Git Source Code Mirror - This is a publish-only repository and all pull requests are ignored. Please follow Documentation/SubmittingPatches procedure for any of your improvements. https://git-scm.com/
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
git/midx.c

2053 lines
52 KiB

#include "cache.h"
#include "config.h"
#include "csum-file.h"
#include "dir.h"
#include "lockfile.h"
#include "packfile.h"
#include "object-store.h"
#include "hash-lookup.h"
#include "midx.h"
#include "progress.h"
#include "trace2.h"
midx: implement midx_repack() To repack with a non-zero batch-size, first sort all pack-files by their modified time. Second, walk those pack-files from oldest to newest, compute their expected size, and add the packs to a list if they are smaller than the given batch-size. Stop when the total expected size is at least the batch size. If the batch size is zero, select all packs in the multi-pack-index. Finally, collect the objects from the multi-pack-index that are in the selected packs and send them to 'git pack-objects'. Write a new multi-pack-index that includes the new pack. Using a batch size of zero is very similar to a standard 'git repack' command, except that we do not delete the old packs and instead rely on the new multi-pack-index to prevent new processes from reading the old packs. This does not disrupt other Git processes that are currently reading the old packs based on the old multi-pack-index. While first designing a 'git multi-pack-index repack' operation, I started by collecting the batches based on the actual size of the objects instead of the size of the pack-files. This allows repacking a large pack-file that has very few referencd objects. However, this came at a significant cost of parsing pack-files instead of simply reading the multi-pack-index and getting the file information for the pack-files. The "expected size" version provides similar behavior, but could skip a pack-file if the average object size is much larger than the actual size of the referenced objects, or can create a large pack if the actual size of the referenced objects is larger than the expected size. Signed-off-by: Derrick Stolee <dstolee@microsoft.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years ago
#include "run-command.h"
#include "repository.h"
#include "chunk-format.h"
#include "pack.h"
#include "pack-bitmap.h"
#include "refs.h"
#include "revision.h"
#include "list-objects.h"
#define MIDX_SIGNATURE 0x4d494458 /* "MIDX" */
#define MIDX_VERSION 1
#define MIDX_BYTE_FILE_VERSION 4
#define MIDX_BYTE_HASH_VERSION 5
#define MIDX_BYTE_NUM_CHUNKS 6
#define MIDX_BYTE_NUM_PACKS 8
#define MIDX_HEADER_SIZE 12
#define MIDX_MIN_SIZE (MIDX_HEADER_SIZE + the_hash_algo->rawsz)
#define MIDX_CHUNK_ALIGNMENT 4
#define MIDX_CHUNKID_PACKNAMES 0x504e414d /* "PNAM" */
#define MIDX_CHUNKID_OIDFANOUT 0x4f494446 /* "OIDF" */
#define MIDX_CHUNKID_OIDLOOKUP 0x4f49444c /* "OIDL" */
#define MIDX_CHUNKID_OBJECTOFFSETS 0x4f4f4646 /* "OOFF" */
#define MIDX_CHUNKID_LARGEOFFSETS 0x4c4f4646 /* "LOFF" */
midx.c: make changing the preferred pack safe The previous patch demonstrates a bug where a MIDX's auxiliary object order can become out of sync with a MIDX bitmap. This is because of two confounding factors: - First, the object order is stored in a file which is named according to the multi-pack index's checksum, and the MIDX does not store the object order. This means that the object order can change without altering the checksum. - But the .rev file is moved into place with finalize_object_file(), which link(2)'s the file into place instead of renaming it. For us, that means that a modified .rev file will not be moved into place if MIDX's checksum was unchanged. This fix is to force the MIDX's checksum to change when the preferred pack changes but the set of packs contained in the MIDX does not. In other words, when the object order changes, the MIDX's checksum needs to change with it (regardless of whether the MIDX is tracking the same or different packs). This prevents a race whereby changing the object order (but not the packs themselves) enables a reader to see the new .rev file with the old MIDX, or similarly seeing the new bitmap with the old object order. But why can't we just stop hardlinking the .rev into place instead adding additional data to the MIDX? Suppose that's what we did. Then when we go to generate the new bitmap, we'll load the old MIDX bitmap, along with the MIDX that it references. That's fine, since the new MIDX isn't moved into place until after the new bitmap is generated. But the new object order *has* been moved into place. So we'll read the old bitmaps in the new order when generating the new bitmap file, meaning that without this secondary change, bitmap generation itself would become a victim of the race described here. This can all be prevented by forcing the MIDX's checksum to change when the object order does. By embedding the entire object order into the MIDX, we do just that. That is, the MIDX's checksum will change in response to any perturbation of the underlying object order. In t5326, this will cause the MIDX's checksum to update (even without changing the set of packs in the MIDX), preventing the