Skip to contentSkip to conversation

Fix coverity issues for AST2700

PATCH Superseded Jamin Lin via 4 participants 11 messages Jun 19, 2024 – Jun 25, 2024 Arm/Aspeed machine support

v1 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

cover Fix coverity issues for AST2700

change from v1:
aspeed/soc: coverity defect: DIVIDE_BY_ZERO aspeed/sdmc: coverity defect: Control flow issues (DEADCODE)

Jamin Lin (2):
aspeed/soc: fix coverity issue
aspeed/sdmc: fix coverity issue

 hw/arm/aspeed_ast27x0.c | 6 ++++++
 hw/misc/aspeed_sdmc.c   | 1 -
 2 files changed, 6 insertions(+), 1 deletion(-)
Signature · 1 line
2.34.1
#1 lore raw

PATCH 1/2 aspeed/soc: fix coverity issue

Fix coverity defect: DIVIDE_BY_ZERO.

---
 hw/arm/aspeed_ast27x0.c | 6 ++++++
 1 file changed, 6 insertions(+)
hw/arm/aspeed_ast27x0.c +6 -0
index b6876b4862..d14a46df6f 100644
@@ -211,6 +211,12 @@ static void aspeed_ram_capacity_write(void *opaque, hwaddr addr, uint64_t data,
     ram_size = object_property_get_uint(OBJECT(&s->sdmc), "ram-size",
                                         &error_abort);
 
+    if (!ram_size) {
+        qemu_log_mask(LOG_GUEST_ERROR,
+                      "%s: ram_size is zero",  __func__);
+        return;
+    }
+
     /*
      * Emulate ddr capacity hardware behavior.
      * If writes the data to the address which is beyond the ram size,
Signature · 1 line
2.34.1
Reviewed-by on v1 1/2
#2 lore raw

Re: [PATCH v1 1/2] aspeed/soc: fix coverity issue

Quoted 3 lines · Jamin Lin

Fix coverity defect: DIVIDE_BY_ZERO.

I have rewritten the commit log :

aspeed/soc: Fix possible divide by zero

Coverity reports a possible DIVIDE_BY_ZERO issue regarding the "ram_size" object property. This can not happen because RAM has predefined valid sizes per SoC. Nevertheless, add a test to close the issue.

Fixes Coverity CID 1547113

with that,

Thanks,

C.

Quoted 21 lines
---
  hw/arm/aspeed_ast27x0.c | 6 ++++++
  1 file changed, 6 insertions(+)
hw/arm/aspeed_ast27x0.c +6 -0
index b6876b4862..d14a46df6f 100644
@@ -211,6 +211,12 @@ static void aspeed_ram_capacity_write(void *opaque, hwaddr addr, uint64_t data,
      ram_size = object_property_get_uint(OBJECT(&s->sdmc), "ram-size",
                                          &error_abort);
 
+    if (!ram_size) {
+        qemu_log_mask(LOG_GUEST_ERROR,
+                      "%s: ram_size is zero",  __func__);
+        return;
+    }
+
      /*
       * Emulate ddr capacity hardware behavior.
       * If writes the data to the address which is beyond the ram size,
#3 lore raw

Re: [PATCH v1 1/2] aspeed/soc: fix coverity issue

In reply to Jamin Lin via
Quoted 13 lines · Jamin Lin

Fix coverity defect: DIVIDE_BY_ZERO.

---
 hw/arm/aspeed_ast27x0.c | 6 ++++++
 1 file changed, 6 insertions(+)
hw/arm/aspeed_ast27x0.c +0 -0
index b6876b4862..d14a46df6f 100644

hw/arm/aspeed_ast27x0.c static void aspeed_ram_capacity_write(void *opaque, hwaddr addr, uint64_t data, · line 219

Quoted 9 lines
hw/arm/aspeed_ast27x0.c +6 -0
@@ -211,6 +211,12 @@ static void aspeed_ram_capacity_write(void *opaque, hwaddr addr, uint64_t data,
     ram_size = object_property_get_uint(OBJECT(&s->sdmc), "ram-size",
                                         &error_abort);
 
+    if (!ram_size) {
+        qemu_log_mask(LOG_GUEST_ERROR,
+                      "%s: ram_size is zero",  __func__);
+        return;
+    }
+

Isn't this a QEMU bug rather than a guest error? The RAM size presumably should never be zero unless the board set the ram-size property on the SDMC incorrectly. So the SDMC device should check (and return an error from its realize method) that the ram-size property is valid, and then here we can just assert(ram_size != 0).

thanks
-- PMM

#4 lore raw

Re: [PATCH v1 1/2] aspeed/soc: fix coverity issue

Quoted 17 lines · Peter Maydell
Quoted 16 lines · Jamin Lin

Fix coverity defect: DIVIDE_BY_ZERO.

