]> code.ossystems Code Review - openembedded-core.git/commitdiff
classes/sstate: Update output hash
authorJoshua Watt <jpewhacker@gmail.com>
Mon, 21 Jan 2019 22:39:19 +0000 (16:39 -0600)
committerRichard Purdie <richard.purdie@linuxfoundation.org>
Mon, 21 Jan 2019 23:45:47 +0000 (23:45 +0000)
Updates the output hash calculation for determining if tasks are
equivalent. The new algorithm does the following based on feedback:
 1) The output hash function was moved to the OE library.
 2) All files are printed in a single line tabular format
 3) Prints the file type and mode in a user-friendly ls-like format
 4) Includes the file owner and group (by name, not ID). These are only
    included if the task is run under pseudo since that is the only time
    they can be consistently determined.
 5) File size is included for regular files

Signed-off-by: Joshua Watt <JPEWhacker@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
meta/classes/sstate.bbclass
meta/lib/oe/sstatesig.py

index 763fce07f95a8e9ad3ac228a53ab626f7e021dfc..2f0bbd2d7dfe2d19c725e71dab41e4d6716fc80a 100644 (file)
@@ -83,9 +83,9 @@ SSTATE_SIG_PASSPHRASE ?= ""
 # Whether to verify the GnUPG signatures when extracting sstate archives
 SSTATE_VERIFY_SIG ?= "0"
 
-SSTATE_HASHEQUIV_METHOD ?= "OEOuthashBasic"
-SSTATE_HASHEQUIV_METHOD[doc] = "The function used to calculate the output hash \
-    for a task, which in turn is used to determine equivalency. \
+SSTATE_HASHEQUIV_METHOD ?= "oe.sstatesig.OEOuthashBasic"
+SSTATE_HASHEQUIV_METHOD[doc] = "The fully-qualified function used to calculate \
+    the output hash for a task, which in turn is used to determine equivalency. \
     "
 
 SSTATE_HASHEQUIV_SERVER ?= ""
@@ -782,65 +782,6 @@ python sstate_sign_package () {
                            d.getVar('SSTATE_SIG_PASSPHRASE'), armor=False)
 }
 
-def OEOuthashBasic(path, sigfile, task, d):
-    import hashlib
-    import stat
-
-    def update_hash(s):
-        s = s.encode('utf-8')
-        h.update(s)
-        if sigfile:
-            sigfile.write(s)
-
-    h = hashlib.sha256()
-    prev_dir = os.getcwd()
-
-    try:
-        os.chdir(path)
-
-        update_hash("OEOuthashBasic\n")
-
-        # It is only currently useful to get equivalent hashes for things that
-        # can be restored from sstate. Since the sstate object is named using
-        # SSTATE_PKGSPEC and the task name, those should be included in the
-        # output hash calculation.
-        update_hash("SSTATE_PKGSPEC=%s\n" % d.getVar('SSTATE_PKGSPEC'))
-        update_hash("task=%s\n" % task)
-
-        for root, dirs, files in os.walk('.', topdown=True):
-            # Sort directories and files to ensure consistent ordering
-            dirs.sort()
-            files.sort()
-
-            for f in files:
-                path = os.path.join(root, f)
-                s = os.lstat(path)
-
-                # Hash file path
-                update_hash(path + '\n')
-
-                # Hash file mode
-                update_hash("\tmode=0x%x\n" % stat.S_IMODE(s.st_mode))
-                update_hash("\ttype=0x%x\n" % stat.S_IFMT(s.st_mode))
-
-                if stat.S_ISBLK(s.st_mode) or stat.S_ISBLK(s.st_mode):
-                    # Hash device major and minor
-                    update_hash("\tdev=%d,%d\n" % (os.major(s.st_rdev), os.minor(s.st_rdev)))
-                elif stat.S_ISLNK(s.st_mode):
-                    # Hash symbolic link
-                    update_hash("\tsymlink=%s\n" % os.readlink(path))
-                else:
-                    fh = hashlib.sha256()
-                    # Hash file contents
-                    with open(path, 'rb') as d:
-                        for chunk in iter(lambda: d.read(4096), b""):
-                            fh.update(chunk)
-                    update_hash("\tdigest=%s\n" % fh.hexdigest())
-    finally:
-        os.chdir(prev_dir)
-
-    return h.hexdigest()
-
 python sstate_report_unihash() {
     report_unihash = getattr(bb.parse.siggen, 'report_unihash', None)
 
index e0eb87e29f04d38b58e85beca113bd239ba832e4..a83af519ec115983ebf9464c4695931e2f39aa0d 100644 (file)
@@ -270,7 +270,7 @@ class SignatureGeneratorOEEquivHash(SignatureGeneratorOEBasicHash):
         super().init_rundepcheck(data)
         self.server = data.getVar('SSTATE_HASHEQUIV_SERVER')
         self.method = data.getVar('SSTATE_HASHEQUIV_METHOD')
-        self.unihashes = bb.persist_data.persist('SSTATESIG_UNIHASH_CACHE_v1_' + self.method, data)
+        self.unihashes = bb.persist_data.persist('SSTATESIG_UNIHASH_CACHE_v1_' + self.method.replace('.', '_'), data)
 
     def get_taskdata(self):
         return (self.server, self.method) + super().get_taskdata()
@@ -355,6 +355,7 @@ class SignatureGeneratorOEEquivHash(SignatureGeneratorOEBasicHash):
         import json
         import tempfile
         import base64
+        import importlib
 
         taskhash = d.getVar('BB_TASKHASH')
         unihash = d.getVar('BB_UNIHASH')
@@ -376,11 +377,14 @@ class SignatureGeneratorOEEquivHash(SignatureGeneratorOEBasicHash):
         sigfile_link = "depsig.do_%s" % task
 
         try:
-            call = self.method + '(path, sigfile, task, d)'
             sigfile = open(os.path.join(tempdir, sigfile_name), 'w+b')
+
             locs = {'path': path, 'sigfile': sigfile, 'task': task, 'd': d}
 
-            outhash = bb.utils.better_eval(call, locs)
+            (module, method) = self.method.rsplit('.', 1)
+            locs['method'] = getattr(importlib.import_module(module), method)
+
+            outhash = bb.utils.better_eval('method(path, sigfile, task, d)', locs)
 
             try:
                 url = '%s/v1/equivalent' % self.server
@@ -581,4 +585,133 @@ def find_sstate_manifest(taskdata, taskdata2, taskname, d, multilibcache):
     bb.warn("Manifest %s not found in %s (variant '%s')?" % (manifest, d2.expand(" ".join(pkgarchs)), variant))
     return None, d2
 
+def OEOuthashBasic(path, sigfile, task, d):
+    """
+    Basic output hash function
+
+    Calculates the output hash of a task by hashing all output file metadata,
+    and file contents.
+    """
+    import hashlib
+    import stat
+    import pwd
+    import grp
+
+    def update_hash(s):
+        s = s.encode('utf-8')
+        h.update(s)
+        if sigfile:
+            sigfile.write(s)
+
+    h = hashlib.sha256()
+    prev_dir = os.getcwd()
+    include_owners = os.environ.get('PSEUDO_DISABLED') == '0'
+
+    try:
+        os.chdir(path)
+
+        update_hash("OEOuthashBasic\n")
+
+        # It is only currently useful to get equivalent hashes for things that
+        # can be restored from sstate. Since the sstate object is named using
+        # SSTATE_PKGSPEC and the task name, those should be included in the
+        # output hash calculation.
+        update_hash("SSTATE_PKGSPEC=%s\n" % d.getVar('SSTATE_PKGSPEC'))
+        update_hash("task=%s\n" % task)
+
+        for root, dirs, files in os.walk('.', topdown=True):
+            # Sort directories to ensure consistent ordering when recursing
+            dirs.sort()
+            files.sort()
+
+            def process(path):
+                s = os.lstat(path)
+
+                if stat.S_ISDIR(s.st_mode):
+                    update_hash('d')
+                elif stat.S_ISCHR(s.st_mode):
+                    update_hash('c')
+                elif stat.S_ISBLK(s.st_mode):
+                    update_hash('b')
+                elif stat.S_ISSOCK(s.st_mode):
+                    update_hash('s')
+                elif stat.S_ISLNK(s.st_mode):
+                    update_hash('l')
+                elif stat.S_ISFIFO(s.st_mode):
+                    update_hash('p')
+                else:
+                    update_hash('-')
+
+                def add_perm(mask, on, off='-'):
+                    if mask & s.st_mode:
+                        update_hash(on)
+                    else:
+                        update_hash(off)
+
+                add_perm(stat.S_IRUSR, 'r')
+                add_perm(stat.S_IWUSR, 'w')
+                if stat.S_ISUID & s.st_mode:
+                    add_perm(stat.S_IXUSR, 's', 'S')
+                else:
+                    add_perm(stat.S_IXUSR, 'x')
+
+                add_perm(stat.S_IRGRP, 'r')
+                add_perm(stat.S_IWGRP, 'w')
+                if stat.S_ISGID & s.st_mode:
+                    add_perm(stat.S_IXGRP, 's', 'S')
+                else:
+                    add_perm(stat.S_IXGRP, 'x')
+
+                add_perm(stat.S_IROTH, 'r')
+                add_perm(stat.S_IWOTH, 'w')
+                if stat.S_ISVTX & s.st_mode:
+                    update_hash('t')
+                else:
+                    add_perm(stat.S_IXOTH, 'x')
+
+                if include_owners:
+                    update_hash(" %10s" % pwd.getpwuid(s.st_uid).pw_name)
+                    update_hash(" %10s" % grp.getgrgid(s.st_gid).gr_name)
+
+                update_hash(" ")
+                if stat.S_ISBLK(s.st_mode) or stat.S_ISCHR(s.st_mode):
+                    update_hash("%9s" % ("%d.%d" % (os.major(s.st_rdev), os.minor(s.st_rdev))))
+                else:
+                    update_hash(" " * 9)
+
+                update_hash(" ")
+                if stat.S_ISREG(s.st_mode):
+                    update_hash("%10d" % s.st_size)
+                else:
+                    update_hash(" " * 10)
+
+                update_hash(" ")
+                fh = hashlib.sha256()
+                if stat.S_ISREG(s.st_mode):
+                    # Hash file contents
+                    with open(path, 'rb') as d:
+                        for chunk in iter(lambda: d.read(4096), b""):
+                            fh.update(chunk)
+                    update_hash(fh.hexdigest())
+                else:
+                    update_hash(" " * len(fh.hexdigest()))
+
+                update_hash(" %s" % path)
+
+                if stat.S_ISLNK(s.st_mode):
+                    update_hash(" -> %s" % os.readlink(path))
+
+                update_hash("\n")
+
+            # Process this directory and all its child files
+            process(root)
+            for f in files:
+                if f == 'fixmepath':
+                    continue
+                process(os.path.join(root, f))
+    finally:
+        os.chdir(prev_dir)
+
+    return h.hexdigest()
+