Skip to contentSkip to conversation

ide-test: Fix endianness problems

PATCH Applied Kevin Wolf 5 participants 9 messages May 15, 2013 – May 16, 2013

lore
Share & tools
lore

Kevin Wolf

author
#0 lore raw

PATCH 1/1 ide-test: Fix endianness problems

The test case passes on big endian hosts now (tested on ppc64)

---
 tests/ide-test.c | 32 +++++++++++++++++++++++++++-----
 1 file changed, 27 insertions(+), 5 deletions(-)
tests/ide-test.c +27 -5
index bdc1da7..365e995 100644
@@ -252,7 +252,10 @@ static void test_bmdma_simple_rw(void)
     uintptr_t guest_buf = guest_alloc(guest_malloc, len);
 
     PrdtEntry prdt[] = {
-        { .addr = guest_buf, .size = len | PRDT_EOT },
+        {
+            .addr = cpu_to_le32(guest_buf),
+            .size = cpu_to_le32(len | PRDT_EOT),
+        },
     };
 
     buf = g_malloc(len);
@@ -304,7 +307,10 @@ static void test_bmdma_short_prdt(void)
     uint8_t status;
 
     PrdtEntry prdt[] = {
-        { .addr = 0, .size = 0x10 | PRDT_EOT },
+        {
+            .addr = 0,
+            .size = cpu_to_le32(0x10 | PRDT_EOT),
+        },
     };
 
     /* Normal request */
@@ -325,7 +331,10 @@ static void test_bmdma_long_prdt(void)
     uint8_t status;
 
     PrdtEntry prdt[] = {
-        { .addr = 0, .size = 0x1000 | PRDT_EOT },
+        {
+            .addr = 0,
+            .size = cpu_to_le32(0x1000 | PRDT_EOT),
+        },
     };
 
     /* Normal request */
@@ -355,6 +364,17 @@ static void test_bmdma_teardown(void)
     ide_test_quit();
 }
 
+static void string_cpu_to_be16(uint16_t *s, size_t bytes)
+{
+    g_assert((bytes & 1) == 0);
+    bytes /= 2;
+
+    while (bytes--) {
+        *s = cpu_to_be16(*s);
+        s++;
+    }
+}
+
 static void test_identify(void)
 {
     uint8_t data;
@@ -389,10 +409,12 @@ static void test_identify(void)
     assert_bit_clear(data, BSY | DF | ERR | DRQ);
 
     /* Check serial number/version in the buffer */
-    ret = memcmp(&buf[10], "ettsidks            ", 20);
+    string_cpu_to_be16(&buf[10], 20);
+    ret = memcmp(&buf[10], "testdisk            ", 20);
     g_assert(ret == 0);
 
-    ret = memcmp(&buf[23], "evsroi n", 8);
+    string_cpu_to_be16(&buf[23], 8);
+    ret = memcmp(&buf[23], "version ", 8);
     g_assert(ret == 0);
 
     /* Write cache enabled bit */
Signature · 1 line
1.8.1.4
author
#2 lore raw

Am 15.05.2013 um 15:15 hat Paolo Bonzini geschrieben:

Quoted 6 lines

Il 15/05/2013 15:00, Kevin Wolf ha scritto:

Quoted 3 lines

The test case passes on big endian hosts now (tested on ppc64)

For 1.5? Do we need an extra -rc?

An extra -rc for a test case fix on big endian hosts is probably too much... But if it can still be applied for the release, sure, why not.

Kevin

#3 lore raw

Il 15/05/2013 15:24, Kevin Wolf ha scritto:

Quoted 10 lines

Am 15.05.2013 um 15:15 hat Paolo Bonzini geschrieben:

Quoted 6 lines

Il 15/05/2013 15:00, Kevin Wolf ha scritto:

Quoted 3 lines

The test case passes on big endian hosts now (tested on ppc64)

For 1.5? Do we need an extra -rc?

An extra -rc for a test case fix on big endian hosts is probably too much... But if it can still be applied for the release, sure, why not.

Well, we have Mac OS X not building, BSDs not supporting the GTK+ release, and SLIRP broken on Windows. At least the first two have patches on the list, the last is bisected but no patch yet.

Paolo

#4 lore raw
Quoted 5 lines · Paolo Bonzini

Il 15/05/2013 15:24, Kevin Wolf ha scritto:

Quoted 4 lines

Am 15.05.2013 um 15:15 hat Paolo Bonzini geschrieben:

Quoted 3 lines

Il 15/05/2013 15:00, Kevin Wolf ha scritto:

Quoted 2 lines

The test case passes on big endian hosts now (tested on ppc64)

Quoted 8 lines
Quoted 6 lines
Quoted 3 lines
Quoted 1 line

For 1.5? Do we need an extra -rc?

An extra -rc for a test case fix on big endian hosts is probably too much... But if it can still be applied for the release, sure, why not.

Well, we have Mac OS X not building,

Peter's patch fixes it and will be in -rc2

Quoted 1 line

BSDs not supporting the GTK+

We don't have a formal support statement, but I don't think it's controversial to say that BSD hosts are a secondary platform from a host point of view.

A build issue in an optional feature on a secondary platform is not something I'd consider a release blocker.

Quoted 1 line

release, and SLIRP broken on Windows.

This is a tough one since it's the default networking backend. That said, it likely has been broken for a while (months) and no one noticed. That makes me think that fixing in stable (particularly if we scheduled a stable release for two weeks after 1.5.0) is reasonable.

Regards,

Anthony Liguori

Quoted 4 lines

At least the first two have
patches on the list, the last is bisected but no patch yet.

Paolo

#5 lore raw
Quoted 27 lines · Anthony Liguori
Quoted 5 lines · Paolo Bonzini

Il 15/05/2013 15:24, Kevin Wolf ha scritto:

Quoted 4 lines

Am 15.05.2013 um 15:15 hat Paolo Bonzini geschrieben:

Quoted 3 lines

Il 15/05/2013 15:00, Kevin Wolf ha scritto:

Quoted 2 lines

The test case passes on big endian hosts now (tested on ppc64)

Quoted 8 lines
Quoted 6 lines
Quoted 3 lines
Quoted 1 line

For 1.5? Do we need an extra -rc?

An extra -rc for a test case fix on big endian hosts is probably too much... But if it can still be applied for the release, sure, why not.

Well, we have Mac OS X not building,

Peter's patch fixes it and will be in -rc2

Quoted 1 line

BSDs not supporting the GTK+

We don't have a formal support statement, but I don't think it's controversial to say that BSD hosts are a secondary platform from a host point of view.

A build issue in an optional feature on a secondary platform is not something I'd consider a release blocker.

Quoted 6 lines
Quoted 1 line

release, and SLIRP broken on Windows.

This is a tough one since it's the default networking backend. That said, it likely has been broken for a while (months) and no one noticed. That makes me think that fixing in stable (particularly if we scheduled a stable release for two weeks after 1.5.0) is reasonable.

This is a bit tight. Freeze for 1.4.2 is the same day as 1.5.0, so we'd be testing 2 stable releases at the same time.

So if we do this, I think it we should restrict it to fixing the specific issues we're considering as potential 1.5.0 release blockers so we can focus on those in testing rather than opening it up for general fixes.

Quoted 10 lines

Regards,

Anthony Liguori

Quoted 4 lines

At least the first two have
patches on the list, the last is bisected but no patch yet.

Paolo

Reviewed-by on v1 1/1
#6 lore raw
In reply to Kevin Wolf
Quoted 20 lines · Kevin Wolf
tests/ide-test.c +11 -0
@@ -355,6 +364,17 @@ static void test_bmdma_teardown(void)
     ide_test_quit();
 }
 
+static void string_cpu_to_be16(uint16_t *s, size_t bytes)
+{
+    g_assert((bytes & 1) == 0);
+    bytes /= 2;
+
+    while (bytes--) {
+        *s = cpu_to_be16(*s);
+        s++;
+    }
+}
+
 static void test_identify(void)
 {
     uint8_t data;
@@ -389,10 +409,12 @@ static void test_identify(void)
     assert_bit_clear(data, BSY | DF | ERR | DRQ);

tests/ide-test.c static void test_identify(void) · line 417

Quoted 10 lines
tests/ide-test.c +4 -2
 
     /* Check serial number/version in the buffer */
-    ret = memcmp(&buf[10], "ettsidks            ", 20);
+    string_cpu_to_be16(&buf[10], 20);
+    ret = memcmp(&buf[10], "testdisk            ", 20);
     g_assert(ret == 0);
 
-    ret = memcmp(&buf[23], "evsroi n", 8);
+    string_cpu_to_be16(&buf[23], 8);
+    ret = memcmp(&buf[23], "version ", 8);

It would have been simpler to specify string_cpu_to_be16() length in "elements" instead of bytes. Then you can drop the assertion and conversion. Not a problem though.

Anthony: Please take this patch without a pull request. I think me sending pull requests for a single late-rc fix doesn't add value.

author
#7 lore raw

Am 16.05.2013 um 11:04 hat Stefan Hajnoczi geschrieben:

Quoted 4 lines

Anthony: Please take this patch without a pull request. I think me sending pull requests for a single late-rc fix doesn't add value.

It's already merged.

Kevin