REGR: interpolate raising if fill_value is given (#54927)

* REGR: interpolate raising if fill_value is given

* Update test and message

* Update pandas/core/generic.py

---------

Co-authored-by: Matthew Roeschke <[email protected]>

Author: phofl

doc/source/whatsnew/v2.1.1.rst

@@ -21,6 +21,7 @@ Fixed regressions
 - Fixed regression in :meth:`DataFrame.to_sql` not roundtripping datetime columns correctly for sqlite (:issue:`54877`)
 - Fixed regression in :meth:`MultiIndex.append` raising when appending overlapping :class:`IntervalIndex` levels (:issue:`54934`)
 - Fixed regression in :meth:`Series.drop_duplicates` for PyArrow strings (:issue:`54904`)
+- Fixed regression in :meth:`Series.interpolate` raising when ``fill_value`` was given (:issue:`54920`)
 - Fixed regression in :meth:`Series.value_counts` raising for numeric data if ``bins`` was specified (:issue:`54857`)
 - Fixed regression when comparing a :class:`Series` with ``datetime64`` dtype with ``None`` (:issue:`54870`)
 

pandas/core/generic.py

@@ -8225,10 +8225,11 @@ def interpolate(
                         stacklevel=find_stack_level(),
                     )
 
-        if "fill_value" in kwargs:
+        if method in fillna_methods and "fill_value" in kwargs:
             raise ValueError(
                 "'fill_value' is not a valid keyword for "
-                f"{type(self).__name__}.interpolate"
+                f"{type(self).__name__}.interpolate with method from "
+                f"{fillna_methods}"
             )
 
         if isinstance(obj.index, MultiIndex) and method != "linear":

pandas/tests/series/methods/test_interpolate.py

@@ -858,3 +858,11 @@ def test_interpolate_asfreq_raises(self):
         with pytest.raises(ValueError, match=msg):
             with tm.assert_produces_warning(FutureWarning, match=msg2):
                 ser.interpolate(method="asfreq")
+
+    def test_interpolate_fill_value(self):
+        # GH#54920
+        pytest.importorskip("scipy")
+        ser = Series([np.nan, 0, 1, np.nan, 3, np.nan])
+        result = ser.interpolate(method="nearest", fill_value=0)
+        expected = Series([np.nan, 0, 1, 1, 3, 0])
+        tm.assert_series_equal(result, expected)