Skip to contentSkip to conversation

habanalabs: Elide a warning on 32-bit targets

PATCH In review Palmer Dabbelt 3 participants 7 messages Apr 1, 2022 – Apr 5, 2022

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 habanalabs: Elide a warning on 32-bit targets

From: Palmer Dabbelt <[email protected]>

This double-cast pattern looks a bit awkward, but it already exists elsewhere in the driver. Without this patch I get

drivers/misc/habanalabs/common/memory.c: In function ‘alloc_device_memory’:
drivers/misc/habanalabs/common/memory.c:153:49: warning: cast from pointer to integer of different size [-Wpointer-to-int-cast]

  153 |                                                 (u64) gen_pool_dma_alloc_align(vm->dram_pg_pool,
      |                                                 ^

which ends up promoted to a build error in my test setup.

---

I don't know anything about this driver, I'm just pattern-matching the warning away.

---
 drivers/misc/habanalabs/common/memory.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)
drivers/misc/habanalabs/common/memory.c +5 -5
index e008d82e4ba3..f1fc79c1fc10 100644
@@ -150,12 +150,12 @@ static int alloc_device_memory(struct hl_ctx *ctx, struct hl_mem_in *args,
 		for (i = 0 ; i < num_pgs ; i++) {
 			if (is_power_of_2(page_size))
 				phys_pg_pack->pages[i] =
-						(u64) gen_pool_dma_alloc_align(vm->dram_pg_pool,
-										page_size, NULL,
-										page_size);
+						(u64) (uintptr_t) gen_pool_dma_alloc_align(vm->dram_pg_pool,
+											   page_size, NULL,
+											   page_size);
 			else
-				phys_pg_pack->pages[i] = (u64) gen_pool_alloc(vm->dram_pg_pool,
-										page_size);
+				phys_pg_pack->pages[i] = (u64) (uintptr_t) gen_pool_alloc(vm->dram_pg_pool,
+											  page_size);
 			if (!phys_pg_pack->pages[i]) {
 				dev_err(hdev->dev,
 					"Failed to allocate device memory (out of memory)\n");
Signature · 7 lines
2.34.1


_______________________________________________
linux-riscv mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-riscv
Reviewed-by on v1 1/1
#1 lore raw
Quoted 46 lines · Palmer Dabbelt

From: Palmer Dabbelt <[email protected]>

This double-cast pattern looks a bit awkward, but it already exists elsewhere in the driver. Without this patch I get

drivers/misc/habanalabs/common/memory.c: In function ‘alloc_device_memory’:
drivers/misc/habanalabs/common/memory.c:153:49: warning: cast from pointer to integer of different size [-Wpointer-to-int-cast]

  153 |                                                 (u64) gen_pool_dma_alloc_align(vm->dram_pg_pool,
      |                                                 ^

which ends up promoted to a build error in my test setup.

---

I don't know anything about this driver, I'm just pattern-matching the warning away.

---
 drivers/misc/habanalabs/common/memory.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)
drivers/misc/habanalabs/common/memory.c +5 -6
index e008d82e4ba3..f1fc79c1fc10 100644
@@ -150,12 +150,12 @@ static int alloc_device_memory(struct hl_ctx *ctx, struct hl_mem_in *args,
                for (i = 0 ; i < num_pgs ; i++) {
                        if (is_power_of_2(page_size))
                                phys_pg_pack->pages[i] =
-                                               (u64) gen_pool_dma_alloc_align(vm->dram_pg_pool,
-                                                                               page_size, NULL,
-                                                                               page_size);
+                                               (u64) (uintptr_t) gen_pool_dma_alloc_align(vm->dram_pg_pool,
+                                                                                          page_size, NULL,
+                                                                                          page_size);
                        else
-                               phys_pg_pack->pages[i] = (u64) gen_pool_alloc(vm->dram_pg_pool,
-                                                                               page_size);
+                               phys_pg_pack->pages[i] = (u64) (uintptr_t) gen_pool_alloc(vm->dram_pg_pool,
+                                                                                         page_size);
                        if (!phys_pg_pack->pages[i]) {
                                dev_err(hdev->dev,
                                        "Failed to allocate device memory (out of memory)\n");
--
Quoted 2 lines

2.34.1

This patch is:

Greg,
Could you please apply this directly to your misc tree and send it to Linus at your next pull request ?
I don't have any other fixes pending for 5.18.

For 5.19 we will do a more elegant solution that Arnd has recommended.

Thanks,
Oded

#2 lore raw
Quoted 51 lines · On Fri, 01 Apr 2022 11:13:48 PDT (-0700), [email protected] wrote:
Quoted 46 lines · Palmer Dabbelt

From: Palmer Dabbelt <[email protected]>

This double-cast pattern looks a bit awkward, but it already exists elsewhere in the driver. Without this patch I get

drivers/misc/habanalabs/common/memory.c: In function ‘alloc_device_memory’:
drivers/misc/habanalabs/common/memory.c:153:49: warning: cast from pointer to integer of different size [-Wpointer-to-int-cast]

  153 |                                                 (u64) gen_pool_dma_alloc_align(vm->dram_pg_pool,
      |                                                 ^

which ends up promoted to a build error in my test setup.

---

I don't know anything about this driver, I'm just pattern-matching the warning away.

---
 drivers/misc/habanalabs/common/memory.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)
drivers/misc/habanalabs/common/memory.c +5 -6
index e008d82e4ba3..f1fc79c1fc10 100644
@@ -150,12 +150,12 @@ static int alloc_device_memory(struct hl_ctx *ctx, struct hl_mem_in *args,
                for (i = 0 ; i < num_pgs ; i++) {
                        if (is_power_of_2(page_size))
                                phys_pg_pack->pages[i] =
-                                               (u64) gen_pool_dma_alloc_align(vm->dram_pg_pool,
-                                                                               page_size, NULL,
-                                                                               page_size);
+                                               (u64) (uintptr_t) gen_pool_dma_alloc_align(vm->dram_pg_pool,
+                                                                                          page_size, NULL,
+                                                                                          page_size);
                        else
-                               phys_pg_pack->pages[i] = (u64) gen_pool_alloc(vm->dram_pg_pool,
-                                                                               page_size);
+                               phys_pg_pack->pages[i] = (u64) (uintptr_t) gen_pool_alloc(vm->dram_pg_pool,
+                                                                                         page_size);
                        if (!phys_pg_pack->pages[i]) {
                                dev_err(hdev->dev,
                                        "Failed to allocate device memory (out of memory)\n");
--
Quoted 2 lines

2.34.1

This patch is:

Quoted 8 lines

Greg,
Could you please apply this directly to your misc tree and send it to Linus at your next pull request ?
I don't have any other fixes pending for 5.18.

For 5.19 we will do a more elegant solution that Arnd has recommended.

Thanks.

Assuming this is too late for rc1, would it be possibe to have it in something I can take into my fixes/for-next without too much diff? I put this on top of the offending commit with a

Fixes e8458e20e0a3 ("habanalabs: make sure device mem alloc is page aligned")

at kernel.org/palmer/habana , if that helps any. No big deal if it goes in another way, it's just nice to keep allyesconfig building on my branches directly.

maintainer
#3 lore raw
Quoted 67 lines · Palmer Dabbelt
Quoted 51 lines · On Fri, 01 Apr 2022 11:13:48 PDT (-0700), [email protected] wrote:
Quoted 46 lines · Palmer Dabbelt

From: Palmer Dabbelt <[email protected]>

This double-cast pattern looks a bit awkward, but it already exists elsewhere in the driver. Without this patch I get

drivers/misc/habanalabs/common/memory.c: In function ‘alloc_device_memory’:
drivers/misc/habanalabs/common/memory.c:153:49: warning: cast from pointer to integer of different size [-Wpointer-to-int-cast]

  153 |                                                 (u64) gen_pool_dma_alloc_align(vm->dram_pg_pool,
      |                                                 ^

which ends up promoted to a build error in my test setup.

