bmap: Initial scaffolding for on-disk block map

This is pretty exploratory work, so I'm going to try to be less thorough
in commit messages until the dust settles.

---

New tag for gbmapdelta:

  LFS3_TAG_GBMAPDELTA   0x0104  v--- ---1 ---- -1rr

New tags for in-bmap block types:

  LFS3_TAG_BMRANGE      0x033u  v--- --11 --11 uuuu
  LFS3_TAG_BMFREE       0x0330  v--- --11 --11 ----
  LFS3_TAG_BMINFLIGHT   0x0331  v--- --11 --11 ---1
  LFS3_TAG_BMINUSE      0x0332  v--- --11 --11 --1-
  LFS3_TAG_BMBAD        0x0333  v--- --11 --11 --11
  LFS3_TAG_BMERASED     0x0334  v--- --11 --11 -1--

New gstate decoding for gbmap:

  .---+- -+- -+- -+- -. cursor: 1 leb128  <=5 bytes
  | cursor            | known:  1 leb128  <=5 bytes
  +---+- -+- -+- -+- -+ block:  1 leb128  <=5 bytes
  | known             | trunk:  1 leb128  <=4 bytes
  +---+- -+- -+- -+- -+ cksum:  1 le32    4 bytes
  | block             | total:            23 bytes
  +---+- -+- -+- -+- -'
  | trunk         |
  +---+- -+- -+- -+
  |     cksum     |
  '---+---+---+---'

New bmap node revdbg string:

  vvv---- -111111- -11---1- -11---1-  (62 62 7e v0  bb~r)  bmap node

New mount/format/info flags (still unsure about these):

  LFS3_M_BMAPMODE     0x03000000  On-disk block map mode
  LFS3_M_BMAPNONE     0x00000000  Don't use the bmap
  LFS3_M_BMAPCACHE    0x01000000  Use the bmap to cache lookahead scans
  LFS3_M_BMAPSLOW     0x02000000  Use the slow bmap algorithm
  LFS3_M_BMAPFAST     0x03000000  Use the fast bmap algorithm

New gbmap wcompat flag:

  LFS3_WCOMPAT_GBMAP  0x00002000  Global block-map in use
This commit is contained in:
Christopher Haster
2025-07-23 15:32:35 -05:00
parent 238dbc705d
commit 88180b6081
13 changed files with 755 additions and 58 deletions
+45 -9
View File
@@ -39,6 +39,7 @@ TAG_NAMELIMIT = 0x0039 # 0x0039 v--- ---- --11 1--1
TAG_FILELIMIT = 0x003a # 0x003a v--- ---- --11 1-1-
TAG_GDELTA = 0x0100 ## 0x01tt v--- ---1 -ttt ttrr
TAG_GRMDELTA = 0x0100 # 0x0100 v--- ---1 ---- ----
TAG_GBMAPDELTA = 0x0104 # 0x0104 v--- ---1 ---- -1rr
TAG_NAME = 0x0200 ## 0x02tt v--- --1- -ttt tttt
TAG_BNAME = 0x0200 # 0x0200 v--- --1- ---- ----
TAG_REG = 0x0201 # 0x0201 v--- --1- ---- ---1
@@ -56,6 +57,12 @@ TAG_BTREE = 0x031c # 0x031c v--- --11 ---1 11rr
TAG_MROOT = 0x0321 # 0x032r v--- --11 --1- --rr
TAG_MDIR = 0x0325 # 0x0324 v--- --11 --1- -1rr
TAG_MTREE = 0x032c # 0x032c v--- --11 --1- 11rr
TAG_BMRANGE = 0x0330 # 0x033u v--- --11 --11 uuuu
TAG_BMFREE = 0x0330 # 0x0330 v--- --11 --11 ----
TAG_BMINFLIGHT = 0x0331 # 0x0331 v--- --11 --11 ---1
TAG_BMINUSE = 0x0332 # 0x0332 v--- --11 --11 --1-
TAG_BMBAD = 0x0333 # 0x0333 v--- --11 --11 --11
TAG_BMERASED = 0x0334 # 0x0334 v--- --11 --11 -1--
TAG_ATTR = 0x0400 ## 0x04aa v--- -1-a -aaa aaaa
TAG_UATTR = 0x0400 # 0x04aa v--- -1-- -aaa aaaa
TAG_SATTR = 0x0500 # 0x05aa v--- -1-1 -aaa aaaa
@@ -340,6 +347,7 @@ def tagrepr(tag, weight=None, size=None, *,
return '%s%s%s%s' % (
'shrub' if tag & TAG_SHRUB else '',
'grmdelta' if (tag & 0xfff) == TAG_GRMDELTA
else 'gbmapdelta' if (tag & 0xfff) == TAG_GBMAPDELTA
else 'gdelta 0x%02x' % (tag & 0xff),
' w%d' % weight if weight else '',
' %s' % size if size is not None else '')
@@ -369,6 +377,13 @@ def tagrepr(tag, weight=None, size=None, *,
else 'mroot' if (tag & 0xfff) == TAG_MROOT
else 'mdir' if (tag & 0xfff) == TAG_MDIR
else 'mtree' if (tag & 0xfff) == TAG_MTREE
else 'bmfree' if (tag & 0xfff) == TAG_BMFREE
else 'bminflight' if (tag & 0xfff) == TAG_BMINFLIGHT
else 'bminuse' if (tag & 0xfff) == TAG_BMINUSE
else 'bmbad' if (tag & 0xfff) == TAG_BMBAD
else 'bmerased' if (tag & 0xfff) == TAG_BMERASED
else 'bmrange 0x%x' % (tag & 0xf)
if (tag & 0xff0) == TAG_BMRANGE
else 'struct 0x%02x' % (tag & 0xff),
' w%d' % weight if weight else '',
' %s' % size if size is not None else '')
@@ -2615,8 +2630,9 @@ class Config:
# lazy gstate object
class Gstate:
def __init__(self, mtree):
def __init__(self, mtree, config):
self.mtree = mtree
self.config = config
# lookup a specific tag
def lookup(self, tag=None, mask=None):
@@ -2692,7 +2708,7 @@ class Gstate:
tag = None
mask = None
def __init__(self, mtree, tag, gdeltas):
def __init__(self, mtree, config, tag, gdeltas):
# replace tag with what we find
self.tag = tag
# keep track of gdeltas for debugging
@@ -2742,8 +2758,8 @@ class Gstate:
class Gcksum(Gstate):
tag = TAG_GCKSUMDELTA
def __init__(self, mtree, tag, gdeltas):
super().__init__(mtree, tag, gdeltas)
def __init__(self, mtree, config, tag, gdeltas):
super().__init__(mtree, config, tag, gdeltas)
self.gcksum = fromle32(self.data)
def __int__(self):
@@ -2756,8 +2772,8 @@ class Gstate:
class Grm(Gstate):
tag = TAG_GRMDELTA
def __init__(self, mtree, tag, gdeltas):
super().__init__(mtree, tag, gdeltas)
def __init__(self, mtree, config, tag, gdeltas):
super().__init__(mtree, config, tag, gdeltas)
queue = []
d = 0
for _ in range(2):
@@ -2775,6 +2791,26 @@ class Gstate:
def repr(self):
return 'grm [%s]' % ', '.join(mid.repr() for mid in self.queue)
# the global block map
class Gbmap(Gstate):
tag = TAG_GBMAPDELTA
def __init__(self, mtree, config, tag, gdeltas):
super().__init__(mtree, config, tag, gdeltas)
d = 0
self.cursor, d_ = fromleb128(self.data, d); d += d_
self.known, d_ = fromleb128(self.data, d); d += d_
block, trunk, cksum, d_ = frombranch(self.data, d); d += d_
self.btree = Btree.fetchck(
mtree.bd, block, trunk,
config.geometry.block_count,
cksum)
def repr(self):
return 'gbmap %s k%s+%s' % (
self.btree.addr(),
self.known, self.cursor)
# keep track of known gstate
_known = [g for g in Gstate.__subclasses__() if g.tag is not None]
@@ -2783,10 +2819,10 @@ class Gstate:
# known config?
for g in self._known:
if (g.tag & ~(g.mask or 0)) == (tag & ~(g.mask or 0)):
return g(self.mtree, tag, gdeltas)
return g(self.mtree, self.config, tag, gdeltas)
# otherwise return a marker class
else:
return Unknown(self.mtree, tag, gdeltas)
return self.Unknown(self.mtree, self.config, tag, gdeltas)
# create cached accessors for known gstate
def _parser(g):
@@ -2807,7 +2843,7 @@ class Lfs3:
# create lazy config/gstate objects
self.config = config or Config(self.mroot)
self.gstate = gstate or Gstate(self.mtree)
self.gstate = gstate or Gstate(self.mtree, self.config)
# go ahead and fetch some expected fields
self.version = self.config.version