From: "Shawn O. Pearce" <spearce@spearce.org>
To: Junio C Hamano <junkio@cox.net>
Cc: git@vger.kernel.org
Subject: [PATCH 3/17] Refactor packed_git to prepare for sliding mmap windows.
Date: Sat, 23 Dec 2006 02:33:44 -0500 [thread overview]
Message-ID: <20061223073344.GD9837@spearce.org> (raw)
In-Reply-To: <53b67707929c7f051f6d384c5d96e653bfa8419c.1166857884.git.spearce@spearce.org>
The idea behind the sliding mmap window pack reader implementation
is to have multiple mmap regions active against the same pack file,
thereby allowing the process to mmap in only the active/hot sections
of the pack and reduce overall virtual address space usage.
To implement this we need to refactor the mmap related data
(pack_base, pack_use_cnt) out of struct packed_git and move them
into a new struct pack_window.
We are refactoring the code to support a single struct pack_window
per packfile, thereby emulating the prior behavior of mmap'ing the
entire pack file.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
---
builtin-pack-objects.c | 4 +-
cache.h | 13 +++++++++--
pack-check.c | 6 ++--
sha1_file.c | 53 ++++++++++++++++++++++-------------------------
4 files changed, 40 insertions(+), 36 deletions(-)
diff --git a/builtin-pack-objects.c b/builtin-pack-objects.c
index 9e15beb..4a00a12 100644
--- a/builtin-pack-objects.c
+++ b/builtin-pack-objects.c
@@ -438,7 +438,7 @@ static unsigned long write_object(struct sha1file *f,
}
use_packed_git(p);
- buf = (char *) p->pack_base
+ buf = p->windows->base
+ entry->in_pack_offset
+ entry->in_pack_header_size;
datalen = find_packed_object_size(p, entry->in_pack_offset)
@@ -943,7 +943,7 @@ static void check_object(struct object_entry *entry)
struct object_entry *base_entry = NULL;
use_packed_git(p);
- buf = p->pack_base;
+ buf = p->windows->base;
buf += entry->in_pack_offset;
/* We want in_pack_type even if we do not reuse delta.
diff --git a/cache.h b/cache.h
index ad94c3f..bf1d776 100644
--- a/cache.h
+++ b/cache.h
@@ -336,14 +336,21 @@ extern struct alternate_object_database {
} *alt_odb_list;
extern void prepare_alt_odb(void);
+struct pack_window {
+ struct pack_window *next;
+ unsigned char *base;
+ off_t offset;
+ size_t len;
+ unsigned int last_used;
+ unsigned int inuse_cnt;
+};
+
extern struct packed_git {
struct packed_git *next;
unsigned long index_size;
unsigned long pack_size;
+ struct pack_window *windows;
unsigned int *index_base;
- void *pack_base;
- unsigned int pack_last_used;
- unsigned int pack_use_cnt;
int pack_local;
unsigned char sha1[20];
/* something like ".git/objects/pack/xxxxx.pack" */
diff --git a/pack-check.c b/pack-check.c
index 491bad2..761cc85 100644
--- a/pack-check.c
+++ b/pack-check.c
@@ -13,7 +13,8 @@ static int verify_packfile(struct packed_git *p)
int nr_objects, err, i;
/* Header consistency check */
- hdr = p->pack_base;
+ pack_base = p->windows->base;
+ hdr = (struct pack_header*)pack_base;
if (hdr->hdr_signature != htonl(PACK_SIGNATURE))
return error("Packfile %s signature mismatch", p->pack_name);
if (!pack_version_ok(hdr->hdr_version))
@@ -26,7 +27,6 @@ static int verify_packfile(struct packed_git *p)
num_packed_objects(p));
SHA1_Init(&ctx);
- pack_base = p->pack_base;
SHA1_Update(&ctx, pack_base, pack_size - 20);
SHA1_Final(sha1, &ctx);
if (hashcmp(sha1, (unsigned char *)pack_base + pack_size - 20))
@@ -78,7 +78,7 @@ static void show_pack_info(struct packed_git *p)
int nr_objects, i;
unsigned int chain_histogram[MAX_CHAIN];
- hdr = p->pack_base;
+ hdr = (struct pack_header*)p->windows->base;
nr_objects = ntohl(hdr->hdr_entries);
memset(chain_histogram, 0, sizeof(chain_histogram));