---

I don't know anything about this driver, I'm just pattern-matching the warning away.

---
 drivers/misc/habanalabs/common/memory.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)
drivers/misc/habanalabs/common/memory.c +5 -6
index e008d82e4ba3..f1fc79c1fc10 100644
@@ -150,12 +150,12 @@ static int alloc_device_memory(struct hl_ctx *ctx, struct hl_mem_in *args,
                for (i = 0 ; i < num_pgs ; i++) {
                        if (is_power_of_2(page_size))
                                phys_pg_pack->pages[i] =
-                                               (u64) gen_pool_dma_alloc_align(vm->dram_pg_pool,
-                                                                               page_size, NULL,
-                                                                               page_size);
+                                               (u64) (uintptr_t) gen_pool_dma_alloc_align(vm->dram_pg_pool,
+                                                                                          page_size, NULL,
+                                                                                          page_size);
                        else
-                               phys_pg_pack->pages[i] = (u64) gen_pool_alloc(vm->dram_pg_pool,
-                                                                               page_size);
+                               phys_pg_pack->pages[i] = (u64) (uintptr_t) gen_pool_alloc(vm->dram_pg_pool,
+                                                                                         page_size);
                        if (!phys_pg_pack->pages[i]) {
                                dev_err(hdev->dev,
                                        "Failed to allocate device memory (out of memory)\n");
--
Quoted 2 lines

2.34.1

This patch is:

Quoted 8 lines

Greg,
Could you please apply this directly to your misc tree and send it to Linus at your next pull request ?
I don't have any other fixes pending for 5.18.

For 5.19 we will do a more elegant solution that Arnd has recommended.

Thanks.

Assuming this is too late for rc1, would it be possibe to have it in something I can take into my fixes/for-next without too much diff? I put this on top of the offending commit with a

Quoted 5 lines
Fixes e8458e20e0a3 ("habanalabs: make sure device mem alloc is page aligned")

at kernel.org/palmer/habana , if that helps any. No big deal if it goes in another way, it's just nice to keep allyesconfig building on my branches directly.

Looks like Guenter sent in a "more complete" version of this. I'll queue it up to my tree and get it to Linus for -rc2.

thanks,
greg k-h

#4 lore raw
Quoted 68 lines · Greg KH
Quoted 67 lines · Palmer Dabbelt
Quoted 51 lines · On Fri, 01 Apr 2022 11:13:48 PDT (-0700), [email protected] wrote:
Quoted 46 lines · Palmer Dabbelt

From: Palmer Dabbelt <[email protected]>

This double-cast pattern looks a bit awkward, but it already exists elsewhere in the driver. Without this patch I get

drivers/misc/habanalabs/common/memory.c: In function ‘alloc_device_memory’:
drivers/misc/habanalabs/common/memory.c:153:49: warning: cast from pointer to integer of different size [-Wpointer-to-int-cast]

  153 |                                                 (u64) gen_pool_dma_alloc_align(vm->dram_pg_pool,
      |                                                 ^

which ends up promoted to a build error in my test setup.

---

I don't know anything about this driver, I'm just pattern-matching the warning away.

