Message ID | pull.1649.git.1706897095273.gitgitgadget@gmail.com (mailing list archive) |
---|---|
State | Accepted |
Commit | 156e28b36d424a26b4548de636fa548b14defa71 |
Headers | show |
Series | sparse-index: pass string length to index_file_exists() | expand |
"Jeff Hostetler via GitGitGadget" <gitgitgadget@gmail.com> writes: > From: Jeff Hostetler <jeffhostetler@github.com> > > The call to index_file_exists() in the loop in expand_to_path() passes > the wrong string length. Let's fix that. > > The loop in expand_to_path() searches the name-hash for each > sub-directory prefix in the provided pathname. That is, by searching > for "dir1/" then "dir1/dir2/" then "dir1/dir2/dir3/" and so on until > it finds a cache-entry representing a sparse directory. > > The code creates "strbuf path_mutable" to contain the working pathname > and modifies the buffer in-place by temporarily replacing the character > following each successive "/" with NUL for the duration of the call to > index_file_exists(). > > It does not update the strbuf.len during this substitution. Meaning we memihash() the full pathname munged with '/' -> NUL through to the end of the original, when we should memihash() the truncated leading pathname. This is bad, and the ... > > Pass the patched length of the prefix path instead. ... fix looks quite straight-forward. > Signed-off-by: Jeff Hostetler <jeffhostetler@github.com> > --- The problem description and the fix makes sense, but did you actually see an end-user visible breakage due to this bug? I am wondering how you found it, and if it is reasonable to have a test demonstrate the breakage. > sparse-index.c | 4 ++-- > 1 file changed, 2 insertions(+), 2 deletions(-) > > diff --git a/sparse-index.c b/sparse-index.c > index 1fdb07a9e69..093708f6220 100644 > --- a/sparse-index.c > +++ b/sparse-index.c > @@ -579,8 +579,9 @@ void expand_to_path(struct index_state *istate, > replace++; > temp = *replace; > *replace = '\0'; > + substr_len = replace - path_mutable.buf; > if (index_file_exists(istate, path_mutable.buf, > - path_mutable.len, icase)) { > + substr_len, icase)) { There is a break out of this loop when the condition for this "if" statement holds, but the value of substr_len does not affect what happens after this index_file_exists() call (correctly) computes its result. The fix looks good. Thanks. > /* > * We found a parent directory in the name-hash > * hashtable, because only sparse directory entries > @@ -593,7 +594,6 @@ void expand_to_path(struct index_state *istate, > } > > *replace = temp; > - substr_len = replace - path_mutable.buf; > } > > cleanup: > > base-commit: 43c8a30d150ecede9709c1f2527c8fba92c65f40
On 2/2/24 1:24 PM, Junio C Hamano wrote: > "Jeff Hostetler via GitGitGadget" <gitgitgadget@gmail.com> writes: > >> From: Jeff Hostetler <jeffhostetler@github.com> >> >> The call to index_file_exists() in the loop in expand_to_path() passes >> the wrong string length. Let's fix that. >> >> The loop in expand_to_path() searches the name-hash for each >> sub-directory prefix in the provided pathname. That is, by searching >> for "dir1/" then "dir1/dir2/" then "dir1/dir2/dir3/" and so on until >> it finds a cache-entry representing a sparse directory. >> >> The code creates "strbuf path_mutable" to contain the working pathname >> and modifies the buffer in-place by temporarily replacing the character >> following each successive "/" with NUL for the duration of the call to >> index_file_exists(). >> >> It does not update the strbuf.len during this substitution. > > Meaning we memihash() the full pathname munged with '/' -> NUL through > to the end of the original, when we should memihash() the truncated > leading pathname. This is bad, and the ... > >> >> Pass the patched length of the prefix path instead. > > ... fix looks quite straight-forward. > >> Signed-off-by: Jeff Hostetler <jeffhostetler@github.com> >> --- > > The problem description and the fix makes sense, but did you > actually see an end-user visible breakage due to this bug? I am > wondering how you found it, and if it is reasonable to have a test > demonstrate the breakage. I'm working on a bug where the fsmonitor client doesn't invalidate the CE flags when there is a case discrepancy between the expected and observed case on a case-insensitive file system. (Fix due shortly.) And I was single-stepping in the debugger inside of `index_file_exists()` while tracking that down and noticed that the length argument was bogus. Something like { name="dir1/", len=10 } I don't remember if this bug caused visible problems or not. It felt like it caused a few extra rounds of mutually-recursive calls between `index_file_exists()` and `expand_to_path()`, but I can't say that for certain (and I was focused on a different problem at the time). I do have some test code in `t/helper/lazy-init-name-hash.c` that I suppose we could extend to beat on it, but I'm not sure it is worth it in this case. Jeff > >> sparse-index.c | 4 ++-- >> 1 file changed, 2 insertions(+), 2 deletions(-) >> >> diff --git a/sparse-index.c b/sparse-index.c >> index 1fdb07a9e69..093708f6220 100644 >> --- a/sparse-index.c >> +++ b/sparse-index.c >> @@ -579,8 +579,9 @@ void expand_to_path(struct index_state *istate, >> replace++; >> temp = *replace; >> *replace = '\0'; >> + substr_len = replace - path_mutable.buf; >> if (index_file_exists(istate, path_mutable.buf, >> - path_mutable.len, icase)) { >> + substr_len, icase)) { > > There is a break out of this loop when the condition for this "if" > statement holds, but the value of substr_len does not affect what > happens after this index_file_exists() call (correctly) computes its > result. The fix looks good. > > Thanks. > >> /* >> * We found a parent directory in the name-hash >> * hashtable, because only sparse directory entries >> @@ -593,7 +594,6 @@ void expand_to_path(struct index_state *istate, >> } >> >> *replace = temp; >> - substr_len = replace - path_mutable.buf; >> } >> >> cleanup: >> >> base-commit: 43c8a30d150ecede9709c1f2527c8fba92c65f40 >
Jeff Hostetler <git@jeffhostetler.com> writes: > I don't remember if this bug caused visible problems or not. It felt > like it caused a few extra rounds of mutually-recursive calls between > `index_file_exists()` and `expand_to_path()`, but I can't say that > for certain (and I was focused on a different problem at the time). > > I do have some test code in `t/helper/lazy-init-name-hash.c` that > I suppose we could extend to beat on it, but I'm not sure it is > worth it in this case. Yeah, if we had a reproduction already handy that would have been a different story, but I agree that it is not worth it. Thanks for a fix. Queued.
diff --git a/sparse-index.c b/sparse-index.c index 1fdb07a9e69..093708f6220 100644 --- a/sparse-index.c +++ b/sparse-index.c @@ -579,8 +579,9 @@ void expand_to_path(struct index_state *istate, replace++; temp = *replace; *replace = '\0'; + substr_len = replace - path_mutable.buf; if (index_file_exists(istate, path_mutable.buf, - path_mutable.len, icase)) { + substr_len, icase)) { /* * We found a parent directory in the name-hash * hashtable, because only sparse directory entries @@ -593,7 +594,6 @@ void expand_to_path(struct index_state *istate, } *replace = temp; - substr_len = replace - path_mutable.buf; } cleanup: