All of lore.kernel.org
 help / color / mirror / Atom feed
From: Jonathan Cameron <Jonathan.Cameron@Huawei.com>
To: Thomas Huth <thuth@redhat.com>
Cc: <qemu-devel@nongnu.org>, Fan Ni <fan.ni@samsung.com>,
	"Michael S. Tsirkin" <mst@redhat.com>,
	Marcel Apfelbaum <marcel.apfelbaum@gmail.com>,
	<qemu-trivial@nongnu.org>
Subject: Re: [PATCH 3/3] hw/mem/cxl_type3: Fix problem with g_steal_pointer()
Date: Mon, 4 Mar 2024 15:10:37 +0000	[thread overview]
Message-ID: <20240304151037.00000f6c@Huawei.com> (raw)
In-Reply-To: <20240304104406.59855-4-thuth@redhat.com>

On Mon,  4 Mar 2024 11:44:06 +0100
Thomas Huth <thuth@redhat.com> wrote:

> When setting GLIB_VERSION_MAX_ALLOWED to GLIB_VERSION_2_58 or higher,
> glib adds type safety checks to the g_steal_pointer() macro. This
> triggers errors in the ct3_build_cdat_entries_for_mr() function which
> uses the g_steal_pointer() for type-casting from one pointer type to
> the other (which also looks quite weird since the local pointers have
> all been declared with g_autofree though they are never freed here).
> Fix it by using a proper typecast instead. For making this possible, we
> have to remove the QEMU_PACKED attribute from some structs since GCC
> otherwise complains that the source and destination pointer might
> have different alignment restrictions. Removing the QEMU_PACKED should
> be fine here since the structs are already naturally aligned. Anyway,
> add some QEMU_BUILD_BUG_ON() statements to make sure that we've got
> the right sizes (without padding in the structs).

I missed these as well when getting rid of the false handling
of failure of g_new0 calls.

Another alternative would be to point to the head structures rather
than the containing structure - would avoid need to cast.
That might be neater?  Should I think also remove the alignment
question?


> 
> Signed-off-by: Thomas Huth <thuth@redhat.com>
> ---
>  include/hw/cxl/cxl_cdat.h |  9 ++++++---
>  hw/mem/cxl_type3.c        | 24 ++++++++++++------------
>  2 files changed, 18 insertions(+), 15 deletions(-)
> 
> diff --git a/include/hw/cxl/cxl_cdat.h b/include/hw/cxl/cxl_cdat.h
> index b44cefaad6..17a09066dc 100644
> --- a/include/hw/cxl/cxl_cdat.h
> +++ b/include/hw/cxl/cxl_cdat.h
> @@ -82,7 +82,8 @@ typedef struct CDATDsmas {
>      uint16_t reserved;
>      uint64_t DPA_base;
>      uint64_t DPA_length;
> -} QEMU_PACKED CDATDsmas;
> +} CDATDsmas;
> +QEMU_BUILD_BUG_ON(sizeof(CDATDsmas) != 24);
>  
>  /* Device Scoped Latency and Bandwidth Information Structure - CDAT Table 5 */
>  typedef struct CDATDslbis {
> @@ -95,7 +96,8 @@ typedef struct CDATDslbis {
>      uint64_t entry_base_unit;
>      uint16_t entry[3];
>      uint16_t reserved2;
> -} QEMU_PACKED CDATDslbis;
> +} CDATDslbis;
> +QEMU_BUILD_BUG_ON(sizeof(CDATDslbis) != 24);
>  
>  /* Device Scoped Memory Side Cache Information Structure - CDAT Table 6 */
>  typedef struct CDATDsmscis {
> @@ -122,7 +124,8 @@ typedef struct CDATDsemts {
>      uint16_t reserved;
>      uint64_t DPA_offset;
>      uint64_t DPA_length;
> -} QEMU_PACKED CDATDsemts;
> +} CDATDsemts;
> +QEMU_BUILD_BUG_ON(sizeof(CDATDsemts) != 24);
>  
>  /* Switch Scoped Latency and Bandwidth Information Structure - CDAT Table 9 */
>  typedef struct CDATSslbisHeader {
> diff --git a/hw/mem/cxl_type3.c b/hw/mem/cxl_type3.c
> index e8801805b9..b679dfae1c 100644
> --- a/hw/mem/cxl_type3.c
> +++ b/hw/mem/cxl_type3.c
> @@ -46,12 +46,12 @@ static void ct3_build_cdat_entries_for_mr(CDATSubHeader **cdat_table,
>                                            int dsmad_handle, MemoryRegion *mr,
>                                            bool is_pmem, uint64_t dpa_base)
>  {
> -    g_autofree CDATDsmas *dsmas = NULL;
> -    g_autofree CDATDslbis *dslbis0 = NULL;
> -    g_autofree CDATDslbis *dslbis1 = NULL;
> -    g_autofree CDATDslbis *dslbis2 = NULL;
> -    g_autofree CDATDslbis *dslbis3 = NULL;
> -    g_autofree CDATDsemts *dsemts = NULL;
> +    CDATDsmas *dsmas;
> +    CDATDslbis *dslbis0;
> +    CDATDslbis *dslbis1;
> +    CDATDslbis *dslbis2;
> +    CDATDslbis *dslbis3;
> +    CDATDsemts *dsemts;
>  
>      dsmas = g_malloc(sizeof(*dsmas));
>      *dsmas = (CDATDsmas) {
> @@ -135,12 +135,12 @@ static void ct3_build_cdat_entries_for_mr(CDATSubHeader **cdat_table,
>      };
>  
>      /* Header always at start of structure */
> -    cdat_table[CT3_CDAT_DSMAS] = g_steal_pointer(&dsmas);
> -    cdat_table[CT3_CDAT_DSLBIS0] = g_steal_pointer(&dslbis0);
> -    cdat_table[CT3_CDAT_DSLBIS1] = g_steal_pointer(&dslbis1);
> -    cdat_table[CT3_CDAT_DSLBIS2] = g_steal_pointer(&dslbis2);
> -    cdat_table[CT3_CDAT_DSLBIS3] = g_steal_pointer(&dslbis3);
> -    cdat_table[CT3_CDAT_DSEMTS] = g_steal_pointer(&dsemts);
> +    cdat_table[CT3_CDAT_DSMAS] = (CDATSubHeader *)dsmas;
Could do
    cdat_table[CT3_CDAT_DSMAS] = &dsmas->header;
etc
> +    cdat_table[CT3_CDAT_DSLBIS0] = (CDATSubHeader *)dslbis0;
> +    cdat_table[CT3_CDAT_DSLBIS1] = (CDATSubHeader *)dslbis1;
> +    cdat_table[CT3_CDAT_DSLBIS2] = (CDATSubHeader *)dslbis2;
> +    cdat_table[CT3_CDAT_DSLBIS3] = (CDATSubHeader *)dslbis3;
> +    cdat_table[CT3_CDAT_DSEMTS] = (CDATSubHeader *)dsemts;
>  }
>  
>  static int ct3_build_cdat_table(CDATSubHeader ***cdat_table, void *priv)



WARNING: multiple messages have this Message-ID (diff)
From: Jonathan Cameron via <qemu-devel@nongnu.org>
To: Thomas Huth <thuth@redhat.com>
Cc: <qemu-devel@nongnu.org>, Fan Ni <fan.ni@samsung.com>,
	"Michael S. Tsirkin" <mst@redhat.com>,
	Marcel Apfelbaum <marcel.apfelbaum@gmail.com>,
	<qemu-trivial@nongnu.org>
Subject: Re: [PATCH 3/3] hw/mem/cxl_type3: Fix problem with g_steal_pointer()
Date: Mon, 4 Mar 2024 15:10:37 +0000	[thread overview]
Message-ID: <20240304151037.00000f6c@Huawei.com> (raw)
In-Reply-To: <20240304104406.59855-4-thuth@redhat.com>