---
 drivers/misc/habanalabs/common/memory.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)
drivers/misc/habanalabs/common/memory.c +5 -6
index e008d82e4ba3..f1fc79c1fc10 100644
@@ -150,12 +150,12 @@ static int alloc_device_memory(struct hl_ctx *ctx, struct hl_mem_in *args,
                for (i = 0 ; i < num_pgs ; i++) {
                        if (is_power_of_2(page_size))
                                phys_pg_pack->pages[i] =
-                                               (u64) gen_pool_dma_alloc_align(vm->dram_pg_pool,
-                                                                               page_size, NULL,
-                                                                               page_size);
+                                               (u64) (uintptr_t) gen_pool_dma_alloc_align(vm->dram_pg_pool,
+                                                                                          page_size, NULL,
+                                                                                          page_size);
                        else
-                               phys_pg_pack->pages[i] = (u64) gen_pool_alloc(vm->dram_pg_pool,
-                                                                               page_size);
+                               phys_pg_pack->pages[i] = (u64) (uintptr_t) gen_pool_alloc(vm->dram_pg_pool,
+                                                                                         page_size);
                        if (!phys_pg_pack->pages[i]) {
                                dev_err(hdev->dev,
                                        "Failed to allocate device memory (out of memory)\n");
--
Quoted 2 lines

2.34.1

This patch is:

Quoted 8 lines

Greg,
Could you please apply this directly to your misc tree and send it to Linus at your next pull request ?
I don't have any other fixes pending for 5.18.

For 5.19 we will do a more elegant solution that Arnd has recommended.

Thanks.

Assuming this is too late for rc1, would it be possibe to have it in something I can take into my fixes/for-next without too much diff? I put this on top of the offending commit with a

Quoted 8 lines
Quoted 5 lines
Fixes e8458e20e0a3 ("habanalabs: make sure device mem alloc is page aligned")

at kernel.org/palmer/habana , if that helps any. No big deal if it goes in another way, it's just nice to keep allyesconfig building on my branches directly.

Looks like Guenter sent in a "more complete" version of this. I'll queue it up to my tree and get it to Linus for -rc2.

I'm fine with anything here, as long as it fixes my build errors ;).

Do you mind giving me a pointer? That way I can use exactly what's going upstream, rather that keeping around this now-dead-end diff (I poked around your kernel.org trees and couldn't find anything, but I've not done that before).

_______________________________________________
linux-riscv mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-riscv

maintainer
#5 lore raw
Quoted 78 lines · Palmer Dabbelt
Quoted 68 lines · Greg KH
Quoted 67 lines · Palmer Dabbelt
Quoted 51 lines · On Fri, 01 Apr 2022 11:13:48 PDT (-0700), [email protected] wrote:
Quoted 46 lines · Palmer Dabbelt

From: Palmer Dabbelt <[email protected]>

This double-cast pattern looks a bit awkward, but it already exists elsewhere in the driver. Without this patch I get

drivers/misc/habanalabs/common/memory.c: In function ‘alloc_device_memory’:
drivers/misc/habanalabs/common/memory.c:153:49: warning: cast from pointer to integer of different size [-Wpointer-to-int-cast]

  153 |                                                 (u64) gen_pool_dma_alloc_align(vm->dram_pg_pool,
      |                                                 ^

which ends up promoted to a build error in my test setup.

---

I don't know anything about this driver, I'm just pattern-matching the warning away.

---
 drivers/misc/habanalabs/common/memory.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)
drivers/misc/habanalabs/common/memory.c +5 -6
index e008d82e4ba3..f1fc79c1fc10 100644
@@ -150,12 +150,12 @@ static int alloc_device_memory(struct hl_ctx *ctx, struct hl_mem_in *args,
                for (i = 0 ; i < num_pgs ; i++) {
                        if (is_power_of_2(page_size))
                                phys_pg_pack->pages[i] =
-                                               (u64) gen_pool_dma_alloc_align(vm->dram_pg_pool,
-                                                                               page_size, NULL,
-                                                                               page_size);
+                                               (u64) (uintptr_t) gen_pool_dma_alloc_align(vm->dram_pg_pool,
+                                                                                          page_size, NULL,
+                                                                                          page_size);
                        else
-                               phys_pg_pack->pages[i] = (u64) gen_pool_alloc(vm->dram_pg_pool,
-                                                                               page_size);
+                               phys_pg_pack->pages[i] = (u64) (uintptr_t) gen_pool_alloc(vm->dram_pg_pool,
+                                                                                         page_size);
                        if (!phys_pg_pack->pages[i]) {
                                dev_err(hdev->dev,
                                        "Failed to allocate device memory (out of memory)\n");
--
Quoted 2 lines

2.34.1

This patch is:

Quoted 8 lines

Greg,
Could you please apply this directly to your misc tree and send it to Linus at your next pull request ?
I don't have any other fixes pending for 5.18.