stale read problem. Note that this makes it safe to continue to link(2) the MIDX .rev file into place, since it is now impossible to have a .rev file that is out-of-sync with the MIDX whose checksum it references. (But we will do away with MIDX .rev files later in this series anyway, so this is somewhat of a moot point). In theory, it is possible to store a "fingerprint" of the full object order here, so long as that fingerprint changes at least as often as the full object order does. Some possibilities here include storing the identity of the preferred pack, along with the mtimes of the non-preferred packs in a consistent order. But storing a limited part of the information makes it difficult to reason about whether or not there are gaps between the two that would cause us to get bitten by this bug again. Signed-off-by: Taylor Blau <me@ttaylorr.com> Reviewed-by: Derrick Stolee <dstolee@microsoft.com> Reviewed-by: Jonathan Tan <jonathantanmy@google.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 months ago
#define MIDX_CHUNKID_REVINDEX 0x52494458 /* "RIDX" */
#define MIDX_CHUNK_FANOUT_SIZE (sizeof(uint32_t) * 256)
#define MIDX_CHUNK_OFFSET_WIDTH (2 * sizeof(uint32_t))
#define MIDX_CHUNK_LARGE_OFFSET_WIDTH (sizeof(uint64_t))
#define MIDX_LARGE_OFFSET_NEEDED 0x80000000
#define PACK_EXPIRED UINT_MAX
const unsigned char *get_midx_checksum(struct multi_pack_index *m)
pack-revindex: read multi-pack reverse indexes Implement reading for multi-pack reverse indexes, as described in the previous patch. Note that these functions don't yet have any callers, and won't until multi-pack reachability bitmaps are introduced in a later patch series. In the meantime, this patch implements some of the infrastructure necessary to support multi-pack bitmaps. There are three new functions exposed by the revindex API: - load_midx_revindex(): loads the reverse index corresponding to the given multi-pack index. - midx_to_pack_pos() and pack_pos_to_midx(): these convert between the multi-pack index and pseudo-pack order. load_midx_revindex() and pack_pos_to_midx() are both relatively straightforward. load_midx_revindex() needs a few functions to be exposed from the midx API. One to get the checksum of a midx, and another to get the .rev's filename. Similar to recent changes in the packed_git struct, three new fields are added to the multi_pack_index struct: one to keep track of the size, one to keep track of the mmap'd pointer, and another to point past the header and at the reverse index's data. pack_pos_to_midx() simply reads the corresponding entry out of the table. midx_to_pack_pos() is the trickiest, since it needs to find an object's position in the psuedo-pack order, but that order can only be recovered in the .rev file itself. This mapping can be implemented with a binary search, but note that the thing we're binary searching over isn't an array of values, but rather a permuted order of those values. So, when comparing two items, it's helpful to keep in mind the difference. Instead of a traditional binary search, where you are comparing two things directly, here we're comparing a (pack, offset) tuple with an index into the multi-pack index. That index describes another (pack, offset) tuple, and it is _those_ two tuples that are compared. Signed-off-by: Taylor Blau <me@ttaylorr.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years ago
{
return m->data + m->data_len - the_hash_algo->rawsz;
}
void get_midx_filename(struct strbuf *out, const char *object_dir)
{
strbuf_addf(out, "%s/pack/multi-pack-index", object_dir);
}
void get_midx_rev_filename(struct strbuf *out, struct multi_pack_index *m)
pack-revindex: read multi-pack reverse indexes Implement reading for multi-pack reverse indexes, as described in the previous patch. Note that these functions don't yet have any callers, and won't until multi-pack reachability bitmaps are introduced in a later patch series. In the meantime, this patch implements some of the infrastructure necessary to support multi-pack bitmaps. There are three new functions exposed by the revindex API: - load_midx_revindex(): loads the reverse index corresponding to the given multi-pack index. - midx_to_pack_pos() and pack_pos_to_midx(): these convert between the multi-pack index and pseudo-pack order. load_midx_revindex() and pack_pos_to_midx() are both relatively straightforward. load_midx_revindex() needs a few functions to be exposed from the midx API. One to get the checksum of a midx, and another to get the .rev's filename. Similar to recent changes in the packed_git struct, three new fields are added to the multi_pack_index struct: one to keep track of the size, one to keep track of the mmap'd pointer, and another to point past the header and at the reverse index's data. pack_pos_to_midx() simply reads the corresponding entry out of the table. midx_to_pack_pos() is the trickiest, since it needs to find an object's position in the psuedo-pack order, but that order can only be recovered in the .rev file itself. This mapping can be implemented with a binary search, but note that the thing we're binary searching over isn't an array of values, but rather a permuted order of those values. So, when comparing two items, it's helpful to keep in mind the difference. Instead of a traditional binary search, where you are comparing two things directly, here we're comparing a (pack, offset) tuple with an index into the multi-pack index. That index describes another (pack, offset) tuple, and it is _those_ two tuples that are compared. Signed-off-by: Taylor Blau <me@ttaylorr.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years ago
{
get_midx_filename(out, m->object_dir);
strbuf_addf(out, "-%s.rev", hash_to_hex(get_midx_checksum(m)));
pack-revindex: read multi-pack reverse indexes Implement reading for multi-pack reverse indexes, as described in the previous patch. Note that these functions don't yet have any callers, and won't until multi-pack reachability bitmaps are introduced in a later patch series. In the meantime, this patch implements some of the infrastructure necessary to support multi-pack bitmaps. There are three new functions exposed by the revindex API: - load_midx_revindex(): loads the reverse index corresponding to the given multi-pack index. - midx_to_pack_pos() and pack_pos_to_midx(): these convert between the multi-pack index and pseudo-pack order. load_midx_revindex() and pack_pos_to_midx() are both relatively straightforward. load_midx_revindex() needs a few functions to be exposed from the midx API. One to get the checksum of a midx, and another to get the .rev's filename. Similar to recent changes in the packed_git struct, three new fields are added to the multi_pack_index struct: one to keep track of the size, one to keep track of the mmap'd pointer, and another to point past the header and at the reverse index's data. pack_pos_to_midx() simply reads the corresponding entry out of the table. midx_to_pack_pos() is the trickiest, since it needs to find an object's position in the psuedo-pack order, but that order can only be recovered in the .rev file itself. This mapping can be implemented with a binary search, but note that the thing we're binary searching over isn't an array of values, but rather a permuted order of those values. So, when comparing two items, it's helpful to keep in mind the difference. Instead of a traditional binary search, where you are comparing two things directly, here we're comparing a (pack, offset) tuple with an index into the multi-pack index. That index describes another (pack, offset) tuple, and it is _those_ two tuples that are compared. Signed-off-by: Taylor Blau <me@ttaylorr.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years ago
}
static int midx_read_oid_fanout(const unsigned char *chunk_start,
size_t chunk_size, void *data)
{
struct multi_pack_index *m = data;
m->chunk_oid_fanout = (uint32_t *)chunk_start;
if (chunk_size != 4 * 256) {
error(_("multi-pack-index OID fanout is of the wrong size"));
return 1;
}
return 0;
}
struct multi_pack_index *load_multi_pack_index(const char *object_dir, int local)
{
struct multi_pack_index *m = NULL;
int fd;
struct stat st;
size_t midx_size;
void *midx_map = NULL;
uint32_t hash_version;
struct strbuf midx_name = STRBUF_INIT;
uint32_t i;
const char *cur_pack_name;
struct chunkfile *cf = NULL;
get_midx_filename(&midx_name, object_dir);
fd = git_open(midx_name.buf);
if (fd < 0)
goto cleanup_fail;
if (fstat(fd, &st)) {
error_errno(_("failed to read %s"), midx_name.buf);
goto cleanup_fail;
}
midx_size = xsize_t(st.st_size);
if (midx_size < MIDX_MIN_SIZE) {
error(_("multi-pack-index file %s is too small"), midx_name.buf);
goto cleanup_fail;
}
strbuf_release(&midx_name);
midx_map = xmmap(NULL, midx_size, PROT_READ, MAP_PRIVATE, fd, 0);
close(fd);
FLEX_ALLOC_STR(m, object_dir, object_dir);
m->data = midx_map;
m->data_len = midx_size;
m->local = local;
m->signature = get_be32(m->data);
if (m->signature != MIDX_SIGNATURE)
die(_("multi-pack-index signature 0x%08x does not match signature 0x%08x"),
m->signature, MIDX_SIGNATURE);
m->version = m->data[MIDX_BYTE_FILE_VERSION];
if (m->version != MIDX_VERSION)
die(_("multi-pack-index version %d not recognized"),
m->version);
hash_version = m->data[MIDX_BYTE_HASH_VERSION];
if (hash_version != oid_version(the_hash_algo)) {
multi-pack-index: use hash version byte Similar to the commit-graph format, the multi-pack-index format has a byte in the header intended to track the hash version used to write the file. This allows one to interpret the hash length without having the context of the repository config specifying the hash length. This was not modified as part of the SHA-256 work because the hash length was automatically up-shifted due to that config. Since we have this byte available, we can make the file formats more obviously incompatible instead of relying on other context from the repository. Add a new oid_version() method in midx.c similar to the one in commit-graph.c. This is specifically made separate from that implementation to avoid artificially linking the formats. The test impact requires a few more things than the corresponding change in the commit-graph format. Specifically, 'test-tool read-midx' was not writing anything about this header value to output. Since the value available in 'struct multi_pack_index' is hash_len instead of a version value, we output "20" or "32" instead of "1" or "2". Since we want a user to not have their Git commands fail if their multi-pack-index has the incorrect hash version compared to the repository's hash version, we relax the die() to an error() in load_multi_pack_index(). This has some effect on 'git multi-pack-index verify' as we need to check that a failed parse of a file that exists is actually a verify error. For that test that checks the hash version matches, we change the corrupted byte from "2" to "3" to ensure the test fails for both hash algorithms. Helped-by: brian m. carlson <sandals@crustytoothpaste.net> Signed-off-by: Derrick Stolee <dstolee@microsoft.com> Reviewed-by: brian m. carlson <sandals@crustytoothpaste.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years ago
error(_("multi-pack-index hash version %u does not match version %u"),
hash_version, oid_version(the_hash_algo));
multi-pack-index: use hash version byte Similar to the commit-graph format, the multi-pack-index format has a byte in the header intended to track the hash version used to write the file. This allows one to interpret the hash length without having the context of the repository config specifying the hash length. This was not modified as part of the SHA-256 work because the hash length was automatically up-shifted due to that config. Since we have this byte available, we can make the file formats more obviously incompatible instead of relying on other context from the repository. Add a new oid_version() method in midx.c similar to the one in commit-graph.c. This is specifically made separate from that implementation to avoid artificially linking the formats. The test impact requires a few more things than the corresponding change in the commit-graph format. Specifically, 'test-tool read-midx' was not writing anything about this header value to output. Since the value available in 'struct multi_pack_index' is hash_len instead of a version value, we output "20" or "32" instead of "1" or "2". Since we want a user to not have their Git commands fail if their multi-pack-index has the incorrect hash version compared to the repository's hash version, we relax the die() to an error() in load_multi_pack_index(). This has some effect on 'git multi-pack-index verify' as we need to check that a failed parse of a file that exists is actually a verify error. For that test that checks the hash version matches, we change the corrupted byte from "2" to "3" to ensure the test fails for both hash algorithms. Helped-by: brian m. carlson <sandals@crustytoothpaste.net> Signed-off-by: Derrick Stolee <dstolee@microsoft.com> Reviewed-by: brian m. carlson <sandals@crustytoothpaste.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years ago
goto cleanup_fail;
}
m->hash_len = the_hash_algo->rawsz;
m->num_chunks = m->data[MIDX_BYTE_NUM_CHUNKS];
m->num_packs = get_be32(m->data + MIDX_BYTE_NUM_PACKS);
cf = init_chunkfile(NULL);
if (read_table_of_contents(cf, m->data, midx_size,
MIDX_HEADER_SIZE, m->num_chunks))
goto cleanup_fail;
if (pair_chunk(cf, MIDX_CHUNKID_PACKNAMES, &m->chunk_pack_names) == CHUNK_NOT_FOUND)
die(_("multi-pack-index missing required pack-name chunk"));
if (read_chunk(cf, MIDX_CHUNKID_OIDFANOUT, midx_read_oid_fanout, m) == CHUNK_NOT_FOUND)
die(_("multi-pack-index missing required OID fanout chunk"));
if (pair_chunk(cf, MIDX_CHUNKID_OIDLOOKUP, &m->chunk_oid_lookup) == CHUNK_NOT_FOUND)
die(_("multi-pack-index missing required OID lookup chunk"));
if (pair_chunk(cf, MIDX_CHUNKID_OBJECTOFFSETS, &m->chunk_object_offsets) == CHUNK_NOT_FOUND)
die(_("multi-pack-index missing required object offsets chunk"));
pair_chunk(cf, MIDX_CHUNKID_LARGEOFFSETS, &m->chunk_large_offsets);
midx: read `RIDX` chunk when present When a MIDX contains the new `RIDX` chunk, ensure that the reverse index is read from it instead of the on-disk .rev file. Since we need to encode the object order in the MIDX itself for correctness reasons, there is no point in storing the same data again outside of the MIDX. So, this patch stops writing separate .rev files, and reads it out of the MIDX itself. This is possible to do with relatively little new code, since the format of the RIDX chunk is identical to the data in the .rev file. In other words, we can implement this by pointing the `revindex_data` field at the reverse index chunk of the MIDX instead of the .rev file without any other changes. Note that we have two knobs that are adjusted for the new tests: GIT_TEST_MIDX_WRITE_REV and GIT_TEST_MIDX_READ_RIDX. The former controls whether the MIDX .rev is written at all, and the latter controls whether we read the MIDX's RIDX chunk. Both are necessary to ensure that the test added at the beginning of this series continues to work. This is because we always need to write the RIDX chunk in the MIDX in order to change its checksum, but we want to make sure reading the existing .rev file still works (since the RIDX chunk takes precedence by default). Arguably this isn't a very interesting mode to test, because the precedence rules mean that we'll always read the RIDX chunk over the .rev file. But it makes it impossible for a user to induce corruption in their repository by adjusting the test knobs (since if we had an either/or knob they could stop writing the RIDX chunk, allowing them to tweak the MIDX's object order without changing its checksum). Signed-off-by: Taylor Blau <me@ttaylorr.com> Reviewed-by: Derrick Stolee <dstolee@microsoft.com> Reviewed-by: Jonathan Tan <jonathantanmy@google.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
8 months ago
if (git_env_bool("GIT_TEST_MIDX_READ_RIDX", 1))
pair_chunk(cf, MIDX_CHUNKID_REVINDEX, &m->chunk_revindex);
m->num_objects = ntohl(m->chunk_oid_fanout[255]);
CALLOC_ARRAY(m->pack_names, m->num_packs);
CALLOC_ARRAY(m->packs, m->num_packs);
cur_pack_name = (const char *)m->chunk_pack_names;
for (i = 0; i < m->num_packs; i++) {
m->pack_names[i] = cur_pack_name;
cur_pack_name += strlen(cur_pack_name) + 1;
if (i && strcmp(m->pack_names[i], m->pack_names[i - 1]) <= 0)
die(_("multi-pack-index pack names out of order: '%s' before '%s'"),
m->pack_names[i - 1],
m->pack_names[i]);
}
trace2_data_intmax("midx", the_repository, "load/num_packs", m->num_packs);
trace2_data_intmax("midx", the_repository, "load/num_objects", m->num_objects);
free_chunkfile(cf);
return m;
cleanup_fail:
free(m);
strbuf_release(&midx_name);
free_chunkfile(cf);
if (midx_map)
munmap(midx_map, midx_size);
if (0 <= fd)
close(fd);
return NULL;
}
void close_midx(struct multi_pack_index *m)
{
uint32_t i;
if (!m)
return;
close_midx(m->next);
munmap((unsigned char *)m->data, m->data_len);
for (i = 0; i < m->num_packs; i++) {
midx: add packs to packed_git linked list The multi-pack-index allows searching for objects across multiple packs using one object list. The original design gains many of these performance benefits by keeping the packs in the multi-pack-index out of the packed_git list. Unfortunately, this has one major drawback. If the multi-pack-index covers thousands of packs, and a command loads many of those packs, then we can hit the limit for open file descriptors. The close_one_pack() method is used to limit this resource, but it only looks at the packed_git list, and uses an LRU cache to prevent thrashing. Instead of complicating this close_one_pack() logic to include direct references to the multi-pack-index, simply add the packs opened by the multi-pack-index to the packed_git list. This immediately solves the file-descriptor limit problem, but requires some extra steps to avoid performance issues or other problems: 1. Create a multi_pack_index bit in the packed_git struct that is one if and only if the pack was loaded from a multi-pack-index. 2. Skip packs with the multi_pack_index bit when doing object lookups and abbreviations. These algorithms already check the multi-pack-index before the packed_git struct. This has a very small performance hit, as we need to walk more packed_git structs. This is acceptable, since these operations run binary search on the other packs, so this walk-and-ignore logic is very fast by comparison. 3. When closing a multi-pack-index file, do not close its packs, as those packs will be closed using close_all_packs(). In some cases, such as 'git repack', we run 'close_midx()' without also closing the packs, so we need to un-set the multi_pack_index bit in those packs. This is necessary, and caught by running t6501-freshen-objects.sh with GIT_TEST_MULTI_PACK_INDEX=1. To manually test this change, I inserted trace2 logging into close_pack_fd() and set pack_max_fds to 10, then ran 'git rev-list --all --objects' on a copy of the Git repo with 300+ pack-files and a multi-pack-index. The logs verified the packs are closed as we read them beyond the file descriptor limit. Signed-off-by: Derrick Stolee <dstolee@microsoft.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years ago
if (m->packs[i])
m->packs[i]->multi_pack_index = 0;
}
FREE_AND_NULL(m->packs);
FREE_AND_NULL(m->pack_names);
free(m);
}
int prepare_midx_pack(struct repository *r, struct multi_pack_index *m, uint32_t pack_int_id)
{
struct strbuf pack_name = STRBUF_INIT;
midx: add packs to packed_git linked list The multi-pack-index allows searching for objects across multiple packs using one object list. The original design gains many of these performance benefits by keeping the packs in the multi-pack-index out of the packed_git list. Unfortunately, this has one major drawback. If the multi-pack-index covers thousands of packs, and a command loads many of those packs, then we can hit the limit for open file descriptors. The close_one_pack() method is used to limit this resource, but it only looks at the packed_git list, and uses an LRU cache to prevent thrashing. Instead of complicating this close_one_pack() logic to include direct references to the multi-pack-index, simply add the packs opened by the multi-pack-index to the packed_git list. This immediately solves the file-descriptor limit problem, but requires some extra steps to avoid performance issues or other problems: 1. Create a multi_pack_index bit in the packed_git struct that is one if and only if the pack was loaded from a multi-pack-index. 2. Skip packs with the multi_pack_index bit when doing object lookups and abbreviations. These algorithms already check the multi-pack-index before the packed_git struct. This has a very small performance hit, as we need to walk more packed_git structs. This is acceptable, since these operations run binary search on the other packs, so this walk-and-ignore logic is very fast by comparison. 3. When closing a multi-pack-index file, do not close its packs, as those packs will be closed using close_all_packs(). In some cases, such as 'git repack', we run 'close_midx()' without also closing the packs, so we need to un-set the multi_pack_index bit in those packs. This is necessary, and caught by running t6501-freshen-objects.sh with GIT_TEST_MULTI_PACK_INDEX=1. To manually test this change, I inserted trace2 logging into close_pack_fd() and set pack_max_fds to 10, then ran 'git rev-list --all --objects' on a copy of the Git repo with 300+ pack-files and a multi-pack-index. The logs verified the packs are closed as we read them beyond the file descriptor limit. Signed-off-by: Derrick Stolee <dstolee@microsoft.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years ago
struct packed_git *p;
if (pack_int_id >= m->num_packs)
die(_("bad pack-int-id: %u (%u total packs)"),
pack_int_id, m->num_packs);
if (m->packs[pack_int_id])
return 0;
strbuf_addf(&pack_name, "%s/pack/%s", m->object_dir,
m->pack_names[pack_int_id]);
midx: add packs to packed_git linked list The multi-pack-index allows searching for objects across multiple packs using one object list. The original design gains many of these performance benefits by keeping the packs in the multi-pack-index out of the packed_git list. Unfortunately, this has one major drawback. If the multi-pack-index covers thousands of packs, and a command loads many of those packs, then we can hit the limit for open file descriptors. The close_one_pack() method is used to limit this resource, but it only looks at the packed_git list, and uses an LRU cache to prevent thrashing. Instead of complicating this close_one_pack() logic to include direct references to the multi-pack-index, simply add the packs opened by the multi-pack-index to the packed_git list. This immediately solves the file-descriptor limit problem, but requires some extra steps to avoid performance issues or other problems: 1. Create a multi_pack_index bit in the packed_git struct that is one if and only if the pack was loaded from a multi-pack-index. 2. Skip packs with the multi_pack_index bit when doing object lookups and abbreviations. These algorithms already check the multi-pack-index before the packed_git struct. This has a very small performance hit, as we need to walk more packed_git structs. This is acceptable, since these operations run binary search on the other packs, so this walk-and-ignore logic is very fast by comparison. 3. When closing a multi-pack-index file, do not close its packs, as those packs will be closed using close_all_packs(). In some cases, such as 'git repack', we run 'close_midx()' without also closing the packs, so we need to un-set the multi_pack_index bit in those packs. This is necessary, and caught by running t6501-freshen-objects.sh with GIT_TEST_MULTI_PACK_INDEX=1. To manually test this change, I inserted trace2 logging into close_pack_fd() and set pack_max_fds to 10, then ran 'git rev-list --all --objects' on a copy of the Git repo with 300+ pack-files and a multi-pack-index. The logs verified the packs are closed as we read them beyond the file descriptor limit. Signed-off-by: Derrick Stolee <dstolee@microsoft.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years ago
p = add_packed_git(pack_name.buf, pack_name.len, m->local);
strbuf_release(&pack_name);
midx: add packs to packed_git linked list The multi-pack-index allows searching for objects across multiple packs using one object list. The original design gains many of these performance benefits by keeping the packs in the multi-pack-index out of the packed_git list. Unfortunately, this has one major drawback. If the multi-pack-index covers thousands of packs, and a command loads many of those packs, then we can hit the limit for open file descriptors. The close_one_pack() method is used to limit this resource, but it only looks at the packed_git list, and uses an LRU cache to prevent thrashing. Instead of complicating this close_one_pack() logic to include direct references to the multi-pack-index, simply add the packs opened by the multi-pack-index to the packed_git list. This immediately solves the file-descriptor limit problem, but requires some extra steps to avoid performance issues or other problems: 1. Create a multi_pack_index bit in the packed_git struct that is one if and only if the pack was loaded from a multi-pack-index. 2. Skip packs with the multi_pack_index bit when doing object lookups and abbreviations. These algorithms already check the multi-pack-index before the packed_git struct. This has a very small performance hit, as we need to walk more packed_git structs. This is acceptable, since these operations run binary search on the other packs, so this walk-and-ignore logic is very fast by comparison. 3. When closing a multi-pack-index file, do not close its packs, as those packs will be closed using close_all_packs(). In some cases, such as 'git repack', we run 'close_midx()' without also closing the packs, so we need to un-set the multi_pack_index bit in those packs. This is necessary, and caught by running t6501-freshen-objects.sh with GIT_TEST_MULTI_PACK_INDEX=1. To manually test this change, I inserted trace2 logging into close_pack_fd() and set pack_max_fds to 10, then ran 'git rev-list --all --objects' on a copy of the Git repo with 300+ pack-files and a multi-pack-index. The logs verified the packs are closed as we read them beyond the file descriptor limit. Signed-off-by: Derrick Stolee <dstolee@microsoft.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
