Add PPDProvider barebones implementation and associated cache skeleton.

chromeos/printing/ppd_cache.cc

Index: chromeos/printing/ppd_cache.cc
diff --git a/chromeos/printing/ppd_cache.cc b/chromeos/printing/ppd_cache.cc
new file mode 100644
index 0000000000000000000000000000000000000000..667d5dbc2c221a82663246884fda648e5e920c2d
--- /dev/null
+++ b/chromeos/printing/ppd_cache.cc
@@ -0,0 +1,174 @@
+// Copyright 2016 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include <vector>
+
+#include "base/files/file_util.h"
+#include "base/memory/ptr_util.h"
+#include "base/path_service.h"
+#include "base/strings/string_util.h"
+#include "chromeos/printing/ppd_cache.h"
+#include "crypto/sha2.h"
+#include "net/base/io_buffer.h"
+#include "net/filter/filter.h"
+#include "net/filter/gzip_header.h"
+
+using base::File;
+using base::FilePath;
+using base::Optional;
+using net::Filter;
+using std::string;
+using std::unique_ptr;
+using std::vector;
+
+namespace chromeos {
+namespace printing {
+namespace {
+
+// Return the ASCII character representing the given nibble, in hex.
+char HexifyNibble(int nibble) {

[skau, 2016/10/14 23:51:49]

Is this equivalent to HexEncode?
https://cs.chromium.org/chromium/src/base/strings/string_number_conversions.h...

[Carlson, 2016/10/17 16:43:20]

I spent seriously like 15 minutes looking for that before rolling my own because
I knew it had to exist, but somehow never found it.

Thanks!

+  DCHECK(nibble >= 0 && nibble <= 0xf);
+  if (nibble <= 9) {
+    return '0' + nibble;
+  }
+  return 'a' + (nibble - 10);
+}
+
+// Return a ascii-hex version of the data in data, interpreted as a list of
+// bytes.  Leading zeros are not removed.  For example, a 3-byte string
+// containing the bytes {0x0a, 0x3f, 0xff} would be converted to the 6-byte
+// string "0a3fff"
+string Hexify(const string& data) {
+  string ret;
+  // Each nibble gets a char, so 2 output bytes per input raw byte.
+  ret.resize(data.size() * 2);
+  int out_idx = 0;
+  for (char c : data) {
+    ret[out_idx++] = HexifyNibble((c >> 4) & 0xf);
+    ret[out_idx++] = HexifyNibble(c & 0xf);
+  }
+  DCHECK_EQ(static_cast<size_t>(out_idx), ret.size());
+  return ret;
+}
+
+// Return true if it looks like contents is already gzipped, false otherwise.
+bool IsGZipped(const string& contents) {
+  const char* ignored;
+  net::GZipHeader header;
+  return header.ReadMore(contents.data(), contents.size(), &ignored) ==
+         net::GZipHeader::COMPLETE_HEADER;
+}
+
+class PpdCacheImpl : public PpdCache {
+ public:
+  explicit PpdCacheImpl(const FilePath& cache_base_dir,
+                        const PpdCache::Options& options)
+      : cache_base_dir_(cache_base_dir) {}
+  ~PpdCacheImpl() override {}
+
+  // Public API functions.
+  Optional<FilePath> Find(
+      const Printer::PpdReference& reference) const override {
+    Optional<FilePath> ret;
+
+    // We can't know here if we have a gzipped or un-gzipped version, so just
+    // look for both.
+    FilePath contents_path_base = GetCachePathBase(reference);
+    for (const string& extension : {".ppd", ".ppd.gz"}) {
+      FilePath contents_path = contents_path_base.AddExtension(extension);
+      if (base::PathExists(contents_path)) {
+        ret = contents_path;
+        break;
+      }
+    }
+    return ret;
+  }
+
+  Optional<FilePath> Store(const Printer::PpdReference& reference,
+                           const string& ppd_contents) override {
+    Optional<FilePath> ret;
+    FilePath contents_path;
+    contents_path = GetCachePathBase(reference).AddExtension(".ppd");
+    if (IsGZipped(ppd_contents)) {
+      contents_path = contents_path.AddExtension(".gz");
+    }
+    if (base::WriteFile(contents_path, ppd_contents.data(),
+                        ppd_contents.size()) ==
+        static_cast<int>(ppd_contents.size())) {
+      ret = contents_path;
+    } else {
+      LOG(ERROR) << "Failed to write " << contents_path.LossyDisplayName();
+      //  Try to clean up the file, as it may have partial contents.  Note
+      // that DeleteFile(nonexistant file) should return true, so failure here
+      // means something is exceptionally hosed.
+      if (!base::DeleteFile(contents_path, false)) {
+        LOG(ERROR) << "Failed to cleanup partially-written file "
+                   << contents_path.LossyDisplayName();
+        return ret;
+      }
+    }
+    return ret;
+  }
+
+ private:
+  // Get the file path at which we expect to find a PPD if it's cached.
+  //
+  // This is, ultimately, just a hash function.  It's extremely infrequently
+  // used (called once when trying to look up information on a printer or store
+  // a PPD), and should be stable, as changing the function will make previously
+  // cached entries unfindable, causing resolve logic to be reinvoked
+  // unnecessarily.
+  //
+  // There's also a faint possibility that a bad actor might try to do something
+  // nefarious by intentionally causing a cache collision that makes the wrong
+  // PPD be used for a printer.  There's no obvious attack vector, but
+  // there's also no real cost to being paranoid here, so we use SHA-256 as the
+  // underlying hash function, and inject fixed field prefixes to prevent
+  // field-substitution spoofing.  This also buys us hash function stability at
+  // the same time.
+  //
+  // Also, care should be taken to preserve the existing hash values if new
+  // fields are added to PpdReference -- that is, if a new field F is added
+  // to PpdReference,  a PpdReference with a default F value should hash to
+  // the same thing as a PpdReference that predates the addition of F to the
+  // structure.
+  //
+  // Note this function expects that the caller will append ".ppd", or ".ppd.gz"
+  // to the output as needed.
+  FilePath GetCachePathBase(const Printer::PpdReference& ref) const {

[skau, 2016/10/14 23:51:49]

Is it desirable that { "user_supplied_ppd_url": "foo" } does not hash to the
same thing as { "user_supplied_url": "foo", "effective_manufacturer": "eee",
"effective_model": "m" }?

[Carlson, 2016/10/17 16:43:20]

I *think* so, but could be convinced otherwise.  My basic thought here is that
*any* change in what we know about a printer should cause us to re-run the
resolution logic, as we might get a different answer this time.  In the case of
the user_supplied field not changing but the other fields changing, we should
just re-resolve to the same thing again, I think. 

The other alternative that I considered as doing a cascading hash, in which we
just hash the highest priority filled in fields and then stop.  It would have
slightly nicer behaviour in the case you cited, but seems like it might be a
little more error-prone in the long term; I like the simplicity of the idea that
changing a field means you miss in the cache as an underlying principle.

If you want me to change it that way, will do.

[skau, 2016/10/17 17:16:26]

Please change it.  While we will need to make sure that fields are grouped
properly for a cascading hash, it doesn't make sense to update a file that we
already have.  We should be explicit about fetching updates since we'd want to
avoid breaking a working configuration.

[Carlson, 2016/10/18 19:05:00]

Done.

+    vector<string> pieces;
+    if (!ref.user_supplied_ppd_url.empty()) {
+      pieces.push_back("user_supplied_ppd_url:");
+      pieces.push_back(ref.user_supplied_ppd_url);
+    }
+    if (!ref.effective_manufacturer.empty()) {
+      pieces.push_back("manufacturer:");
+      pieces.push_back(ref.effective_manufacturer);
+    }
+    if (!ref.effective_model.empty()) {
+      pieces.push_back("model:");
+      pieces.push_back(ref.effective_model);
+    }
+    // The separator here is not needed, but makes debug output more readable.
+    string full_key = base::JoinString(pieces, "|");
+    string ascii_hash = Hexify(crypto::SHA256HashString(full_key));
+    VLOG(3) << "PPD Cache key is " << full_key << " which hashes to "
+            << ascii_hash;
+
+    return cache_base_dir_.Append(ascii_hash);
+  }
+
+  const FilePath cache_base_dir_;
+};

[Lei Zhang, 2016/10/17 17:54:06]

Add DISALLOW_COPY_AND_ASSIGN(...);

[Carlson, 2016/10/18 19:05:01]

Done.

+
+}  // namespace
+
+// static
+unique_ptr<PpdCache> PpdCache::Create(const FilePath& cache_base_dir,
+                                      const PpdCache::Options& options) {
+  return ::base::MakeUnique<PpdCacheImpl>(cache_base_dir, options);
+}
+
+}  // namespace printing
+}  // namespace chromeos