For 5.19 we will do a more elegant solution that Arnd has recommended.

Thanks.

Assuming this is too late for rc1, would it be possibe to have it in something I can take into my fixes/for-next without too much diff? I put this on top of the offending commit with a

Quoted 8 lines
Quoted 5 lines
Fixes e8458e20e0a3 ("habanalabs: make sure device mem alloc is page aligned")

at kernel.org/palmer/habana , if that helps any. No big deal if it goes in another way, it's just nice to keep allyesconfig building on my branches directly.

Looks like Guenter sent in a "more complete" version of this. I'll queue it up to my tree and get it to Linus for -rc2.

Quoted 6 lines

I'm fine with anything here, as long as it fixes my build errors ;).

Do you mind giving me a pointer? That way I can use exactly what's going upstream, rather that keeping around this now-dead-end diff (I poked around your kernel.org trees and couldn't find anything, but I've not done that before).

The commit I took is here:

	https://lore.kernel.org/all/[email protected]/

and you can find it in my char-misc.git tree with commit:
94865e2dcb46 ("habanalabs: Fix test build failures")
and it should show up in the next linux-next tree as well.

thanks,

greg k-h

maintainer
#6 lore raw
In reply to Oded Gabbay
Quoted 51 lines · Oded Gabbay
Quoted 46 lines · Palmer Dabbelt

From: Palmer Dabbelt <[email protected]>

This double-cast pattern looks a bit awkward, but it already exists elsewhere in the driver. Without this patch I get

drivers/misc/habanalabs/common/memory.c: In function ‘alloc_device_memory’:
drivers/misc/habanalabs/common/memory.c:153:49: warning: cast from pointer to integer of different size [-Wpointer-to-int-cast]

  153 |                                                 (u64) gen_pool_dma_alloc_align(vm->dram_pg_pool,
      |                                                 ^

which ends up promoted to a build error in my test setup.

---

I don't know anything about this driver, I'm just pattern-matching the warning away.

---
 drivers/misc/habanalabs/common/memory.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)
drivers/misc/habanalabs/common/memory.c +5 -6
index e008d82e4ba3..f1fc79c1fc10 100644
@@ -150,12 +150,12 @@ static int alloc_device_memory(struct hl_ctx *ctx, struct hl_mem_in *args,
                for (i = 0 ; i < num_pgs ; i++) {
                        if (is_power_of_2(page_size))
                                phys_pg_pack->pages[i] =
-                                               (u64) gen_pool_dma_alloc_align(vm->dram_pg_pool,
-                                                                               page_size, NULL,
-                                                                               page_size);
+                                               (u64) (uintptr_t) gen_pool_dma_alloc_align(vm->dram_pg_pool,
+                                                                                          page_size, NULL,
+                                                                                          page_size);
                        else
-                               phys_pg_pack->pages[i] = (u64) gen_pool_alloc(vm->dram_pg_pool,
-                                                                               page_size);
+                               phys_pg_pack->pages[i] = (u64) (uintptr_t) gen_pool_alloc(vm->dram_pg_pool,
+                                                                                         page_size);
                        if (!phys_pg_pack->pages[i]) {
                                dev_err(hdev->dev,
                                        "Failed to allocate device memory (out of memory)\n");
--
Quoted 2 lines

2.34.1

This patch is:

Quoted 5 lines

Greg,
Could you please apply this directly to your misc tree and send it to Linus at your next pull request ?

So ignore https://lore.kernel.org/r/[email protected] ?

Quoted 1 line

I don't have any other fixes pending for 5.18.

Yes, I will queue this up after -rc1 is out.

thanks,

greg k-h