---
  hw/arm/aspeed_ast27x0.c | 6 ++++++
  1 file changed, 6 insertions(+)
hw/arm/aspeed_ast27x0.c +0 -0
index b6876b4862..d14a46df6f 100644
@@ -211,6 +211,12 @@ static void aspeed_ram_capacity_write(void *opaque, hwaddr addr, uint64_t data,
      ram_size = object_property_get_uint(OBJECT(&s->sdmc), "ram-size",
                                          &error_abort);
 
Quoted 12 lines

hw/arm/aspeed_ast27x0.c static void aspeed_ram_capacity_write(void *opaque, hwaddr addr, uint64_t data, · line 219

Quoted 6 lines
hw/arm/aspeed_ast27x0.c +6 -0
+    if (!ram_size) {
+        qemu_log_mask(LOG_GUEST_ERROR,
+                      "%s: ram_size is zero",  __func__);
+        return;
+    }
+

Isn't this a QEMU bug rather than a guest error? The RAM size presumably should never be zero unless the board set the ram-size property on the SDMC incorrectly. So the SDMC device should check (and return an error from its realize method) that the ram-size property is valid,

That's the case in aspeed_sdmc_set_ram_size() which is called from the aspeed machine init routine when the ram size is set.

Setting the machine ram size to zero on the command line doesn't report an error though and the size is the default.

Quoted 1 line

and then here we can just assert(ram_size != 0).

Yes.

Jamin, could you please send a v2 with the commit logs update I proposed ? See the patches on my aspeed-9.1 branch.

Thanks,

C.

#5 lore raw

Re: [PATCH v1 1/2] aspeed/soc: fix coverity issue

Quoted 32 lines · Cédric Le Goater
Quoted 17 lines · Peter Maydell
Quoted 16 lines · Jamin Lin

Fix coverity defect: DIVIDE_BY_ZERO.

---
  hw/arm/aspeed_ast27x0.c | 6 ++++++
  1 file changed, 6 insertions(+)
hw/arm/aspeed_ast27x0.c +0 -0
index b6876b4862..d14a46df6f 100644
@@ -211,6 +211,12 @@ static void aspeed_ram_capacity_write(void *opaque, hwaddr addr, uint64_t data,
      ram_size = object_property_get_uint(OBJECT(&s->sdmc), "ram-size",
                                          &error_abort);
 
Quoted 12 lines

hw/arm/aspeed_ast27x0.c static void aspeed_ram_capacity_write(void *opaque, hwaddr addr, uint64_t data, · line 219

Quoted 6 lines
hw/arm/aspeed_ast27x0.c +6 -0
+    if (!ram_size) {
+        qemu_log_mask(LOG_GUEST_ERROR,
+                      "%s: ram_size is zero",  __func__);
+        return;
+    }
+

Isn't this a QEMU bug rather than a guest error? The RAM size presumably should never be zero unless the board set the ram-size property on the SDMC incorrectly. So the SDMC device should check (and return an error from its realize method) that the ram-size property is valid,

Quoted 2 lines

That's the case in aspeed_sdmc_set_ram_size() which is called from the aspeed machine init routine when the ram size is set.

True, but if the property is never set at all then the struct field will be left at whatever value it had, which is 0, I think. So if that's not valid then it either needs to be a different default or else the realize method should complain that the property was never set.

thanks
-- PMM

#6 lore raw

Re: [PATCH v1 1/2] aspeed/soc: fix coverity issue

Quoted 33 lines · Peter Maydell
Quoted 32 lines · Cédric Le Goater
Quoted 29 lines · Peter Maydell
Quoted 13 lines · Jamin Lin

Fix coverity defect: DIVIDE_BY_ZERO.

---
   hw/arm/aspeed_ast27x0.c | 6 ++++++
   1 file changed, 6 insertions(+)
hw/arm/aspeed_ast27x0.c +0 -0
index b6876b4862..d14a46df6f 100644

hw/arm/aspeed_ast27x0.c static void aspeed_ram_capacity_write(void *opaque, hwaddr addr, uint64_t data, · line 219

Quoted 9 lines
hw/arm/aspeed_ast27x0.c +6 -0
@@ -211,6 +211,12 @@ static void aspeed_ram_capacity_write(void *opaque, hwaddr addr, uint64_t data,
       ram_size = object_property_get_uint(OBJECT(&s->sdmc), "ram-size",
                                           &error_abort);
 
+    if (!ram_size) {
+        qemu_log_mask(LOG_GUEST_ERROR,
+                      "%s: ram_size is zero",  __func__);
+        return;
+    }
+

Isn't this a QEMU bug rather than a guest error? The RAM size presumably should never be zero unless the board set the ram-size property on the SDMC incorrectly. So the SDMC device should check (and return an error from its realize method) that the ram-size property is valid,

Quoted 8 lines
Quoted 2 lines

That's the case in aspeed_sdmc_set_ram_size() which is called from the aspeed machine init routine when the ram size is set.

True, but if the property is never set at all then the struct field will be left at whatever value it had, which is 0, I think. So if that's not valid then it either needs to be a different default or else the realize method should complain that the property was never set.

Ah, yes, and the issue has been there for while. I will send a separate patch for this.

Thanks,

C.

#7 lore raw

RE: [PATCH v1 1/2] aspeed/soc: fix coverity issue

In reply to Cédric Le Goater

Hi Cedric,

Quoted 52 lines

-----Original Message-----
From: Cédric Le Goater <[email protected]>
Sent: Monday, June 24, 2024 9:58 PM
To: Peter Maydell <[email protected]>; Jamin Lin <[email protected]>
Cc: Steven Lee <[email protected]>; Troy Lee <[email protected]>; Andrew Jeffery <[email protected]>; Joel Stanley <[email protected]>; open list:ASPEED BMCs <[email protected]>; open list:All patches CC here <[email protected]>; Troy Lee <[email protected]>; Yunlin Tang <[email protected]>
Subject: Re: [PATCH v1 1/2] aspeed/soc: fix coverity issue

Quoted 1 line · Peter Maydell

On Wed, 19 Jun 2024 at 10:35, Jamin Lin <[email protected]>

Quoted 8 lines · wrote:
Quoted 8 lines

Fix coverity defect: DIVIDE_BY_ZERO.

---
  hw/arm/aspeed_ast27x0.c | 6 ++++++
  1 file changed, 6 insertions(+)
Quoted 5 lines

hw/arm/aspeed_ast27x0.c static void aspeed_ram_capacity_write(void

Quoted 5 lines
hw/arm/aspeed_ast27x0.c index renamed from hw/arm/aspeed_ast27x0.c +0 -0

b6876b4862..d14a46df6f 100644

hw/arm/aspeed_ast27x0.c +0 -0
@@ -211,6 +211,12 @@ static void aspeed_ram_capacity_write(void

*opaque, hwaddr addr, uint64_t data,

Quoted 1 line

hw/arm/aspeed_ast27x0.c static void aspeed_ram_capacity_write(void · line 211

Quoted 1 line
hw/arm/aspeed_ast27x0.c +0 -0
      ram_size = object_property_get_uint(OBJECT(&s->sdmc),

"ram-size",

Quoted 9 lines

hw/arm/aspeed_ast27x0.c static void aspeed_ram_capacity_write(void *opaque, hwaddr addr, uint64_t data, · line 219

Quoted 8 lines
hw/arm/aspeed_ast27x0.c +6 -0
                                          &error_abort);
 
+    if (!ram_size) {
+        qemu_log_mask(LOG_GUEST_ERROR,
+                      "%s: ram_size is zero",  __func__);
+        return;
+    }
+
Quoted 5 lines

Isn't this a QEMU bug rather than a guest error? The RAM size presumably should never be zero unless the board set the ram-size property on the SDMC incorrectly. So the SDMC device should check (and return an error from its realize
method) that the ram-size property is valid,

That's the case in aspeed_sdmc_set_ram_size() which is called from the aspeed machine init routine when the ram size is set.

Setting the machine ram size to zero on the command line doesn't report an error though and the size is the default.

Quoted 6 lines
Quoted 1 line

and then here we can just assert(ram_size != 0).

Yes.

Jamin, could you please send a v2 with the commit logs update I proposed ? See the patches on my aspeed-9.1 branch.

I resend v2 patch with your commit log, https://www.mail-archive.com/[email protected]/msg1050302.html
Do we need to drop this patch, https://www.mail-archive.com/[email protected]/msg1050301.html?

Thanks-Jamin

Quoted 4 lines

Thanks,

C.

#8 lore raw

Re: [PATCH v1 1/2] aspeed/soc: fix coverity issue

Hello

Quoted 53 lines · Jamin Lin

Hi Cedric,

Quoted 52 lines

-----Original Message-----
From: Cédric Le Goater <[email protected]>
Sent: Monday, June 24, 2024 9:58 PM
To: Peter Maydell <[email protected]>; Jamin Lin <[email protected]>
Cc: Steven Lee <[email protected]>; Troy Lee <[email protected]>; Andrew Jeffery <[email protected]>; Joel Stanley <[email protected]>; open list:ASPEED BMCs <[email protected]>; open list:All patches CC here <[email protected]>; Troy Lee <[email protected]>; Yunlin Tang <[email protected]>
Subject: Re: [PATCH v1 1/2] aspeed/soc: fix coverity issue

Quoted 1 line · Peter Maydell

On Wed, 19 Jun 2024 at 10:35, Jamin Lin <[email protected]>

Quoted 8 lines · wrote:
Quoted 8 lines

Fix coverity defect: DIVIDE_BY_ZERO.

---
   hw/arm/aspeed_ast27x0.c | 6 ++++++
   1 file changed, 6 insertions(+)
Quoted 5 lines

hw/arm/aspeed_ast27x0.c static void aspeed_ram_capacity_write(void

Quoted 5 lines
hw/arm/aspeed_ast27x0.c index renamed from hw/arm/aspeed_ast27x0.c +0 -0

b6876b4862..d14a46df6f 100644

hw/arm/aspeed_ast27x0.c +0 -0
@@ -211,6 +211,12 @@ static void aspeed_ram_capacity_write(void

*opaque, hwaddr addr, uint64_t data,

Quoted 1 line

hw/arm/aspeed_ast27x0.c static void aspeed_ram_capacity_write(void · line 211

Quoted 1 line
hw/arm/aspeed_ast27x0.c +0 -0
       ram_size = object_property_get_uint(OBJECT(&s->sdmc),

"ram-size",

Quoted 9 lines

hw/arm/aspeed_ast27x0.c static void aspeed_ram_capacity_write(void *opaque, hwaddr addr, uint64_t data, · line 219

Quoted 8 lines
hw/arm/aspeed_ast27x0.c +6 -0
                                           &error_abort);
 
+    if (!ram_size) {
+        qemu_log_mask(LOG_GUEST_ERROR,
+                      "%s: ram_size is zero",  __func__);
+        return;
+    }
+
Quoted 5 lines

Isn't this a QEMU bug rather than a guest error? The RAM size presumably should never be zero unless the board set the ram-size property on the SDMC incorrectly. So the SDMC device should check (and return an error from its realize
method) that the ram-size property is valid,

That's the case in aspeed_sdmc_set_ram_size() which is called from the aspeed machine init routine when the ram size is set.

Setting the machine ram size to zero on the command line doesn't report an error though and the size is the default.

Quoted 7 lines
Quoted 6 lines
Quoted 1 line

and then here we can just assert(ram_size != 0).

Yes.

Jamin, could you please send a v2 with the commit logs update I proposed ? See the patches on my aspeed-9.1 branch.

I resend v2 patch with your commit log, https://www.mail-archive.com/[email protected]/msg1050302.html

We'll need a v3 with the assert. No hurries. We still have some time before the soft freeze.

Thanks,

C.

Quoted 7 lines

Do we need to drop this patch, https://www.mail-archive.com/[email protected]/msg1050301.html?

Thanks-Jamin

Quoted 4 lines

Thanks,

C.

#9 lore raw

PATCH 2/2 aspeed/sdmc: fix coverity issue

In reply to Jamin Lin via

Fix coverity defect: Control flow issues (DEADCODE)

---
 hw/misc/aspeed_sdmc.c | 1 -
 1 file changed, 1 deletion(-)
hw/misc/aspeed_sdmc.c +0 -1
index 93e2e29ead..94eed9264d 100644
@@ -589,7 +589,6 @@ static void aspeed_2700_sdmc_write(AspeedSDMCState *s, uint32_t reg,
     case R_INT_STATUS:
     case R_INT_CLEAR:
     case R_INT_MASK:
-    case R_MAIN_STATUS:
     case R_ERR_STATUS:
     case R_ECC_FAIL_STATUS:
     case R_ECC_FAIL_ADDR:
Signature · 1 line
2.34.1
Reviewed-by on v1 2/2
#10 lore raw

Re: [PATCH v1 2/2] aspeed/sdmc: fix coverity issue

Quoted 1 line · Jamin Lin

Fix coverity defect: Control flow issues (DEADCODE)

aspeed/sdmc: Remove extra R_MAIN_STATUS case

Coverity reports that the newly added 'case R_MAIN_STATUS' is DEADCODE because it can not be reached. This is because R_MAIN_STATUS is handled before in the "Unprotected registers" switch statement. Remove it.

Fixes Coverity CID 1547112

with that,

Thanks,

C.

Quoted 18 lines
---
  hw/misc/aspeed_sdmc.c | 1 -
  1 file changed, 1 deletion(-)
hw/misc/aspeed_sdmc.c +0 -1
index 93e2e29ead..94eed9264d 100644
@@ -589,7 +589,6 @@ static void aspeed_2700_sdmc_write(AspeedSDMCState *s, uint32_t reg,
      case R_INT_STATUS:
      case R_INT_CLEAR:
      case R_INT_MASK:
-    case R_MAIN_STATUS:
      case R_ERR_STATUS:
      case R_ECC_FAIL_STATUS:
      case R_ECC_FAIL_ADDR: