object can computed it's object header needs to be computed. One of the members of the object header is the entire size of the object. To know the size of an equivalent object an entire pass over the original object needs to be made, as trees, commits, and tags are composed of a variable number of variable sized pieces. Unfortunately there is no formula to compute the size of an equivalent object from just the size of the original object. Avoid all of those future complications by limiting index_bulk_checkin to only work on blobs. Inspired-by: brian m. carlson <sandals@crustytoothpaste.net> Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com> --- bulk-checkin.c | 35 +++++++++++++++++------------------ bulk-checkin.h | 6 +++--- object-file.c | 12 ++++++------ 3 files changed, 26 insertions(+), 27 deletions(-)