summary refs log tree commit diff stats
path: root/results/classifier/105/other/714629
diff options
context:
space:
mode:
authorChristian Krinitsin <mail@krinitsin.com>2025-06-03 12:04:13 +0000
committerChristian Krinitsin <mail@krinitsin.com>2025-06-03 12:04:13 +0000
commit256709d2eb3fd80d768a99964be5caa61effa2a0 (patch)
tree05b2352fba70923126836a64b6a0de43902e976a /results/classifier/105/other/714629
parent2ab14fa96a6c5484b5e4ba8337551bb8dcc79cc5 (diff)
downloadqemu-analysis-256709d2eb3fd80d768a99964be5caa61effa2a0.tar.gz
qemu-analysis-256709d2eb3fd80d768a99964be5caa61effa2a0.zip
add new classifier result
Diffstat (limited to 'results/classifier/105/other/714629')
-rw-r--r--results/classifier/105/other/71462923
1 files changed, 23 insertions, 0 deletions
diff --git a/results/classifier/105/other/714629 b/results/classifier/105/other/714629
new file mode 100644
index 000000000..26204b8da
--- /dev/null
+++ b/results/classifier/105/other/714629
@@ -0,0 +1,23 @@
+other: 0.806
+instruction: 0.766
+device: 0.755
+network: 0.747
+vnc: 0.635
+socket: 0.629
+mistranslation: 0.619
+semantic: 0.591
+boot: 0.505
+graphic: 0.379
+assembly: 0.307
+KVM: 0.253
+
+BIOS doesn't load when read() returns less than the full ROM length
+
+When qemu is running over a 9p filesystem (e.g. when running underneath -virtfs of another qemu), and probably some other network filesystems, it fails to read its BIOS image.  This is because it uses a single low-level read() call on the bios.bin, asking for the full file.  However read() may return less than the full length, and it's the caller's responsibility to call it repeatedly if necessary.  When read does come up short, qemu doesn't repeat the call, and reports an error instead.  The attached patch fixes the one problem I saw, but I haven't tried to cover the general case (e.g. extension ROMs).
+
+
+
+This seems to have been fixed by commit 1f40547f5ce0c135faa7d14f066b97002fd8c204.
+
+Fix included in QEMU v4.0.0.
+