3 years ago
if (!p)
return 1;
p->multi_pack_index = 1;
m->packs[pack_int_id] = p;
install_packed_git(r, p);
list_add_tail(&p->mru, &r->objects->packed_git_mru);
return 0;
}
int bsearch_midx(const struct object_id *oid, struct multi_pack_index *m, uint32_t *result)
{
return bsearch_hash(oid->hash, m->chunk_oid_fanout, m->chunk_oid_lookup,
the_hash_algo->rawsz, result);
}
struct object_id *nth_midxed_object_oid(struct object_id *oid,
struct multi_pack_index *m,
uint32_t n)
{
if (n >= m->num_objects)
return NULL;
oidread(oid, m->chunk_oid_lookup + m->hash_len * n);
return oid;
}
off_t nth_midxed_offset(struct multi_pack_index *m, uint32_t pos)
{
const unsigned char *offset_data;
uint32_t offset32;
offset_data = m->chunk_object_offsets + (off_t)pos * MIDX_CHUNK_OFFSET_WIDTH;
offset32 = get_be32(offset_data + sizeof(uint32_t));
if (m->chunk_large_offsets && offset32 & MIDX_LARGE_OFFSET_NEEDED) {
if (sizeof(off_t) < sizeof(uint64_t))
die(_("multi-pack-index stores a 64-bit offset, but off_t is too small"));
offset32 ^= MIDX_LARGE_OFFSET_NEEDED;
return get_be64(m->chunk_large_offsets + sizeof(uint64_t) * offset32);
}
return offset32;
}
uint32_t nth_midxed_pack_int_id(struct multi_pack_index *m, uint32_t pos)
{
return get_be32(m->chunk_object_offsets +
(off_t)pos * MIDX_CHUNK_OFFSET_WIDTH);
}
int fill_midx_entry(struct repository * r,
const struct object_id *oid,
struct pack_entry *e,
struct multi_pack_index *m)
{
uint32_t pos;
uint32_t pack_int_id;
struct packed_git *p;
if (!bsearch_midx(oid, m, &pos))
return 0;
if (pos >= m->num_objects)
return 0;
pack_int_id = nth_midxed_pack_int_id(m, pos);
if (prepare_midx_pack(r, m, pack_int_id))
midx.c: protect against disappearing packs When a packed object is stored in a multi-pack index, but that pack has racily gone away, the MIDX code simply calls die(), when it could be returning an error to the caller, which would in turn lead to re-scanning the pack directory. A pack can racily disappear, for example, due to a simultaneous 'git repack -ad', You can also reproduce this with two terminals, where one is running: git init while true; do git commit -q --allow-empty -m foo git repack -ad git multi-pack-index write done (in effect, constantly writing new MIDXs), and the other is running: obj=$(git rev-parse HEAD) while true; do echo $obj | git cat-file --batch-check='%(objectsize:disk)' || break done That will sometimes hit the error preparing packfile from multi-pack-index message, which this patch fixes. Right now, that path to discovering a missing pack looks something like 'find_pack_entry()' calling 'fill_midx_entry()' and eventually making its way to call 'nth_midxed_pack_entry()'. 'nth_midxed_pack_entry()' already checks 'is_pack_valid()' and propagates an error if the pack is invalid. So, this works if the pack has gone away between calling 'prepare_midx_pack()' and before calling 'is_pack_valid()', but not if it disappears before then. Catch the case where the pack has already disappeared before 'prepare_midx_pack()' by returning an error in that case, too. Co-authored-by: Jeff King <peff@peff.net> Signed-off-by: Taylor Blau <me@ttaylorr.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years ago
return 0;
p = m->packs[pack_int_id];
/*
* We are about to tell the caller where they can locate the
* requested object. We better make sure the packfile is
* still here and can be accessed before supplying that
* answer, as it may have been deleted since the MIDX was
* loaded!
*/
if (!is_pack_valid(p))
return 0;
if (oidset_size(&p->bad_objects) &&
oidset_contains(&p->bad_objects, oid))
return 0;
e->offset = nth_midxed_offset(m, pos);
e->p = p;
return 1;
}
/* Match "foo.idx" against either "foo.pack" _or_ "foo.idx". */
static int cmp_idx_or_pack_name(const char *idx_or_pack_name,
const char *idx_name)
{
/* Skip past any initial matching prefix. */
while (*idx_name && *idx_name == *idx_or_pack_name) {
idx_name++;
idx_or_pack_name++;
}
/*
* If we didn't match completely, we may have matched "pack-1234." and
* be left with "idx" and "pack" respectively, which is also OK. We do
* not have to check for "idx" and "idx", because that would have been
* a complete match (and in that case these strcmps will be false, but
* we'll correctly return 0 from the final strcmp() below.
*
* Technically this matches "fooidx" and "foopack", but we'd never have
* such names in the first place.
*/
if (!strcmp(idx_name, "idx") && !strcmp(idx_or_pack_name, "pack"))
return 0;
/*
* This not only checks for a complete match, but also orders based on
* the first non-identical character, which means our ordering will
* match a raw strcmp(). That makes it OK to use this to binary search
* a naively-sorted list.
*/
return strcmp(idx_or_pack_name, idx_name);
}
int midx_contains_pack(struct multi_pack_index *m, const char *idx_or_pack_name)
{
uint32_t first = 0, last = m->num_packs;
while (first < last) {
uint32_t mid = first + (last - first) / 2;
const char *current;
int cmp;
current = m->pack_names[mid];
cmp = cmp_idx_or_pack_name(idx_or_pack_name, current);
if (!cmp)
return 1;
if (cmp > 0) {
first = mid + 1;
continue;
}
last = mid;
}
return 0;
}
int prepare_multi_pack_index_one(struct repository *r, const char *object_dir, int local)
{
struct multi_pack_index *m;
struct multi_pack_index *m_search;
prepare_repo_settings(r);
if (!r->settings.core_multi_pack_index)
return 0;
for (m_search = r->objects->multi_pack_index; m_search; m_search = m_search->next)
if (!strcmp(object_dir, m_search->object_dir))
return 1;
m = load_multi_pack_index(object_dir, local);
if (m) {
midx: traverse the local MIDX first When a repository has an alternate object directory configured, callers can traverse through each alternate's MIDX by walking the '->next' pointer. But, when 'prepare_multi_pack_index_one()' loads multiple MIDXs, it places the new ones at the front of this pointer chain, not at the end. This can be confusing for callers such as 'git repack -ad', causing test failures like in t7700.6 with 'GIT_TEST_MULTI_PACK_INDEX=1'. The occurs when dropping a pack known to the local MIDX with alternates configured that have their own MIDX. Since the alternate's MIDX is returned via 'get_multi_pack_index()', 'midx_contains_pack()' returns true (which is correct, since it traverses through the '->next' pointer to find the MIDX in the chain that does contain the requested object). But, we call 'clear_midx_file()' on 'the_repository', which drops the MIDX at the path of the first MIDX in the chain, which (in the case of t7700.6 is the one in the alternate). This patch addresses that by: - placing the local MIDX first in the chain when calling 'prepare_multi_pack_index_one()', and - introducing a new 'get_local_multi_pack_index()', which explicitly returns the repository-local MIDX, if any. Don't impose an additional order on the MIDX's '->next' pointer beyond that the first item in the chain must be local if one exists so that we avoid a quadratic insertion. Likewise, use 'get_local_multi_pack_index()' in 'remove_redundant_pack()' to fix the formerly broken t7700.6 when run with 'GIT_TEST_MULTI_PACK_INDEX=1'. Finally, note that the MIDX ordering invariant is only preserved by the insertion order in 'prepare_packed_git()', which traverses through the ODB's '->next' pointer, meaning we visit the local object store first. This fragility makes this an undesirable long-term solution if more callers are added, but it is acceptable for now since this is the only caller. Helped-by: Jeff King <peff@peff.net> Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Taylor Blau <me@ttaylorr.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2 years ago
struct multi_pack_index *mp = r->objects->multi_pack_index;
if (mp) {
m->next = mp->next;
mp->next = m;
} else
r->objects->multi_pack_index = m;
return 1;
}
return 0;
}
static size_t write_midx_header(struct hashfile *f,
unsigned char num_chunks,
uint32_t num_packs)
{
hashwrite_be32(f, MIDX_SIGNATURE);
hashwrite_u8(f, MIDX_VERSION);
hashwrite_u8(f, oid_version(the_hash_algo));
hashwrite_u8(f, num_chunks);
hashwrite_u8(f, 0); /* unused */
hashwrite_be32(f, num_packs);
return MIDX_HEADER_SIZE;
}
struct pack_info {
uint32_t orig_pack_int_id;
char *pack_name;
struct packed_git *p;
unsigned expired : 1;
};
static int pack_info_compare(const void *_a, const void *_b)
{
struct pack_info *a = (struct pack_info *)_a;
struct pack_info *b = (struct pack_info *)_b;
return strcmp(a->pack_name, b->pack_name);
}
static int idx_or_pack_name_cmp(const void *_va, const void *_vb)
{
const char *pack_name = _va;
const struct pack_info *compar = _vb;
return cmp_idx_or_pack_name(pack_name, compar->pack_name);
}
struct write_midx_context {
struct pack_info *info;
uint32_t nr;
uint32_t alloc;
struct multi_pack_index *m;
struct progress *progress;
unsigned pack_paths_checked;
struct pack_midx_entry *entries;
uint32_t entries_nr;
uint32_t *pack_perm;
uint32_t *pack_order;
unsigned large_offsets_needed:1;
uint32_t num_large_offsets;
int preferred_pack_idx;
struct string_list *to_include;
};
static void add_pack_to_midx(const char *full_path, size_t full_path_len,
const char *file_name, void *data)
{
struct write_midx_context *ctx = data;
if (ends_with(file_name, ".idx")) {
display_progress(ctx->progress, ++ctx->pack_paths_checked);
/*
* Note that at most one of ctx->m and ctx->to_include are set,
* so we are testing midx_contains_pack() and
* string_list_has_string() independently (guarded by the
* appropriate NULL checks).
*
* We could support passing to_include while reusing an existing
* MIDX, but don't currently since the reuse process drags
* forward all packs from an existing MIDX (without checking
* whether or not they appear in the to_include list).
*
* If we added support for that, these next two conditional
* should be performed independently (likely checking
* to_include before the existing MIDX).
*/
if (ctx->m && midx_contains_pack(ctx->m, file_name))
return;
else if (ctx->to_include &&
!string_list_has_string(ctx->to_include, file_name))
return;