On Mon,  4 Mar 2024 11:44:06 +0100
Thomas Huth <thuth@redhat.com> wrote:

> When setting GLIB_VERSION_MAX_ALLOWED to GLIB_VERSION_2_58 or higher,
> glib adds type safety checks to the g_steal_pointer() macro. This
> triggers errors in the ct3_build_cdat_entries_for_mr() function which
> uses the g_steal_pointer() for type-casting from one pointer type to
> the other (which also looks quite weird since the local pointers have
> all been declared with g_autofree though they are never freed here).
> Fix it by using a proper typecast instead. For making this possible, we
> have to remove the QEMU_PACKED attribute from some structs since GCC
> otherwise complains that the source and destination pointer might
> have different alignment restrictions. Removing the QEMU_PACKED should
> be fine here since the structs are already naturally aligned. Anyway,
> add some QEMU_BUILD_BUG_ON() statements to make sure that we've got
> the right sizes (without padding in the structs).

I missed these as well when getting rid of the false handling
of failure of g_new0 calls.

Another alternative would be to point to the head structures rather
than the containing structure - would avoid need to cast.
That might be neater?  Should I think also remove the alignment
question?


> 
> Signed-off-by: Thomas Huth <thuth@redhat.com>
> ---
>  include/hw/cxl/cxl_cdat.h |  9 ++++++---
>  hw/mem/cxl_type3.c        | 24 ++++++++++++------------
>  2 files changed, 18 insertions(+), 15 deletions(-)
> 
> diff --git a/include/hw/cxl/cxl_cdat.h b/include/hw/cxl/cxl_cdat.h
> index b44cefaad6..17a09066dc 100644
> --- a/include/hw/cxl/cxl_cdat.h
> +++ b/include/hw/cxl/cxl_cdat.h
> @@ -82,7 +82,8 @@ typedef struct CDATDsmas {
>      uint16_t reserved;
>      uint64_t DPA_base;
>      uint64_t DPA_length;
> -} QEMU_PACKED CDATDsmas;
> +} CDATDsmas;
> +QEMU_BUILD_BUG_ON(sizeof(CDATDsmas) != 24);
>  
>  /* Device Scoped Latency and Bandwidth Information Structure - CDAT Table 5 */
>  typedef struct CDATDslbis {
> @@ -95,7 +96,8 @@ typedef struct CDATDslbis {
>      uint64_t entry_base_unit;
>      uint16_t entry[3];
>      uint16_t reserved2;
> -} QEMU_PACKED CDATDslbis;
> +} CDATDslbis;
> +QEMU_BUILD_BUG_ON(sizeof(CDATDslbis) != 24);
>  
>  /* Device Scoped Memory Side Cache Information Structure - CDAT Table 6 */
>  typedef struct CDATDsmscis {
> @@ -122,7 +124,8 @@ typedef struct CDATDsemts {
>      uint16_t reserved;
>      uint64_t DPA_offset;
>      uint64_t DPA_length;
> -} QEMU_PACKED CDATDsemts;
> +} CDATDsemts;
> +QEMU_BUILD_BUG_ON(sizeof(CDATDsemts) != 24);
>  
>  /* Switch Scoped Latency and Bandwidth Information Structure - CDAT Table 9 */
>  typedef struct CDATSslbisHeader {
> diff --git a/hw/mem/cxl_type3.c b/hw/mem/cxl_type3.c
> index e8801805b9..b679dfae1c 100644
> --- a/hw/mem/cxl_type3.c
> +++ b/hw/mem/cxl_type3.c
> @@ -46,12 +46,12 @@ static void ct3_build_cdat_entries_for_mr(CDATSubHeader **cdat_table,
>                                            int dsmad_handle, MemoryRegion *mr,
>                                            bool is_pmem, uint64_t dpa_base)
>  {
> -    g_autofree CDATDsmas *dsmas = NULL;
> -    g_autofree CDATDslbis *dslbis0 = NULL;
> -    g_autofree CDATDslbis *dslbis1 = NULL;
> -    g_autofree CDATDslbis *dslbis2 = NULL;
> -    g_autofree CDATDslbis *dslbis3 = NULL;
> -    g_autofree CDATDsemts *dsemts = NULL;
> +    CDATDsmas *dsmas;
> +    CDATDslbis *dslbis0;
> +    CDATDslbis *dslbis1;
> +    CDATDslbis *dslbis2;
> +    CDATDslbis *dslbis3;
> +    CDATDsemts *dsemts;
>  
>      dsmas = g_malloc(sizeof(*dsmas));
>      *dsmas = (CDATDsmas) {
> @@ -135,12 +135,12 @@ static void ct3_build_cdat_entries_for_mr(CDATSubHeader **cdat_table,
>      };
>  
>      /* Header always at start of structure */
> -    cdat_table[CT3_CDAT_DSMAS] = g_steal_pointer(&dsmas);
> -    cdat_table[CT3_CDAT_DSLBIS0] = g_steal_pointer(&dslbis0);
> -    cdat_table[CT3_CDAT_DSLBIS1] = g_steal_pointer(&dslbis1);
> -    cdat_table[CT3_CDAT_DSLBIS2] = g_steal_pointer(&dslbis2);
> -    cdat_table[CT3_CDAT_DSLBIS3] = g_steal_pointer(&dslbis3);
> -    cdat_table[CT3_CDAT_DSEMTS] = g_steal_pointer(&dsemts);
> +    cdat_table[CT3_CDAT_DSMAS] = (CDATSubHeader *)dsmas;
Could do
    cdat_table[CT3_CDAT_DSMAS] = &dsmas->header;
etc
> +    cdat_table[CT3_CDAT_DSLBIS0] = (CDATSubHeader *)dslbis0;
> +    cdat_table[CT3_CDAT_DSLBIS1] = (CDATSubHeader *)dslbis1;
> +    cdat_table[CT3_CDAT_DSLBIS2] = (CDATSubHeader *)dslbis2;
> +    cdat_table[CT3_CDAT_DSLBIS3] = (CDATSubHeader *)dslbis3;
> +    cdat_table[CT3_CDAT_DSEMTS] = (CDATSubHeader *)dsemts;
>  }
>  
>  static int ct3_build_cdat_table(CDATSubHeader ***cdat_table, void *priv)



  reply	other threads:[~2024-03-04 15:10 UTC|newest]

Thread overview: 16+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-03-04 10:44 [PATCH 0/3] cxl: Fix issues with g_steal_pointer() Thomas Huth
2024-03-04 10:44 ` [PATCH 1/3] hw/cxl/cxl-cdat: Fix type of buf in ct3_load_cdat() Thomas Huth
2024-03-04 14:57   ` Jonathan Cameron
2024-03-04 14:57     ` Jonathan Cameron via
2024-03-04 10:44 ` [PATCH 2/3] hw/pci-bridge/cxl_upstream: Fix problem with g_steal_pointer() Thomas Huth
2024-03-04 15:06   ` Jonathan Cameron
2024-03-04 15:06     ` Jonathan Cameron via
2024-03-04 15:12     ` Jonathan Cameron
2024-03-04 15:12       ` Jonathan Cameron via
2024-03-04 10:44 ` [PATCH 3/3] hw/mem/cxl_type3: " Thomas Huth
2024-03-04 15:10   ` Jonathan Cameron [this message]
2024-03-04 15:10     ` Jonathan Cameron via
2024-03-05  7:27     ` Thomas Huth
2024-03-05 15:52       ` Jonathan Cameron
2024-03-05 15:52         ` Jonathan Cameron via
2024-03-08  6:14 ` [PATCH 0/3] cxl: Fix issues " Michael Tokarev

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20240304151037.00000f6c@Huawei.com \
    --to=jonathan.cameron@huawei.com \
    --cc=fan.ni@samsung.com \
    --cc=marcel.apfelbaum@gmail.com \
    --cc=mst@redhat.com \
    --cc=qemu-devel@nongnu.org \
    --cc=qemu-trivial@nongnu.org \
    --cc=thuth@redhat.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.