diff --git a/sha1_file.c b/sha1_file.c
index 4183f59..8377874 100644
--- a/sha1_file.c
+++ b/sha1_file.c
@@ -455,21 +455,23 @@ static int unuse_one_packed_git(void)
struct packed_git *p, *lru = NULL;
for (p = packed_git; p; p = p->next) {
- if (p->pack_use_cnt || !p->pack_base)
+ if (!p->windows || p->windows->inuse_cnt)
continue;
- if (!lru || p->pack_last_used < lru->pack_last_used)
+ if (!lru || p->windows->last_used < lru->windows->last_used)
lru = p;
}
if (!lru)
return 0;
- munmap(lru->pack_base, lru->pack_size);
- lru->pack_base = NULL;
+ munmap(lru->windows->base, lru->windows->len);
+ free(lru->windows);
+ lru->windows = NULL;
return 1;
}
void unuse_packed_git(struct packed_git *p)
{
- p->pack_use_cnt--;
+ if (p->windows)
+ p->windows->inuse_cnt--;
}
int use_packed_git(struct packed_git *p)
@@ -482,10 +484,10 @@ int use_packed_git(struct packed_git *p)
die("packfile %s not a regular file", p->pack_name);
p->pack_size = st.st_size;
}
- if (!p->pack_base) {
+ if (!p->windows) {
int fd;
struct stat st;
- void *map;
+ struct pack_window *win;
struct pack_header *hdr;
pack_mapped += p->pack_size;
@@ -500,16 +502,18 @@ int use_packed_git(struct packed_git *p)
}
if (st.st_size != p->pack_size)
die("packfile %s size mismatch.", p->pack_name);
- map = mmap(NULL, p->pack_size, PROT_READ, MAP_PRIVATE, fd, 0);
+ win = xcalloc(1, sizeof(*win));
+ win->len = p->pack_size;
+ win->base = mmap(NULL, p->pack_size, PROT_READ, MAP_PRIVATE, fd, 0);
close(fd);
- if (map == MAP_FAILED)
+ if (win->base == MAP_FAILED)
die("packfile %s cannot be mapped.", p->pack_name);
- p->pack_base = map;
+ p->windows = win;
/* Check if we understand this pack file. If we don't we're
* likely too old to handle it.
*/
- hdr = map;
+ hdr = (struct pack_header*)win->base;
if (hdr->hdr_signature != htonl(PACK_SIGNATURE))
die("packfile %s isn't actually a pack.", p->pack_name);
if (!pack_version_ok(hdr->hdr_version))
@@ -522,13 +526,13 @@ int use_packed_git(struct packed_git *p)
*/
if (hashcmp((unsigned char *)(p->index_base) +
p->index_size - 40,
- (unsigned char *)p->pack_base +
+ p->windows->base +
p->pack_size - 20)) {
die("packfile %s does not match index.", p->pack_name);
}
}
- p->pack_last_used = pack_used_ctr++;
- p->pack_use_cnt++;
+ p->windows->last_used = pack_used_ctr++;
+ p->windows->inuse_cnt++;
return 0;
}
@@ -557,10 +561,7 @@ struct packed_git *add_packed_git(char *path, int path_len, int local)
p->index_size = idx_size;
p->pack_size = st.st_size;
p->index_base = idx_map;
- p->next = NULL;
- p->pack_base = NULL;
- p->pack_last_used = 0;
- p->pack_use_cnt = 0;
+ p->windows = NULL;
p->pack_local = local;
if ((path_len > 44) && !get_sha1_hex(path + path_len - 44, sha1))
hashcpy(p->sha1, sha1);
@@ -590,10 +591,7 @@ struct packed_git *parse_pack_index_file(const unsigned char *sha1, char *idx_pa
p->index_size = idx_size;
p->pack_size = 0;
p->index_base = idx_map;
- p->next = NULL;
- p->pack_base = NULL;
- p->pack_last_used = 0;
- p->pack_use_cnt = 0;
+ p->windows = NULL;
hashcpy(p->sha1, sha1);
return p;
}
@@ -882,7 +880,7 @@ static unsigned long get_delta_base(struct packed_git *p,
unsigned long delta_obj_offset,
unsigned long *base_obj_offset)
{
- unsigned char *base_info = (unsigned char *) p->pack_base + offset;
+ unsigned char *base_info = p->windows->base + offset;
unsigned long base_offset;
/* there must be at least 20 bytes left regardless of delta type */
@@ -949,7 +947,7 @@ static int packed_delta_info(struct packed_git *p,
memset(&stream, 0, sizeof(stream));
- stream.next_in = (unsigned char *) p->pack_base + offset;
+ stream.next_in = p->windows->base + offset;
stream.avail_in = p->pack_size - offset;
stream.next_out = delta_head;
stream.avail_out = sizeof(delta_head);
@@ -984,8 +982,7 @@ static unsigned long unpack_object_header(struct packed_git *p, unsigned long of
if (p->pack_size <= offset)
die("object offset outside of pack file");
- used = unpack_object_header_gently((unsigned char *)p->pack_base +
- offset,
+ used = unpack_object_header_gently(p->windows->base + offset,
p->pack_size - offset, type, sizep);
if (!used)
die("object offset outside of pack file");
@@ -1031,7 +1028,7 @@ void packed_object_info_detail(struct packed_git *p,
if (p->pack_size <= offset + 20)
die("pack file %s records an incomplete delta base",
p->pack_name);
- next_sha1 = (unsigned char *) p->pack_base + offset;
+ next_sha1 = p->windows->base + offset;
if (*delta_chain_length == 0)
hashcpy(base_sha1, next_sha1);
offset = find_pack_entry_one(next_sha1, p);
@@ -1081,7 +1078,7 @@ static void *unpack_compressed_entry(struct packed_git *p,
buffer = xmalloc(size + 1);
buffer[size] = 0;
memset(&stream, 0, sizeof(stream));
- stream.next_in = (unsigned char*)p->pack_base + offset;
+ stream.next_in = p->windows->base + offset;
stream.avail_in = p->pack_size - offset;
stream.next_out = buffer;
stream.avail_out = size;
--
1.4.4.3.g87d8
next prev parent reply other threads:[~2006-12-23 7:33 UTC|newest]
Thread overview: 25+ messages / expand[flat|nested] mbox.gz Atom feed top
[not found] <53b67707929c7f051f6d384c5d96e653bfa8419c.1166857884.git.spearce@spearce.org>
2006-12-23 7:33 ` [PATCH 1/17] Replace unpack_entry_gently with unpack_entry Shawn O. Pearce
2006-12-23 7:33 ` [PATCH 2/17] Introduce new config option for mmap limit Shawn O. Pearce
2006-12-23 7:33 ` Shawn O. Pearce [this message]
2006-12-23 7:33 ` [PATCH 4/17] Use off_t for index and pack file lengths Shawn O. Pearce
2006-12-23 7:33 ` [PATCH 5/17] Create read_or_die utility routine Shawn O. Pearce
2006-12-23 7:34 ` [PATCH 6/17] Refactor how we open pack files to prepare for multiple windows Shawn O. Pearce
2006-12-23 7:34 ` [PATCH 7/17] Replace use_packed_git with window cursors Shawn O. Pearce
2006-12-23 7:34 ` [PATCH 8/17] Loop over pack_windows when inflating/accessing data Shawn O. Pearce
2006-12-23 7:34 ` [PATCH 9/17] Document why header parsing won't exceed a window Shawn O. Pearce
2006-12-23 7:34 ` [PATCH 10/17] Unmap individual windows rather than entire files Shawn O. Pearce
2006-12-23 7:34 ` [PATCH 11/17] Fully activate the sliding window pack access Shawn O. Pearce
2006-12-23 18:44 ` Linus Torvalds
2006-12-23 19:34 ` Eric Blake
2006-12-24 0:58 ` Johannes Schindelin
2006-12-23 19:45 ` Junio C Hamano
2006-12-23 20:10 ` Linus Torvalds
2006-12-24 1:23 ` Johannes Schindelin
2006-12-24 2:23 ` Shawn Pearce
2006-12-24 2:35 ` Shawn Pearce
2006-12-23 7:34 ` [PATCH 12/17] Load core configuration in git-verify-pack Shawn O. Pearce
2006-12-23 7:34 ` [PATCH 13/17] Ensure core.packedGitWindowSize cannot be less than 2 pages Shawn O. Pearce
2006-12-23 7:34 ` [PATCH 14/17] Improve error message when packfile mmap fails Shawn O. Pearce
2006-12-23 7:34 ` [PATCH 15/17] Support unmapping windows on 'temporary' packfiles Shawn O. Pearce
2006-12-23 7:34 ` [PATCH 16/17] Create pack_report() as a debugging aid Shawn O. Pearce
2006-12-23 7:34 ` [PATCH 17/17] Test suite for sliding window mmap implementation Shawn O. Pearce
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20061223073344.GD9837@spearce.org \
--to=spearce@spearce.org \
--cc=git@vger.kernel.org \
--cc=junkio@cox.net \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).